1. Introduction
Artificial Intelligence (AI) revenues in insurance are expected to grow 23% to $3.4 billion between 2019–2024, yet the suitability of black-box AI models in insurance practices remains questionable (Bean 2021; Chen et al. 2019; GlobalData 2021). The growth of AI as an intelligent decision-making methodology that can perform complex computational tasks is revolutionising financial services, particularly within insurance practices. Data and its potential use are seen as a primary strategic asset and a source of competitive advantage in financial services firms, with AI models’ leverage of such data providing numerous advantages (Kim and Gardner 2015). Such advantages of AI use in the insurance industry include enhanced fraud detection in claims management, granularity and personalisation when pricing insurance premiums, the creation of smart contracts, analysis of legal documents, virtual assistants (chatbots) and office operations (EIOPA 2021; Eling et al. 2021; McFall et al. 2020; Ngai et al. 2011; OECD 2020; Riikkinen et al. 2018; Zarifis et al. 2019). AI encompasses the collation of multiple technologies in a single system which enables machines to interpret data and aid complex computational decision-making (Chi et al. 2020). Although AI models’ advantages abound, recent literature highlights the AI models’ opacity which is coined as black-box thinking (Adadi and Berrada 2018; Carabantes 2020). The Insurance Value Chain (IVC) makes extensive use of AI methods at every stage of the value creation process, with AI particularly impactful in claims management and underwriting and pricing departments (Eling et al. 2021). This research systematically reviews all peer-reviewed applications of (X)AI in insurance between 2000 and 2021 with a critical focus on explainability of the models. This is the first study to investigate XAI in an applied, insurance industry context.
The rationale for Explainable Artificial Intelligence (XAI) development is primarily driven by three main reasons: (i) demand for the production of more transparent models, (ii) necessity of techniques to allow for humans to interact with them, and (iii) trustworthy inferences from such transparent models (Došilović et al. 2018; Fox et al. 2017; Mullins et al. 2021). Decision-makers require an explanation of the AI system to aid in their understanding of their decision-making processes (Biran and Cotton 2017; Hoffman et al. 2018). Throughout this systematic review, AI is defined using recent recommendations by AI experts: “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real virtual environments such AI systems are designed to operate with varying levels of autonomy” (Krafft et al. 2020). As an extension of AI models, XAI involves enhancing current AI models by developing their transparency, interpretability and explicability, with such AI advancements ultimately aiming to make AI models more understandable to humans (Adadi and Berrada 2018; Floridi et al. 2018). By presenting an analysis of insurance’s AI applications’ degree of explainability, the reader gleans an insight into the progress made to-date in insurance practice and research to satisfy the want for transparency and explanations of AI-driven decisions. Practically, end-insurance consumers affected by AI-enhanced decisions will be less likely to trust in the decisions made by machines when they do not trust and understand the AI processes involved (Burrell 2016; Ribeiro et al. 2016).
Insurance’s influence on socio-economic development cannot be understated, with the sound development of national insurance markets allowing for the promotion of financial stability, improved welfare and business innovation (Ferguson 2008; Ungur 2017). Insurance affordability is a key determinant of societal progress, with the modelling of insurance pricing practices playing a key role in this affordability (Daniels 2011), with actuarially fair pricing of insurance premiums allowing for a population to access insurance at rates which they can reasonably afford (Grant 2012). Transparency and explainability of AI models are core requirements to achieve impactful trustworthy AI in society (Felzmann et al. 2019; Maynard et al. 2022; Moradi and Samwald 2021). Trustworthiness is a core concept within the insurance industry, with enhanced XAI explanations directly affecting trust levels amongst insurance companies and their stakeholders.
This paper is structured as follows; Section 2 presents related works of this review while analysing current research on XAI’s definition and related taxonomies, also outlining related work on (X)AI’s impact on the IVC. Section 3 presents the methodological system to collect and analyse relevant literature on (X)AI use along the IVC. The search technique to arrive at relevant articles is especially emphasised to ensure the validity of eventual research results and allow for future research reproducibility. Section 4 outlines the review’s findings of the systematically chosen sample of literature and their AI methods through the lens of defined XAI criteria. Section 5 presents a novel discussion of the review’s results on the prevalence of XAI along the IVC, focusing on the extent to which AI applications along the IVC are explainable. Section 6 concludes the systematic review, reiterating points of interest regarding the future of XAI applications in insurance practices.
XAI Terminology
Kelley et al. (2018) define AI as “a computer system that can sense its environment, comprehend, learn, and take action from what it’s learning”, with XAI intuitively expanding on this description by allowing humans to be present at every stage of this AI decision-making lifespan. A common misnomer of AI models’ explainability is that it is simply the improvement of trust in AI systems and their decision processes, through developing “causal structures in observational data” (Goodman and Flaxman 2017; Lipton 2018). Models’ explainability enhances the interpretability, i.e., understanding how a model came to a certain decision (Lou et al. 2013), while also positively impacting fair and ethical decision making for high-computational tasks (Srihari 2020). Table 1 outlines the XAI variables and categories used within the systematic review to analyse the degree of explainability present in AI methods applied within the insurance industry. Additionally, the following categories of XAI methods are used to classify published applications of AI in insurance: (1) Feature Interaction and Importance, (2) Attention Mechanism, (3) Dimensionality Reduction, (4) Knowledge Distillation and Rule Extraction, and (5) Intrinsically Interpretable Models1. Additional categorisations and terminology determinations are summarised in Clinciu and Hastie (2019) and Arrieta et al. (2020).
2. Fundamental Concepts & Background
2.1. Artificial Intelligence Applications in Insurance
AI use abounds across the entirety of the IVC with Eling et al. (2021) and EIOPA (2021) providing a thorough examination of the six main stages of the IVC and their goals. Tekaya et al. (2020) preface AI research in financial services by offering an overview of current use-cases and advantages of implementing Big Data and AI models in banking, credit risk management, fraud detection and the insurance industry. Several other articles highlight the importance and advantages of AI applications in the insurance industry, predicting major shifts in operations in the coming years (Paruchuri 2020; Riikkinen et al. 2018; Umamaheswari and Janakiraman 2014). Popular areas within insurance research where AI has been applied include fraud detection (Sithic and Balasubramanian 2013; Verma et al. 2017) and claims reserving (Baudry and Robert 2019; Blier-Wong et al. 2021; Lopez and Milhaud 2021; Wüthrich 2018). Grize et al. (2020) focus on ML applications in non-life insurance, highlighting AI’s positive impact on risk assessment to improve the insurance companies’ overall profitability in the long run.
Fang et al. (2016) used Big Data to develop a new profitability method for insurers using historical customer data, where they found that the Random Forest (RF) model outperformed other methods of forecasting (linear regression and SVM). Shapiro (2007) documents the extent to which fuzzy logic (FL) has been applied to insurance practices, which prompted Baser and Apaydin (2010)’s later research on claims reserving using hybrid fuzzy least squares regression and Khuong and Tuan (2016)’s creation of a neuro-fuzzy inference system for insurance forecasting. NallamReddy et al. (2014) present a robust review of clustering techniques used in insurance. Quan and Valdez (2018) use another understandable and transparent AI method, Decision Trees (DT), to investigate their use in insurance claims prediction. Interestingly, later research acknowledges the low predictive power of DTs and boosts their intrinsic interpretability to provide a more robust insurance pricing model (Henckaerts et al. 2021).
Sarkar (2020) argues that the insurance industry holds the potential for algorithmic capabilities to enhance each stage of the industry’s value chain. Through highlighting AI’s offerings at each stage of the IVC, the research prompted further studies from Walsh and Taylor (2020) and Eling et al. (2021) to determine precise AI opportunities available to the insurance industry. Walsh and Taylor (2020) highlight AI models’ ability to mimic, or augment, human capabilities with NLP, Internet of Things (IoT) and computer vision. Eling et al. (2021) analyse AI’s impact at each step on IVC and specifically highlights the potential for AI to enhance revenue streams, loss prediction and loss prevention measures for insurance practitioners.
Bias inherent to black-box AI systems threatens trust within the insurance industry, with this bias primarily driven by either humans’ input or algorithmic bias (Koster 2020; Ntoutsi et al. 2020). There is potential for these models’ impediments to compound and extenuate bias in their decision-making processes with unfair outcomes possible within the insurance industry (Confalonieri et al. 2021; Koster et al. 2021). This issue of bias is further aggravated when the lacking transparency in systems makes it difficult to dispute or appeal a biased decision by AI algorithms (von Eschenbach 2021). Bias in AI models could potentially lead to discriminatory behaviour of the AI system, caused by the model’s tendency to use sensitive information resulting in unfair decisions (Barocas and Selbst 2016). There is strong research conducted on the determination of responsible AI, with Koster et al. (2021) providing a framework to create a responsible AI system, and Arrieta et al. (2020) outlining degrees of fairness to be implemented in an AI system to reduce discriminatory issues. Although a thorough examination of trust as it pertains to social sciences, leading into its importance in human-AI relationships, is beyond the scope of the current review, trust in AI systems is considered critical for the sustained use of AI technologies in insurance (Mayer et al. 1995; Siau and Wang 2018). Toreini et al. (2020) propose a Chain of Trust framework to further enhance users’ trust in AI and ML technologies, while research on explanations in AI’s use in medical diagnostic settings proves advantageous for clinician’s trust and understanding in these technologies (Diprose et al. 2020; Tonekaboni et al. 2019). Jacovi et al. (2021) outline that the agreement between a human and AI system is contractual, therefore the interaction between a human and AI system must be explicit for trust to be present in the relationship between both parties (Hawley 2014; Tallant 2017). Trust derived from explanations in AI systems is enhanced by the provision of explanations and understandability, supporting the growth of XAI demand within the insurance industry.
2.2. Explainable Artificial Intelligence
XAI’s recent history is firmly rooted in the field of AI, with contributions of explainability and transparency paving the way for XAI’s growth. Lundberg and Lee (2017) cited explainability as the “interpretable approximation of the original complex [AI] model”, while later Al-Shedivat et al. (2020) reference explainability as a “local approximation of a complex model (by another model)”. What is clear from the increased research focus on AI in the late 2010’s is that the notion of explainability did not drastically mature—research continues to ask the same questions pertaining to AI. Such issues include the fairness of an AI system, the transparency of decision pathways, and the explanation to be provided to the end user. A further important consideration is that XAI is merely the process of making AI understandable to humans, including its actions, recommendations and underlying decisions (Anjomshoae et al. 2019). Neither AI, nor XAI, are on the cusp of machine-led moral decisions or understanding (Ford 2018). Humans are still at the core of (X)AI, with bias and fairness central issues to contend with. This section outlines current research in XAI and its impact on the research field of AI.
The evaluation of the insurance industry’s (X)AI applications’ explainability contributes to the interdisciplinary literature on XAI. Through presenting the current discussion and taxonomies of XAI in the literature, the authors highlight the necessity of defined XAI criteria and categories in line with those used in this paper’s analysis. Gade et al. (2019) outline the main challenges for XAI researchers which include (1) ‘defining model explainability’, (2) ‘formulating explainability tasks for understanding model behaviour and developing solutions for these tasks’, and (3) ‘designing measures for evaluating the performance of models in explainability tasks’. Vilone and Longo (2020)’s later systematic study contributed a classification system for published XAI literature, aiming to establish boundaries in the field of XAI research. Four main clusters of research were found by Vilone and Longo (2020); (1) ‘reviews focused on specific aspects of XAI’, (2) ‘the theories and notions related to the concept of explainability’, (3) ‘the methods aimed at explaining the inferential process of data-driven and knowledge-based modelling approaches’, and 4) ‘the ways to evaluate the methods for explainability’.
Extending on the above, the literature on XAI is attempting to determine a sound definition of XAI, which is commonly referred to as ‘explainability’ rather than ‘interpretability’. Islam et al. (2020) note that explainability is more than interpretability in terms of importance and trust in the prediction. Interpretability is often the end goal with explanations acting as tools to reach interpretability (Honegger 2018). Additionally, the General Data Protection Regulations (GDPR) (EU 2016) which is discussed later in this paper covers only explainability (Došilović et al. 2018). These considerations encourage the authors to focus on the need for a domain-specific definition of XAI relevant to insurance practices. Instead of offering actionable definitions of XAI, other works classify the requirements that an explainable system should meet (Lipton 2018; Xie et al. 2020) or the methods of evaluations underhich an AI system can be deemed explainable (Doshi-Velez and Kim 2017; Hoffman et al. 2018; Lipton 2018; Rosenfeld 2021).
Reviews of XAI in medicine ignited the XAI research field, with many studies on the technology’s effects on disease diagnosis, classification and treatment published in recent years. Payrovnaziri et al. (2020) involved the review of 49 articles published in the period 2009–2019 to group XAI methods used in the medical field. In this study, Payrovnaziri et al. (2020) grouped XAI methods into 5 different groups: (1) ‘Knowledge Distillation and Rule Extraction’, (2) ‘Intrinsically Interpretable Models’, (3) ‘Data Dimensionality Reduction’, (4) ‘Attention Mechanism’ and (5) ‘Feature Interaction and Importance’. Antoniadi et al. (2021) outline challenges pertaining to AI’s use for clinical decision support systems, emphasising lacking transparency as a key issue. Notwithstanding the obvious advantages of XAI methods to enhance understandability and aid medical practitioners’ decisions which abound, their research finds a distinct lack of XAI applications in medicine.
Finance-related studies on XAI include Demajo et al. (2020); Hadji Misheva et al. (2021) and Biecek et al. (2021)’s research on credit scoring and risk management. Similarly, Bussmann et al. (2020) explore XAI in fintech risk management and peer-to-peer lending platforms, while Kute et al. (2021) also focus on risk management in finance applications through their review of DL and XAI technologies in the identification of suspicious money laundering practices. Gramegna and Giudici (2020) analyse XAI’s potential to identify policyholders’ reasons for buying or abandoning non-life insurance coverage. The grouping and assessment of like-minded policyholders allows for additional high-quality information on policyholders to be obtained, with transparent and accessible AI models used. Adadi and Berrada (2018) provide a foundational background to the main concepts and implications of an XAI system, citing data security and fair lending in financial services as key issues surrounding XAI use in financial services. Concerning banking and accounting practices, Burgt (2020) states that trust in AI systems in the banking industry is paramount and provides a discussion on the trade-off between explainability and predictability of AI systems. Gramespacher and Posth (2021) then utilise XAI to optimise the return target function of a loan portfolio, while Mehdiyev et al. (2021) add to the conversation by analysing tax auditing practices and public administration’s appetite for XAI. Albeit the obvious advantages of developing transparent decision-making systems in public administration, this research cites the requirements of safe, reliable, and trustworthy AI systems as creating additional complexity in AI systems which take some time to implement widely. The interest in human-centred decision-making machines reaches beyond medical and finance domains. Putnam and Conati (2019) provide a survey that finds students seek additional explanations from their Intelligent Tutoring System to aid their education prospects. Natural Language Processing (NLP) is another research area with significant interest in XAI methods as revealed by Danilevsky et al. (2020) with sarcasm detection in dialogues later reviewed by Kumar et al. (2021). Anjomshoae et al. (2019) reviews inter-robot explainability and addresses the issue of explainability to non-users of ML robots through personalisation and context awareness.
The current systematic review builds upon previous research on XAI methods’ classification and analysis of XAI literature during the systematic selection of literature. Although the above literature does provide a brief overview of the current understanding of XAI and related key concerns highlighted in the literature, this is the first paper to review XAI applications in the insurance industry.
2.3. The Importance of Explainability in Insurance Analytics
The personal data of EU citizens is described as a fundamental right by the EU Charter of Fundamental Rights and has been addressed since 1995 by the Data Protection Directive (Taylor 2017; Yeung et al. 2019). Citizen rights to privacy are operationalised through a number of data governance mechanisms ranging from consent platforms and data management systems which produce compliance measures to the control, use and lifespan of personal data. Accordingly, the data regulation environment is one of the most robust and sophisticated that is built on a strategy to both empower citizens to engage with the digital world and also to inform and guide commercial use of personal data. Data is protected by several regulatory instruments that provide a specific response to data use. These range from the data governance and the digital markets act to the GDPR (Andrew and Baker 2021; Goddard 2017). The range of different instruments speaks to the complexity of data use and data commercialisation scenarios. Insurance analytics often concerns the use of citizen and customer data to provide value to both the insureds and the insurance business model. Insurance analytics already uses personal data to optimise front- and back-end operations, risk modelling and risk pricing (Hollis and Strauss 2007; Keller et al. 2018; Ma et al. 2018; Mizgier et al. 2018; Naylor 2017). Furthermore, insurance analytics can provide important value in fraud management, claims management and better managing risk pooling by creating more accurate behavioural profiles of insureds (Barry and Charpentier 2020; Cevolini and Esposito 2020; Tanninen 2020). The commercial promise of insurance analytics also raises questions and concerns regarding the potential harms of undermining the core social solidarity of insurance by changing the pricing structure and limiting access to insurance products and services to those that meet stricter parameters of risk pricing. The importance of access to insurance is evident in compulsory products such as motor and, in some states, life insurance. Health insurance and insurance analytics are becoming a more controversial issue as increased reliance on private health care in parallel with increased use of insurance analytics are highlighting the tension between affordability and welfare. In short, insurance analytics offers scalable optimisation and high-value commercial solutions to IVCs and business models. Still, EU regulation is seeking to govern the use by the steering industry to more equitable, transparent and explainable (Kuo and Lupton 2020) uses of data analytics (EIOPA 2021; Mullins et al. 2021; van den Boom 2021).
3. Methodology
3.1. Literature Search Strategy
This literature search plan and related inclusion and exclusion criteria build upon the framework applied within Eling et al. (2021), with the aim of expanding upon their research to assess the prevalence of XAI methods in the IVC’s AI applications. Eling et al. (2021)’s research assessed AI’s impact on the IVC and the insurability of risks. The research presented in this paper expands on the abovementioned research to determine not only the impact on the IVC of AI systems being used, but also their degree of explainability. This framework is a suitable addition to the current study as a guide to literature inclusion criteria: inclusion of AI literature concerned with different stages along the IVC.
Analysis was conducted on a systematically selected body of literature from the following databases: EBSCOhost (Business Source Complete and EconLit), ACM Digital Library2, Scopus, Web of Science and IEEE Xplore. These databases were chosen due to their wide breadth of content spanning both insurance and finance-related research, while also accounting for computer science journals to access research on AI applications. The above databases were chosen to feasibly and approximately align the current review with Eling et al. (2021)’s research, while considering database accessibility limitations.
Table 2 outlines the key search terms used interchangeably with AI in the abovementioned databases, alongside ‘Insurance’ OR ‘Insurer’ using Boolean terminology. This broad set of search terms ensures an all-encompassing article-base of the IVC’s use of AI and are adapted from Eling et al. (2021)’s literature search method.
Figure 1 outlines the systematic literature search process where an initial 419 articles were scanned for relevancy to this paper. Key relevancy criteria included the assessment of articles’ contents concerning their place along the IVC. The IVC stages are extensively outlined in Table 3, as adapted from both Eling et al. (2021) and EIOPA (2021)’s research. The articles included in the systematic study of XAI in insurance are categorised according to the specific stage of the IVC which they refer to. This categorisation allows for further assessment of XAI use within the entire IVC process.
In addition to the above, the articles’ relevancy was filtered using the following criteria set:
Time Period: Articles3 published between 1 January 2000–31 December 2021 are included,
Relevancy: The presence of keywords (Table 2) in the abstract is necessary for the article’s inclusion. Additionally, the articles need to be relevant to the assessment of AI applications along the IVC directly (e.g., articles concerned with determining drivers’ behaviour using telematics information, which may later inform insurance companies’ pricing practices were excluded, as well as generalised surveys on AI uses in insurance4),
Singularity: Duplicate articles found across the various databases are excluded,
Accessibility: Only peer-reviewed articles that are accessible through the aforementioned databases and are accessible in full text are included (i.e., extended abstracts are not included),
Language: Only articles published in English are included.
Articles published before 2000 are not included in the current review due to the increased understanding of AI from 2000 onwards (Liao et al. 2012), and the creation of the European GDPR in 2016 (implemented in the European Union in 2018) which is especially applicable to conversations on future XAI regulation.
The initial screening process included the assessment of 419 articles (following duplicate removal) based on their title, source, and abstract for the presence of the key search terms. In all, 66 articles were included for final review at this stage of the literature search. A backward search of the relevant articles (n = 66) was then conducted, which identified a further 37 articles. The backward search is a popular method of rigorous literature searching within systematic reviews in a range of disciplines including medicine (Mohamadloo et al. 2017), law (Siegel et al. 2021) and finance (Eckert and Hüsig 2021). The backward search entailed the assessment of the 66 relevant articles’ bibliographies for additional articles of relevance to the current review. Based on this rigorous selection process, a total of 103 articles were identified as relevant for the current study (Reference Appendix A for the complete database of articles meeting the relevance threshold for inclusion in this systematic review). Figure 2 provides a breakdown of the publication year dispersion of each of these 103 articles. These results are comprised of ~75% journal articles (n = 77 and ~25% conference papers/proceedings (n = 26). The PRISMA flow diagram depicts the systematic review process (Figure 3). The PRISMA statement enhances transparency of systematic reviews, to ensure the research conducted during the course of a systematic review is robust and reliable (Page and Moher 2017). Each stage of the literature search for the systematic review is highlighted within the PRISMA diagram (Figure 3).
3.2. Literature Extraction Process
The evaluation of the full-text articles is sub sectioned into two distinct phases in line with both core contributions of this review. Initially, the articles’ applied AI method was distinguished, alongside the prediction task(s) of this AI method. Secondly, the degree of explainability of the AI method employed is analysed. Here, the degree of explainability is evident in the XAI criteria applicable to each AI method employed in each article.
The criteria used in evaluating the AI methods’ degree of explainability (Table 4) are adapted from Payrovnaziri et al. (2020)’s systematic review methodology and modified to suit this review on the insurance industry. The inclusion of the XAI variables and criteria is supported by previous research in XAI, with the criteria synthesised from Mueller et al. (2019); Du et al. (2019); Carvalho et al. (2019) and Payrovnaziri et al. (2020).
3.3. Limitations of the Research
Limitations of the current review are outlined to ensure the validity and reliable reproducibility of results. In particular, the authors are unable to access 18 references which Eling et al. (2021) presented following their literature search process, while the industry reports reviewed within the same article are not included in the current systematic review. The lack of industry reports’ analysis in this paper leads to an absence of articles concerning the Support Activities stage on the IVC. In Eling et al. (2021)’s research, all articles found pertaining to insurance companies’ Support Activities were industry reports.
Industry reports were not included in this paper as access to articles with complete methodological processes outlined is pertinent to the current systematic review, a section which industry reports regularly omit in their publications. The inclusion of academic articles and conference articles ensures the methods of AI integration in each of the reviewed articles is outlined, in particular a coherent methodology discussion which can be assessed using the XAI criteria outlined in this paper.
The authors note the limitations of Payrovnaziri et al. (2020)’s research framework pertaining to XAI literature. In particular, the XAI categorisations presented feature some overlap across various XAI categories. For example, attention mechanism targets feature attribution, a category which is also covered under the feature interaction and importance categorisation. Nevertheless, this framework provides optimal categorisations for the scope of this work to assess the degree of explainability within AI applications in insurance, as defined boundaries of each XAI categorisation is provided.
4. Systematic Review Results
4.1. AI Methods and Prediction Tasks
The systematically chosen articles are first assessed based on the AI method employed and associated prediction task, with a focus on then distinguishing the degree of explainability evident in the literature. The stage of the IVC each article refers to is also clarified in the systematic research findings. Research on AI’s use along the IVC over the twenty-one-year period of this review revealed AI is popular at every stage of the IVC, except for insurance companies’ Support Activities. Such activities include general HR, IT and Public Relations departments in insurance companies. As mentioned above, a viable reason for the lack of articles concerned with this stage of the IVC is that Eling et al. (2021)’s study found articles on this subject through their review of industry reports, which the present systematic review did not include in the systematic review. The Underwriting and Pricing stage reveals significant research results (40%), with Claim Management (34%) also making extensive use of AI methods for fraud management and identification in particular.
Table 5 lists all the articles alongside the AI method employed and prediction task. A range of AI methods are used in the articles including; (1) Ensemble, (2) Neural Network (NN), (3) Clustering, (4) Regression (Linear and Logistic), (5) Fuzzy Logic, (6) Bayesian Network (BN), (7) Decision Tree, (8) Support Vector Machine (SVM). Other methods used include Instance- and Rule-based, Regularisation and Reinforcement Learning. The most popular AI method used is Ensemble (23%), with both NNs (20%) and Clustering (14%) also proving popular.
The line of insurance business the research in each article refers to is also classified, with non-life insurance lines returning a high number of articles in the systematic review (55%). Motor insurance prediction problems are popular areas of research, including driving behaviour classification and automobile insurance fraud (44%). Articles concerning insurers’ life-business shows health(care) insurance as a popular area of research (13%), with health insurance fraud prevention and the classification of health insureds the most prominent research areas.
4.2. XAI Categories along the IVC
The following categories of XAI methods are highlighted within the article database; (1) Feature Interaction and Importance, (2) Attention Mechanism, (3) Dimensionality Reduction, (4) Knowledge Distillation and Rule Extraction, and (5) Intrinsically Interpretable Models. Figure 4 shows each stage on the IVC and the corresponding XAI method employed in the reviewed articles. The XAI methods’ interpretability techniques are then categorised into (1) intrinsic or post hoc, (2) local or global and (3) model-specific or model-agnostic (Table 6). According to the reviewed articles, most of the research on AI applications in insurance is concerned with Knowledge Discovery and Distillation, which is also grouped with Rule Extraction (35%) XAI methods for the purpose of the current review.
4.3. Feature Interaction and Importance
Analysing (X)AI models’ input features’ importance and interaction is a popular XAI method, with ~27% of reviewed articles utilising this method. The determination of features’ importance contributed to the development of thorough XAI methods to complete many prediction tasks at each stage on the IVC. Smith et al. (2000) utilise Artificial Neural Networks (ANN) to gain an insight into customer policies which were likely to renew or terminate at the close of the policy period through analysing those factors which contribute to policy termination. This assessment of optimal premium pricing through data mining and ML methods instructs research on insurance customer retention and profitability. Additionally, addressing customer retention is Larivière and Van den Poel (2005)’s research which explored three predictor variables which encompass potential explanatory variables to inform insurance customer retention. Their RF model provides an importance measure between the explanatory and dependence variables for the prediction task.
Claim management and insurance fraud detection are areas which benefit from analysing the interaction and importance of feature inputs in AI applications through the isolation of important features which contribute to fraud (Belhadji et al. 2000). Similarly, Tao et al. (2012) avoid the curse of dimensionality through using the kernel function for SVMs in their XAI approach for insurance fraud identification, while Supraja and Saritha (2017) use this XAI method to ready their data for automobile fraud detection using fuzzy rule-based predictive techniques.
Feature interaction and importance is also useful in assessing risk across a wide range of insurance activities and informing underwriting and pricing of premiums. Biddle et al. (2018) add to literature on automated underwriting in life insurance applications using the XAI method of Feature Interaction and Importance. Recursive Feature Elimination is used to reduce the feature space through iteratively wrapping and training a classifier on several feature subsets and then providing feature rankings for each subset. Premium pricing of automobile insurance is researched by Yeo et al. (2002)’s, where cluster grouping of policyholders according to relative features aids in determining the price sensitivity of policyholder groups to premium prices.
4.4. Attention Mechanism
The Attention Mechanism within an AI model primarily attempts to find a set of positions in a sequence with the most relevant information on a prediction task (Payrovnaziri et al. 2020), which in turn enhances interpretability, according to Mascharka et al. (2018).
In line with the current review, Attention Mechanism is used to compute the weight of claim occurrences to inform fraud detection (Viaene et al. 2004) and inform insurer insolvency prediction (Ibiwoye et al. 2012). Lin and Chang (2009) apply Attention Mechanism in their determination of premium rates of ‘in-between’ risks through weight classification of different tariff classes. The method also aids in the determination of litigation risk of liability insurance within the accountancy profession, as Sevim et al. (2016) incorporate Attention Mechanism in their development of an ANN model, while Deprez et al. (2017) apply Attention Mechanism to mortality modelling through back-testing parametric mortality models. Samonte et al. (2018) use this XAI method for automatic document classification of medical record notes using NLP. The enhancement of the Hierarchical Attention Network model (EnHAN) assigns topics for each word in a given text and learns topical word embedding in a hierarchical manner. Topical word embedding models solve the multi-label, multi-class classification problem within medical records to inform cluster processes for billing and insurance claims.
Wei and Dan (2019) apply Attention Mechanism to parameter optimisation of SVM features, while Zhang and Kong (2020) also optimised parameters for input in NB model to inform insurance product recommendations. In terms of sequence generation, this XAI method was used by Matloob et al. (2020) to inform their predictive model for fraudulent behaviour in health insurance.
4.5. Dimensionality Reduction
Researchers typically use dimensionality reduction techniques in order to reduce the set of features inputted in the model principally to improve a model’s efficiency (Motoda and Liu 2002). Kumar et al. (2010), for instance, use a frequency-based feature selection technique to reduce the dataset dimensions. This action aided in developing a model for error prevention in health insurance claims processing through reducing data storage requirements and improved model execution time. They found that using a lower frequency threshold and limiting the input feature improved the predictive accuracy. Finding similar results in terms of improved predictive accuracy, Li et al. (2018) use Principal Component Analysis (PCA) to increase the diversity of each of the 100 trees used in a RF model. This action improves the overall accuracy of the algorithm. In this instance, PCA transforms the data at each node to another space when computing the best split at that node which contributed to satisfactory feature selection in the development of the RF algorithm for fraud detection. PCA is also used in Underwriting and Pricing of life insurance through model development for risk assessment of life insurance customers (Boodhun and Jayabalan 2018).
For the popular prediction tasks related to automobile insurance, the reduction in dataset dimensionality is also useful. Liu et al. (2014) reduce their large claim frequency prediction to a multi-class prediction problem to aid the eventual implementation of Adaptive Boosting (AdaBoost) to automobile insurance data. The act of reducing the number of frequency classes contributes to AdaBoost presenting as superior to SVM, NN, DTs and GLM in terms of prediction ability and interpretability. Huang and Meng (2019) bin variables to approximate continuous variables in the dataset and construct tariff classes with high-level predictive power which enhances the model’s accuracy and predictive power in the classification of usage-based insurance (UBI) products. An ANN model is optimised in Vassiljeva et al. (2017) to inform automobile contract development through assessing drivers’ risk, while Bian et al. (2018) reduced their data dimensions to include only the five most relevant factors in determining drivers’ behaviour.
Other stages on the IVC benefit from data dimensionality reduction, with Desik et al. (2016)’s identification of relevant data clusters to inform model development of marketing strategies within different insurance product groups proving successful. The Sales and Distribution stage of the IVC uses a similar reduction of dataset features which hold no bearing on insurance customers’ likelihood of renewal (Kwak et al. 2020).
4.6. Knowledge Distillation and Rule Extraction
Knowledge Distillation and Rule Extraction components of AI models refers to the combination of large models to create a smaller, more manageable model (Hinton et al. 2015). For instance, both Cheng et al. (2020) and Jin et al. (2021) investigate optimal insurance strategies (insurance, reinsurance and investment) using the MCAM to develop adequate NN models for their respective prediction tasks. In another work concerning NNs and Knowledge Distillation XAI methods, Kiermayer and Weiß (2021) approximate representative portfolios of both term life insurance plans and Defined Contribution pension plans to aid in determining the insurer’s solvency capital requirements. These representative portfolios are inputted in a NN model, which significantly outperforms k-means clustering for insurance portfolio grouping and the evaluation of insurers’ investment surplus. The combination of models was also utilised by Xu et al. (2011) where a random rough subspace method is incorporated into a NN to aid optimised insurance fraud detection.
In terms of extracting actionable knowledge from models, Lee et al. (2020) propose a methodology for extracting variables from textual data (word similarities) to use such variables in claims analyses, thus improving actuarial modelling. Similarly, Wang and Xu (2018) apply LDA-based deep learning for the extraction of text features in claims data to detect automobile insurance fraud.
The development of association rules aids in building XAI models which are regularly understandable and useful for prediction tasks across the entirety of the IVC. Ravi et al. (2017) develop a model for analysing insurance customer complaints and categorising them for insurance customer service offices. Customer grievances are assigned an association rule which are categorised by treating grievance variables as holding a certain degree of membership with the different rules. Association rule learning is also implemented in fraud detection through the identification of frequent fraud occurrence patterns (Verma et al. 2017) and the computation of relative weights of variables related to suspicious claim activity using Adaboost AI methods (Viaene et al. 2004).
4.7. Intrinsically Interpretable Models
Aside from the interpretability techniques outlined above, other researchers have relied on the intrinsic predictive capabilities of models in their research. Through preserving the predictive capabilities of less complex AI models using boosting and optimisation techniques, the predictive power of Intrinsically Interpretable Models proves useful along the IVC.
Researchers implemented Intrinsically Interpretable Models for a range of prediction tasks including; (1) double GLMs to model insurance costs’ dispersion and mean (Smyth and Jørgensen 2002), (2) prediction of insurance losses through boosting trees (Guelman 2012), (3) prediction of insurance customers’ profitability (Fang et al. 2016), and (4) cluster identification and classification (Karamizadeh and Zolfagharifar 2016; Lin et al. 2017).
Carfora et al. (2019) identified clusters of driver behaviour to inform UBI pricing through unsupervised ML classification techniques and cluster analysis. K-means clustering is used to classify driver aggressiveness to inform a risk index of driving behaviour on different road types (primarily urban vs. highway). Benedek and László (2019) compare several interpretable AI techniques in their identification of insurance fraud indicators, which each facilitate the segmentation of such fraud indicators. DTs are highlighted as suitable AI methods for such indicator identification and classification.
5. Discussion
5.1. AI’s Application on the Insurance Value Chain
The use of AI applications at each stage on the IVC is promising, with a variety of prediction tasks fulfilled by AI applications. In line with Eling et al. (2021)’s findings, AI is disrupting the insurance industry in a number of ways. The automation of underwriting tasks and the identification and prevention of fraudulent behaviour are key areas where AI is impacting the IVC. This is in line with a survey by the Coalition Against Insurance Fraud (2020) reporting 56% of insurance companies’ surveyed AI as their primary mode of insurance fraud detection. An interesting note is the distinction between Eling et al. (2021)’s findings on AI’s use in Support Activities and the presence of XAI methods in such activities. The literature search process for this review did not result in any articles concerning XAI use in insurance Support Activities (including HR, IT, Legal and General Management). The authors accept that this finding is likely attributed to restricted keyword searches which do not consider Support Activities, opening the possibility of further research on XAI’s presence in insurance companies’ Support Activities.
5.2. XAI Definition, Evaluation and Regulatory Compliance
Research on XAI (Section 2.1) highlight the disjointed understanding of XAI both across and within industries, thus providing motivation for the current review. There appears no consistent definition of XAI in the reviewed insurance literature, a finding which is in line with Payrovnaziri et al. (2020)’s findings of XAI’s use and definition in medicine research. The main issue posed by this finding is that the evaluation of XAI methods is made increasingly difficult when there is no defined definition and scope of XAI. This review develops an XAI evaluation criteria, incorporating interpretability evaluation as either (i) intrinsic or post hoc, (ii) local or global and (iii) model-specific or model-agnostic. The results provide an extension to XAI survey research conducted by Adadi and Berrada (2018), Arrieta et al. (2020) and Das and Rad (2020) who each defined inter-related taxonomies of XAI. The development of an all-encompassing XAI definition for insurers and AI experts will allow for further adoption of XAI methods in the insurance industry.
Each definition of XAI discussed in Section 2.2 is derived from the early definition of explainability as the “assignment of causal responsibility” originally cited in Josephson and Josephson (1996). Although each paper providing additional insight into XAI definitions is useful, the lacking cohesion amongst these studies hampers the consolidation of each individual contribution into an interdisciplinarily accepted XAI definition. The authors acknowledge that an all-purpose XAI definition is difficult to determine, as both notions of explainability and interpretability (which are often used interchangeably and used in creating XAI definitions) are domain-specific notions (Freitas 2014; Rudin 2018). Lipton (2018) cites interpretability as an ill-defined concept as interpretability is not a fixed notion in and of itself. In efforts to define XAI specifically within the insurance industry, the authors accept all referenced definitions of XAI and findings of XAI use on the IVC to-date and propose the following XAI definition specific to the insurance industry:
“XAI is the transfer of understanding to AI models’ end-users by highlighting key decision- pathways in the model and allowing for human interpretability at various stages of the model’s decision-process. XAI involves outlining the relationship between model inputs and prediction, meanwhile maintaining predictive accuracy of the model throughout”
In addition to benefitting XAI research, the authors note that a solid definition of XAI pertaining directly to the insurance industry (and financial services at large) will aid the development of adapted regulation, which is in line with recommendations from Palacio et al. (2021). The GDPR (EU 2016) established a regime of “algorithmic accountability” and (insureds) “right to explanation” from decision-making algorithms (Bayamlıoğlu 2021; Wulf and Seizov 2022). XAI promotes such transparent and interpretable traits, yet a comprehensive implementation of these methods necessitates regulatory compliance (Henckaerts et al. 2020). In the momentary absence of specific regulation of XAI models, the authors highlight the potential for XAI methods to be paired with existing governance measures in the insurance industry to mitigate concerns surrounding the use of novel AI methods until satisfactory regulation is developed. This recommendation is in line with governance guidelines from EIOPA (2021), for example the maintenance of human oversight in decision-making processes.
5.3. The Relationship between Explanation and Trust
The recent proliferation of XAI literature is partly driven by the need to maintain users’ trust in AI to further develop AI adoption (Jacovi et al. 2021; Robinson 2020). Despite this rationale, prior XAI research has not considered the notion of trust in much detail. As a multidimensional and dynamic construct, a concise definition of trust has received considerable critical attention but remained elusive. The interplay between explainability and trust can be further substantiated by exploring what constitutes user trust in AI. So far, it has been established that explanations can positively affect users’ trustworthiness assessment in several use cases, such as recommendation agents (Xiao and Benbasat 2007) or information security (Pieters 2011). In particular, explanations can foster cognitive-based trust that prevails early in the human-AI relationship. This initial trust development phase is often referred to as swift trust (Meyerson et al. 1996). This notion of interpersonal trust, following the common act of anthropomorphising machines, affects how humans interact with such machines (Hoffman 2017). Users are affected by the reliability of their ‘partner’ in the interpersonal relationship (the machine), however the lack of humane empathy and ability to apologise for mistakes during automated decision-making hinders the fostering of a truly anthropomorphised machine being involved in a real interpersonal relationship with a human (Beck et al. 2002). As interaction history is lacking, the extent to which a user can understand a given process or decision is paramount (Colaner 2022). However, the question remains whether there is a threshold after which this positive effect can be reversed. If users suffered from such explanation overload, more explanations would not be significantly associated with trust. This assessment is subjective and perceptual in nature and might well be influenced by a user’s general propensity to trust AI models. This assumption accords with previous findings by McKnight et al. (2002) that the disposition to trust positively influences the trustworthiness assessment in e-commerce. Further work is thus required to examine how, precisely, the trust construct can be integrated into XAI research.
6. Conclusions
The primary contribution of this systematic review to widespread XAI understanding is an in-depth analysis of published literature on XAI in insurance practices. The growing commercialisation of AI applications leads to the potential of insurers to create high-value solutions in response to the industry’s efficiency issues and respond appropriately to changes in the business landscape (Balasubramanian et al. 2018). The necessity to highlight transparent and understandable AI processes applied within the insurance industry prompts this investigation of XAI applications and their current use cases. This review of key literature provides a comprehensive analysis of XAI applications in insurance for both key insurance regulators and insurance practitioners which will allow for extensive application in future regulatory decision-making. Legally, the opacity of black-box AI systems hinders regulatory bodies from determining whether data is processed fairly (Carabantes 2020; Rieder and Simon 2017), with XAI enhancing the potential for AI systems’ regulation under the GDPR in Europe (EU 2016).
This review assesses 103 articles (comprised of journal articles and conference papers/proceedings) which outline XAI applications at each stage of the IVC. The lack of explainability evaluation and consensus on XAI definitions hinders the potential progress of the XAI research field in insurance practices as there is no clear way to evaluate the degree of explainability in XAI. This review attempts to bridge this gap by defining XAI criteria and incorporating such criteria into a systematic review of XAI applications in insurance literature. Utilising this XAI criteria the degree of explainability in each XAI application is provided, assigning each AI method to a grouped XAI approach, and then evaluating the model’s interpretability as either (i) intrinsic or post hoc, (ii) local or global, and (iii) model-specific or model-agnostic. Findings reiterate the authors’ hypothesis that XAI methods are popular within insurance research, enabling the transparent use of AI methods in industry research. The transparency XAI methods afford insurance companies enhances the application of AI models in an industry striving for a basis of trust with multiple stakeholders.
Additionally, this paper analyses XAI definitions and proposes a revised definition of XAI. This proposed definition is informed by previous XAI definitions in XAI literature and systematic reviews of literature on AI applications on the IVC. The authors acknowledge this definition will not be widely applicable to a wide range of industries, therefore it’s reiterated that the proposed XAI definition is applicable to financial services and the insurance industry. This definition will aid in adapting regulation in the insurance industry to suit an AI-rich insurance industry. Further clarification is necessary on the relationship between explanation and trust as both concepts pertain to XAI, with research recommendations centered on the extent to which explanations assist in the development of trust in AI models.
To achieve a substantial understanding of the entire potential of XAI research requires an interdisciplinary effort. The systematic review of XAI methods in different research areas is a stepping-stone to full understanding of the research field, with medicine reviews providing the bulk of knowledge on the topic at the time of writing. Considering the research gap regarding XAI applications along the IVC, this paper is one of the first attempts to provide an overview of XAI’s current use within the insurance industry.
Conceptualization, E.O., B.S., M.M. and M.C.; methodology, E.O. and B.S.; investigation, E.O. and B.S.; data curation, E.O.; writing—original draft preparation, E.O. and J.R.; writing—review and editing, B.S., M.M., M.C., J.R. and G.C.; visualization, E.O.; supervision, B.S., M.M. and M.C.; funding acquisition, G.C. All authors have read and agreed to the published version of the manuscript.
The authors can confirm that all relevant data are included in the article.
The authors declare no conflict of interest.
| AdaBoost | Adaptive Boosting |
| AI | Artificial Intelligence |
| ANN | Artificial Neural Network |
| BN | Bayesian Network |
| BPNN | Back Propagation Neural Network |
| CHAID | Chi-Squared Automatic Interaction Detection |
| CNN | Convolutional Neural Networks |
| CPLF | Cost-Sensitive Parallel Learning Framework |
| CRM | Customer Relationship Management |
| DFSVM | Dual Membership Fuzzy Support Vector Machine |
| DL | Deep Learning |
| ESIM | Evolutionary Support Vector Machine Inference Model |
| EvoDM | Evolutionary Data Mining |
| FL | Fuzzy Logic |
| GAM | Generalised Additive Model |
| GLM | Generalised Linear Model |
| HVSVM | Hull Vector Support Vector Machine |
| IoT | Internet of Things |
| IVC | Insurance Value Chain |
| KDD | |
| LASSO | Least Absolute Shrinkage and Selection Operator |
| MCAM | Markov Chain Approximation Method |
| ML | Machine Learning |
| NB | Naïve Bayes |
| NCA | Neighbourhood Component Analysis |
| NLP | Natural Language Processing |
| NN | Neural Network |
| PCA | Principal Component Analysis |
| RF | Random Forest |
| SBS | Sequential Backward Selection |
| SFS | Sequential Forward Selection |
| SHAP | Shapley Additive exPlanations |
| SOFM | Self-Organising Feature Map |
| SOM | Self-Organising Map |
| UBI | Usage-Based Insurance |
| WEKA | Waikato Environment for Knowledge Analysis |
| XAI | Explainable Artificial Intelligence |
| XGBoost | Extreme Gradient Boosting Algorithms |
Footnotes
1. The five XAI categories used were introduced to XAI literature by
2. Searched ‘The ACM Guide to Computing Literature’.
3. ‘Articles’ throughout this review refers to both academic articles and conference papers.
4. Several such surveys and reviews are discussed in
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Enlarge this image.Figure 1. Literature Search Process. Backward Searching includes the assessment of the references in each of the 103 relevant articles for additional articles of relevance to the current review. Note: Eling et al. (2021).
Enlarge this image.Figure 2. Insurance AI Articles Meeting Relevance Threshold (2000–2021) outlines the number of systematically reviewed articles by year according to the inclusion and exclusion criteria outlined in Section 3.1.
Enlarge this image.Figure 3. The PRISMA Flow Diagram is a recognised standard for systematic review literature search processes (Stovold et al. 2014). * ‘Source’ refers to the article inclusion criteria for this systematic review: journal articles and conference papers/proceedings are included.
Enlarge this image.Figure 4. IVC Stage and Corresponding XAI Method Employed present the seven IVC stages assessed in the systematically chosen articles and the XAI method used in their methodology. Support Activities is not included in this paper as no articles returned in the systematic literature search presented prediction tasks in line with insurance companies’ Support Activities.
XAI Variables used during the literature analysis to assess the explainability of AI systems applied within insurance industry practices. Reference
| Intrinsic vs. Post hoc | Intrinsic Interpretability | Describes how a model works and is interpretable by itself. Interpretability is achieved through imposing constraints on the model. | |
| Post hoc Interpretability | Analyses what else the (original) model can tell us, necessitating additional models to achieve explainability. The original model’s explainability is analysed after training. | ||
| Local vs. Global | Local Interpretability | Reveals the impact of input features on the overall model’s prediction. | |
| Global Interpretability | Local explanations are combined to present the overall AI model’s rules or features which determine their predictive outcome. | ||
| Model-Specific vs. Model-Agnostic | Model-specific Interpretation | Interpretation is limited to specific model classes as each interpretation method is based on a specific model’s internals. | |
| Model-agnostic Interpretation | Applied to any AI model after the model’s training. Analyses relationships between AI model’s feature inputs and outputs. |
Key Search Terms Interchangeable with (Explainable) Artificial Intelligence in the Literature Search Process.
| Artificial Intelligence (AI) | Smart Devices | Analytics | Support Vector Machine (SVM) |
| Genetic Algorithm | Neural Network (NN) | Computational Intelligence | Machine Learning (ML) |
| Convolutional Neural Network (CNN) | Artificial Neural Network (ANN) | Explainable Artificial Intelligence (XAI) | Deep Learning |
| Data Mining | Big Data | Fuzzy Systems | Fuzzy Logic |
| Swarm Intelligence | Natural Language Processing (NLP) | Image Analysis | Machine Vision |
The Insurance Value Chain. The stages of the insurance industry’s IVC is adapted from
| Value Chain Stage | Main Tasks | Impact of Artificial Intelligence Applications |
|---|---|---|
| Marketing | Market and customer research |
- Improved prediction of customer lifetime value |
| Product Development | Configuration of products Verification of legal requirements | - The establishment of add-on services such as early detection of new diseases and their prevention enables the development of new revenue streams in addition to risk coverage |
| Sales and Distribution | Customer acquisition and consultation Sales conversations |
- Support of human sales agents by offering advanced sales insights (e.g., cross- and up-selling opportunities) through smart data-driven virtual sales assistants (chatbots) for improved customer consultation and tailored product recommendations |
| Underwriting and Pricing | Product pricing (actuarial methods) |
- Automated application handling, underwriting and risk assessment processes enable accurate insurance quotes within minutes |
| Contract Administration and Customer Services | Change of contract data |
- Development of chatbots for the automated answering of written and verbal customer queries using Natural Language Processing (NLP) |
| Claim Management | Claim settlement Investigation of fraud | - Automated claims management leads to decreasing claim settlement life cycles and increased payout accuracy |
| Asset and Risk Management | Asset allocation |
- Automated investment research with more accurate and detailed market data enables portfolio management to make better-informed decisions due to new insights and more sophisticated analysis of data |
AI Methods and XAI Criteria used for the systematic analysis of the literature.
| AI Method | XAI Criteria | |
|---|---|---|
| Bayesian Network | Instance-based | Feature Interaction and Importance |
| Clustering | Regression | Attention Mechanism |
| Neural Network | Reinforcement Learning | (Data) Dimensionality Reduction |
| Decision Tree | Regularisation | Knowledge Distillation & Rule Extraction |
| Ensemble | Rule-based | Intrinsically Interpretable Models |
| Fuzzy Logic | Support Vector Machine | |
AI Methods and Prediction Tasks. Abbreviations in
| AI Method | Prediction Task(s) | Life/Non-Life | Line of Insurance |
||
|---|---|---|---|---|---|
| Marketing | |||||
| 1 | Neural Network | ANNs used to predict the propensity of consumers to purchase an insurance policy | - | - | |
| 2 | Regression | Develop a predictive modelling solution to aid the identification of the best insurance product group for current insurance product group of customers | - | - | |
| 3 | Ensemble | Prediction of insurance customer profitability | Life | Health | |
| 4 | Ensemble | Prediction of customer retention and profitability | - | - | |
| 5 | Ensemble | Classification to enhance the marketing of insurance products | Life | - | |
| 6 | Rule-based | Extraction of low-level knowledge data to answer high-level questions on customer acquisition, customer up- and cross-selling and customer retention within insurance companies | - | - | |
| Product Development | |||||
| 7 | Ensemble | Prediction of automobile insurance policies chosen by customers using Random Forest (RF) | Non-life | Motor | |
| 8 | Clustering | K-means used to identify clusters which contribute to the profit and loss of auto insurance companies | Non-life | Motor | |
| 9 | Neural Network | Driving behaviour classification | Non-life | Motor | |
| 10 | Neural Network | Calculation of life expectancy (mortality forecasting) based on the individual’s health status | Life | Health | |
| 11 | Bayesian Network | BN risk estimation approach for the emergence of new risk structures, including autonomous vehicles | Non-life | Motor and ProductLiability | |
| Sales and Distribution | |||||
| 12 | Decision Tree | Creation of business rules from customer-led data to improve insurer competitiveness | - | - | |
| 13 | Ensemble | XGBoost predictive classification algorithm provides Shapley values | Non-life | - | |
| 14 | Rule-based | Association between policyholder switching after a claim and the associated change in premium | Non-life | Motor | |
| 15 | Bayesian Network | Selection of promising prospective insurance customers from a vendor’s address list | - | - | |
| 16 | Ensemble | Prediction of auto-renewal using RF | Non-life | Motor | |
| 17 | Ensemble | Ensemble of DTs used to maximise the expected net profit of customers | - | - | |
| 18 | Clustering | Grouping of health insured population | Life | Health | |
| 19 | Bayesian Network | Estimation of insurance product recommendation | - | - | |
| Underwriting and Pricing | |||||
| 20 | Fuzzy Logic | Encoded the underwriting guidelines to automate the underwriting procedures of long-term care and life insurance policies | Life | Long Term Care | |
| 21 | Regression | Assess the enhanced accuracy of risk selection predictive models utilising driving behaviour variables in addition to traditional accident risk predictors | Non-life | Motor | |
| 22 | Ensemble | Ensemble learning-based approach to obtain information on a user’s risk classification which informs the compensation payout | Non-life | Motor | |
| 23 | Instance-based | Prediction of the applications of exclusions in life insurance policies when automated underwriting methods are employed | Life | - | |
| 24 | Fuzzy Logic | Automation of underwriting practices | - | - | |
| 25 | Neural Network | Predict the risk level of life insurance applicants | Life | - | |
| 26 | Rule-based | Predetermined feature values provided | Non-life | Motor | |
| 27 | Clustering | Evaluation of UBI automobile insurance policies | Non-life | Motor | |
| 28 | Support Vector Machine | Evaluation of loss risk and development of criteria for optimal insurance deductible decision making | Non-life | Construction | |
| 29 | Ensemble | Indirect estimation of the pure premium in motor vehicle insurance | Non-Life | Motor | |
| 30 | Regression | Use of the GLM to establish policyholders’ pure premium | Non-life | Motor | |
| 31 | Regression | GAMs used for rate-making | Non-life | Motor | |
| 32 | Ensemble | Mortality modelling using boosting regression techniques | Life | - | |
| 33 | Regularisation | LASSO penalty development to aid regularisation techniques in ML | - | - | |
| 34 | Clustering | Selection of representative policies for the assessment of variable annuity policy pricing | Life | - | |
| 35 | Clustering | Valuation of variable annuity policies | Life | - | |
| 36 | Reinforcement Learning | Monte Carlo-based modelling for variable annuity portfolios | Life | - | |
| 37 | Ensemble | Gradient Boosting Trees used to predict insurance losses | Non-life | Motor | |
| 38 | Ensemble | Bias-corrected bagging method used to improve predictive performance of regression trees | Non-life | - | |
| 39 | Regression | Risk probability prediction based on telematics driving data | Non-life | Motor | |
| 40 | Ensemble | Risk assessment of potential policyholders using risk scores within numerous ensembles of AI methods | Life | - | |
| 41 | Instance-based | A novel model for analysis of imbalanced datasets in end-to-end insurance processes | Life | - | |
| 42 | Rule-based | Knowledge-based system to enhance life underwriting processes | Life | - | |
| 43 | Clustering | Assessment and classification of premiums | Non-life | Motor | |
| 44 | Clustering | Deal with inadequately labelled data trajectories with drivers’ identifiers | Non-life | Motor | |
| 45 | Support Vector Machine | Prediction of claims which need reworking due to errors | Life | Health | |
| 46 | Ensemble | Driver identification using RF | Non-life | Motor | |
| 47 | Neural Network | Price the correct premium rate for ‘in-between’ risks between predefined tariff rates | Non-life | Property & Casualty | |
| 48 | Ensemble | Adaboost to predict claim frequency of auto insurance | Non-life | Motor | |
| 49 | Decision Tree | Prediction of insurance customers’ decisions following an automobile accident | Non-life | Motor | |
| 50 | Neural Network | Prediction of an insurance portfolio’s claim frequency for forthcoming years | Non-life | Motor | |
| 51 | Neural Network | Automatic multi-class labelling of ICD-9 codes of patient notes | Life | Health | |
| 52 | Neural Network | Determination of litigation risks for accounting professional liability insurance | Non-life | Professional Liability | |
| 53 | Instance-Based | Unsupervised pattern recognition framework for mobile telematics data to propose a solution to unlabelled telematics data | Non-life | Motor | |
| 54 | Neural Network | NNs used to classify policyholders as likely to renew or terminate, to aid the achievement of maximum potential profitability for the insurance company | Non-life | Motor | |
| 55 | Support Vector Machine | Stock price prediction | Non-life | Agriculture | |
| 56 | Neural Network | Optimisation of NN insurance pricing models | Non-life | Motor | |
| 57 | Neural Network | Classification to enhance NN functionality for automated insurance underwriting | - | - | |
| 58 | Rule-based | Rating model for UBI automobile insurance rates | - | - | |
| 59 | Ensemble | Gradient Boosting Trees used to predict insurance premiums | Non-life | Motor | |
| 60 | Clustering | Optimisation of insurance premium pricing | Non-life | Motor | |
| Contract Administration and Customer Services | |||||
| 61 | Fuzzy Logic | Creation of association rules which analyse customer grievances and summarise them | - | - | |
| 62 | Clustering | Prediction of airline customer clusters and appropriate Cancellation Protection Service insurance fee per customer group | Non-life | Airline | |
| 63 | Bayesian Network | The optimal set of hyperparameters for the later used ML model is found using Bayesian optimisation methods | - | - | |
| 64 | Neural Network | Automobile insurance customers’ risk estimate using ANN to inform contract development | Non-life | Motor | |
| 65 | Fuzzy Logic | Value creation for insurance customers | Life | - | |
| Claim Management | |||||
| 66 | Ensemble | Estimation of outstanding liabilities on a given policy using an ensemble of regression trees | - | - | |
| 67 | Regression | Calculate the probability of fraud in insurance files | Non-life | Motor | |
| 68 | Rule-based | Identification of fraud indicators | Non-Life | Motor | |
| 69 | Bayesian Network | Bayesian skewed logit model used to fit an insurance database (binary data) | Non-life | Motor | |
| 70 | Instance-Based | SOFM NN used to extract characteristics of medical insurance fraud behaviour | Life | Health | |
| 71 | Neural Network | NNs testing of regression models | Non-life | Liability | |
| 72 | Regression | Assessment of claim frequency | |||
| 73 | Ensemble | XGBoost used to detect automobile insurance fraudulent claims | Non-life | Motor | |
| 74 | Regression | Assessment of claim frequency | Non-life | Motor | |
| 75 | Neural Network | Estimation of claims reserves for individual reported claims | Non-life | ||
| 76 | Support Vector Machine | Error detection in insurance claims | Life | Health | |
| 77 | Clustering | Detection of fraud patterns | Non-life | Motor | |
| 78 | Ensemble | Medicare provider claims fraud | Life | Health | |
| 79 | Neural Network | Automation of fraud detection using ANN | Life | Health | |
| 80 | Clustering | Detection of fraudulent claims | Life | Health | |
| 81 | Rule-based | Fraudulent claim detection | Non-life | Motor | |
| 82 | Neural Network | CNN used to prevent claims leakage | Non-life | Motor | |
| 83 | Rule-based | Association Rules’ provision of actionable business insights for insurance claims data | Non-life | Liability | |
| 84 | Regression | GLM and GAM used in NLP to extract variables from text and use these variables in claims analysis | Non-life | Property &Casualty | |
| 85 | Ensemble | Random Forest for automobile insurance fraud detection | Non-life | Motor | |
| 86 | Clustering | Enhance the accuracy of claims fraud prediction | Non-life | Motor | |
| 87 | Rule-based | Fraud detection | Life | Health | |
| 88 | Fuzzy Logic | To distinguish whether fraudulent actions are involved in insurance claims settlement | - | - | |
| 89 | Regression | GLM to model insurance costs’ dispersion | Non-life | Motor | |
| 90 | Instance-based | Determination of joint medical fraud through reducing the occurrence of false positives caused by non-fraudulent abnormal behaviour | Life | Health | |
| 91 | Fuzzy Logic | Utilising fuzzy rule-based techniques to improve fraud detection | Non-life | Motor | |
| 92 | Fuzzy Logic | DFSVM used to solve the issue of misdiagnosed fraud detection due to the ‘overlap’ problem in insurance fraud samples | Non-life | Motor | |
| 93 | Clustering | K-means used to increase performance and reduce the complexity of the model | Life | Health | |
| 94 | Regression | Fraud detection | Non-life | Motor | |
| 95 | Ensemble | Adaboost used in insurance claim fraud detection | Non-life | Motor | |
| 96 | Bayesian Network | NN for fraud detection | Non-life | Motor | |
| 97 | Neural Network | NN used to detect automobile insurance fraud | Non-life | Motor | |
| 98 | Ensemble | Random rough subspace method | Non-life | Motor | |
| 99 | Ensemble | Optimisation of BP Neural Network by combining it with an improved genetic algorithm | Non-life | Motor | |
| Asset and Risk Management | |||||
| 100 | Neural Network | Optimal reinsurance and dividend strategies for insurance companies | - | - | |
| 101 | Neural Network | Insurer insolvency prediction | - | - | |
| 102 | Neural Network | Determine the optimal insurance, reinsurance, and investment strategies of an insurance company | - | - | |
| 103 | Clustering | Grouping of insurance contracts | Life | Life | |
XAI Methods and their approach in the articles is outlined, with the additional XAI assessment of (i) intrinsic or post hoc, (ii) local or global, and (iii) model-specific or model-agnostic interpretability methods. Abbreviations in
| XAI Category | XAI Approach | Intrinsic/Post- hoc | Local/Global | Model- Specific/Agnostic | ||
|---|---|---|---|---|---|---|
| Marketing | ||||||
| 1 | Feature Interaction and Importance | Dataset is pre-processed with three feature selection methods; (1) Neighbourhood Component Analysis (NCA), (2) Sequential Forward Selection (SFS) and, (3) Sequential Backward Selection (SBS) | Intrinsic | Global | Model-agnostic | |
| 2 | Dimensionality Reduction | Identification of relevant data clusters to inform model development for differing product groups | Post hoc | Local | Model-agnostic | |
| 3 | Intrinsically Interpretable Model | RF regression | Intrinsic | Global | Model-specific | |
| 4 | Feature Interaction and Importance | Exploration of three major predictor categories as explanatory variables | Intrinsic | Local | Model-specific | |
| 5 | Intrinsically Interpretable Model | RF provides automatic feature selection which aids interpretability of the model | Intrinsic | Global | Model-specific | |
| 6 | Knowledge Distillation and Rule Extraction | Bridge the gap between databases and their users by implementing KDD methods | Intrinsic | Local | Model-specific | |
| Product Development | ||||||
| 7 | Feature Interaction and Importance | Classification of data into different sets according to different policy options available | Intrinsic | Local | Model-specific | |
| 8 | Intrinsically Interpretable Model | Pattern recognition with clustering algorithms to find missing data to minimise insurance losses | Intrinsic | Global | Model-specific | |
| 9 | Feature Interaction and Importance | Extraction of relevant features | Post hoc | Local | Model-agnostic | |
| 10 | Feature Interaction and Importance | NN proposed as a better predictor of life expectancy than the Lee–Carter model due to the ability to adapt for each sex and each cause of life expectancy through a learning algorithm using historical data | Post hoc | Local | Model-agnostic | |
| 11 | Knowledge Distillation and Rule Extraction | Determination of causal and probabilistic dependencies through subjective assumptions (of the data) | Intrinsic | Local | Model-specific | |
| Sales and Distribution | ||||||
| 12 | Feature Interaction and Importance | CHAID used to create groups and gain an understanding of their impact on the dependent variable | Intrinsic | Local | Model-specific | |
| 13 | Intrinsically Interpretable Model | Similarity clustering of the returned Shapley values to analyse customers’ insurance buying behaviour | Intrinsic | Global | Model-specific | |
| 14 | Knowledge Distillation and Rule Extraction | Association rule learning to identify relationships among variables | Intrinsic | Global | Model-specific | |
| 15 | Feature Interaction and Importance | PCA is used to reduce the dimensionality of the features and reduce the chance of overfitting | Post hoc | Local | Model-agnostic | |
| 16 | Dimensionality Reduction | Removal of dataset features which have no bearing on the customers’ likelihood to renew | Intrinsic | Local | Model-specific | |
| 17 | Knowledge Distillation and Rule Extraction | Development of postprocessing step to extract actionable knowledge from DTs to obtain actions which are associated with attribute-value changes | Intrinsic | Local | Model-specific | |
| 18 | Intrinsically Interpretable Model | Clustering the insured population using k-means | Intrinsic | Global | Model-specific | |
| 19 | Attention Mechanism | Parameter optimisation for NB model | Post hoc | Local | Model-agnostic | |
| Underwriting and Pricing | ||||||
| 20 | Feature Interaction and Importance | Use of NLP and explanation of the interaction of different model features which alters the model | Intrinsic | Global | Model-specific | |
| 21 | Feature Interaction and Importance | Stepwise feature selection | Intrinsic | Global | Model-specific | |
| 22 | Dimensionality Reduction | Found the 5 most relevant features to inform driving behaviour | Intrinsic | Local | Model-specific | |
| 23 | Feature Interaction and Importance | Recursive Feature Elimination to provide feature rankings for feature subsets | Post hoc | Global | Model-agnostic | |
| 24 | Knowledge Distillation and Rule Extraction | Fuzzy rule-based decision systems used to encode risk classification of complex underwriting tasks | Intrinsic | Local | Model-specific | |
| 25 | Dimensionality Reduction | Correlation-Based Feature Selection and PCA | Intrinsic | Local | Model-specific | |
| 26 | Feature Interaction and Importance | SHAP is used to provide the contribution of each feature value to the prediction in comparison to the average prediction | Post hoc | Local | Model-agnostic | |
| 27 | Intrinsically Interpretable Model | Identification of driver behaviour using ML algorithms | Intrinsic | Global | Model-specific | |
| 28 | Knowledge Distillation and Rule Extraction | Development of loss prediction model using the ESIM | Intrinsic | Global | Model-specific | |
| 29 | Dimensionality Reduction | Exploitation of knowledge from certain characteristics of datasets to estimate conditional probabilities and conditional expectations given the knowledge of the variable representing the pure premium | Intrinsic | Local | Model-specific | |
| 30 | Dimensionality Reduction | Use of policyholders’ relevant characteristics to determine the pure premium | Intrinsic | Local | Model-specific | |
| 31 | Knowledge Distillation and Rule Extraction | Bayesian GAMs developed using MCAM inference | Intrinsic | Local | Model-specific | |
| 32 | Attention Mechanism | Back-testing parametric mortality models | Post hoc | Global | Model-agnostic | |
| 33 | Knowledge Distillation and Rule Extraction | Development of SMuRF algorithm to allow for Sparse Multi-type Regularised Feature modelling | Intrinsic | Global | Model-specific | |
| 34 | Knowledge Distillation and Rule Extraction | Gaussian Process Regression employed to value variable annuity policies | Intrinsic | Local | Model-specific | |
| 35 | Knowledge Distillation and Rule Extraction | Kriging Regression method employed to value variable annuity policies | Intrinsic | Local | Model-specific | |
| 36 | Knowledge Distillation and Rule Extraction | Generalised Beta of the Second Kind (GB2) Regression method employed to value variable annuity policies | Intrinsic | Local | Model-specific | |
| 37 | Intrinsically Interpretable Model | Interpretable results given by the simple linear model through showcasing the relative influence of the input variables and their partial dependence plots | Intrinsic | Global | Model-specific | |
| 38 | Knowledge Distillation and Rule Extraction | Bagging creates several regression trees which fits a bootstrap sample of the training data and makes a prediction through averaging the predicted outcomes from the bootstrapped trees | Post hoc | Local | Model-agnostic | |
| 39 | ( |
Dimensionality Reduction | Variables are binned to discretise continuous variables and construct tariff classes with significant predictive effects to improve interpretability of UBI predictive models | Post hoc | Intrinsic | Model-agnostic |
| 40 | Feature Interaction and Importance | Using WEKA software, the dimensional feature set was reduced for use | Intrinsic | Global | Model-specific | |
| 41 | Feature Interaction and Importance | Imbalanced data trend forecasting using learning descriptions and sequences and adjusting the CPLF | Post hoc | Local | Model-specific | |
| 42 | Knowledge Distillation and Rule Extraction | Containment of the sets of rules with similar purpose and/or structure which defines the knowledge bases | Intrinsic | Global | Model-agnostic | |
| 43 | Intrinsically Interpretable Model | Clustering provides homogeneity within classifications of risk and heterogeneity between risk classifications | Intrinsic | Global | Model-specific | |
| 44 | Intrinsically Interpretable Model | Gradient Boosting DTs used to classify (unlabelled) trajectories | Post hoc | Local | Model-specific | |
| 45 | Dimensionality Reduction | Frequency-based feature selection technique | Intrinsic | Global | Model-specific | |
| 46 | Dimensionality Reduction | Reduction in feature values’ noise (normalisation of sensing data) | Intrinsic | Local | Model-specific | |
| 47 | Attention Mechanism | Use of premium rate determination rules as network inputs in the BPNN to create the ‘missing rates’ of in-between risks | Post hoc | Local | Model-specific | |
| 48 | Dimensionality Reduction | Reduction in claim frequency prediction problem to multi-class problem | Post hoc | Global | Model-specific | |
| 49 | Knowledge Distillation and Rule Extraction | Combination of simple linear weights and residual components to replicate non-linear effects to resemble a fully parametrised PPCI-like (Payments per Claim Incurred) model | Intrinsic | Local | Model-specific | |
| 50 | Knowledge Distillation and Rule Extraction | Built a predictive model using previous Bayesian credibility inputs to predict the value of another field | Post hoc | Local | Model-specific | |
| 51 | Attention Mechanism | NLP used for document classification of medical record notes, with RNNs employed to encode vectors in Bi-LTSM model | Intrinsic | Local | Model-specific | |
| 52 | Attention Mechanism | Model is developed from the relationships between the variables gained from previous data and then tested | Post hoc | Local | Model-specific | |
| 53 | Feature Interaction and Importance | SOM to reduce data complexity | Intrinsic | Global | Model-specific | |
| 54 | Feature Interaction and Importance | Assessed the variables of relevance to the current task through rejecting variables with x |
Post hoc | Local | Model-agnostic | |
| 55 | Attention Mechanism | Parameter optimisation for SVM model | Intrinsic | Global | Model-specific | |
| 56 | Feature Interaction and Importance | Enhancement of neural network efficiency through feature selection | Intrinsic | Global | Model-specific | |
| 57 | Knowledge Distillation and Rule Extraction | Comparison of four NN models for automated insurance underwriting | Post hoc | Local | Model-specific | |
| 58 | Knowledge Distillation and Rule Extraction | Combination of the CNN and HVSVM models to create a model with higher discrimination accuracy than either model presents alone | Post hoc | Global | Model-specific | |
| 59 | Intrinsically Interpretable Model | TDBoost package provides interpretable results | Intrinsic | Local | Model-specific | |
| 60 | Feature Interaction and Importance | Grouping of important clusters to input in NN model for insurance retention rates and price sensitivity prediction | Intrinsic | Local | Model-specific | |
| Contract Administration and Customer Services | ||||||
| 61 | Knowledge Distillation and Rule Extraction | Treatment of each variable as having a certain degree of membership with certain rules to categorise complaints | Intrinsic | Global | Model-specific | |
| 62 | Feature Interaction and Importance | Cancellation Protection Service insurance fee is calculated based on the relevant weight of each cluster | Intrinsic | Global | Model-specific | |
| 63 | Feature Interaction and Importance | SHAP is used in evaluating the feature importance in predicting the output level | Post hoc | Global | Model-agnostic | |
| 64 | Dimensionality Reduction | Only relevant parameters are considered in the ANN model | Intrinsic | Local | Model-specific | |
| 65 | Knowledge Distillation and Rule Extraction | Development of integrated ML model to carry out the prediction task | Intrinsic | Local | Model-specific | |
| Claim Management | ||||||
| 66 | Feature Interaction and Importance | Definition of policy subsets within the synthetic dataset | Post hoc | Local | Model-agnostic | |
| 67 | Feature Interaction and Importance | Regression used to isolate significant contributory variables in fraud | Intrinsic | Local | Model-specific | |
| 68 | Intrinsically Interpretable Model | Comparison of various intrinsic AI methods for fraud indicator identification | Intrinsic | Local | Model-specific | |
| 69 | Knowledge Distillation and Rule Extraction | Use of a skewed logit model to more accurately classify fraudulent insurance claims | Post hoc | Global | Model-agnostic | |
| 70 | Dimensionality Reduction | PCA in the reduction in data’s dimensionality | Post hoc | Local | Model-agnostic | |
| 71 | Dimensionality Reduction | Extraction of relevant features | Post hoc | Global | Model-specific | |
| 72 | Attention Mechanism | Describe the joint development process of individual claim payments and claims incurred | Intrinsic | Global | Model-agnostic | |
| 73 | Knowledge Distillation and Rule Extraction | Combination of many regression trees together in order to optimise the objective function and then learn a prediction function | Intrinsic | Global | Model-agnostic | |
| 74 | Knowledge Distillation and Rule Extraction | Comparison of various fitted models which summarise all the covariates’ effects on claim frequency | Intrinsic | Global | Model-specific | |
| 75 | Knowledge Distillation and Rule Extraction | NN proposed which is modelled through learning from one probability/regression function to the other via parameter sharing | Post hoc | Local | Model-specific | |
| 76 | Knowledge Distillation and Rule Extraction | Development of an interactive prioritisation component to aid auditors in their fraud detection | Post hoc | Local | Model-specific | |
| 77 | Knowledge Distillation and Rule Extraction | Definition of rules based on each cluster to determine future fraud propensity (using WEKA) | Intrinsic | Global | Model-specific | |
| 78 | Feature Interaction and Importance | Removed unnecessary data features | Intrinsic | Local | Model-specific | |
| 79 | Feature Interaction and Importance | Class imbalance within the dataset is rectified using one-hot encoding | Post hoc | Local | Model-specific | |
| 80 | Knowledge Distillation and Rule Extraction | Development of an electronic fraud & abuse detection model | Post hoc | Global | Model-agnostic | |
| 81 | Feature Interaction and Importance | Classifier construction using NB | Intrinsic | Local | Model-specific | |
| 82 | Feature Interaction and Importance | Fine-tuning of the dataset | Post hoc | Local | Model-specific | |
| 83 | Knowledge Distillation and Rule Extraction | Development of Association Rules function for Workers’ Compensation claim data analysis | Intrinsic | Global | Model-specific | |
| 84 | Knowledge Distillation and Rule Extraction | Transformation of words to vectors, where each vector represents some feature of the word | Intrinsic | Local | Model-specific | |
| 85 | Dimensionality Reduction | PCA used to transform data at each node to another space when computing the best split at that node | Intrinsic | Global | Model-specific | |
| 86 | Knowledge Distillation and Rule Extraction | Sequence generation to inform predictive model for fraudulent behaviour | Intrinsic | Local | Model-specific | |
| 87 | Knowledge Distillation and Rule Extraction | Two evolutionary data mining (EvoDM) algorithms developed to improve insurance fraud prediction; (1) GAK-means (combination of K-means algorithm with genetic algorithm) and, (2) MPSO-K-means (combination of K-means algorithm with Momentum-type Particle Swarm Optimisation (MPSO)) | Post-hoc | Local | Model-specific | |
| 88 | Knowledge Distillation andRule Extraction | Mimic the expertise of the human insurance auditors in real life insurance claim settlement scenarios | Post-hoc | Local | Model-agnostic | |
| 89 | Intrinsically Interpretable Model | Modelling of insurance costs’ dispersion and mean | Intrinsic | Local | Model-specific | |
| 90 | Feature Interaction and Importance | Formulation of compact clusters of individual behaviour in a large dataset | Intrinsic | Local | Model-specific | |
| 91 | Feature Interaction and Importance | K-means clustering used to prepare dataset prior to FL technique application | Intrinsic | Local | Model-specific | |
| 92 | Feature Interaction and Importance | Avoidance of curse of dimensionality problem through kernel function use for SVM’s calculation | Post hoc | Global | Model-agnostic | |
| 93 | Knowledge Distillation and Rule Extraction | Association rule learning to identify frequent fraud occurring patterns for varying groups | Intrinsic | Local | Model-specific | |
| 94 | Dimensionality Reduction | Removal of fraud indicators with 10 or less instances to aid model convergence and stability during estimation | Intrinsic | Global | Model-specific | |
| 95 | Attention Mechanism | Computation of the relative importance (weight) of individual components of suspicious claim occurrences | Intrinsic | Global | Model-specific | |
| 96 | Feature Interaction and Importance | Determination of relevant inputs for the NN model | Post hoc | Local | Model-agnostic | |
| 97 | Dimensionality Reduction | Extraction of text features hiding in the text descriptions of claims (Latent Dirichlet Allocation-based deep learning for text analytics) | Post hoc | Local | Model-agnostic | |
| 98 | Knowledge Distillation and Rule Extraction | Random rough subspace method incorporated into NN to detect insurance fraud | Intrinsic | Global | Model-specific | |
| 99 | Dimensionality Reduction | PCA used to reduce dimensions of the multi-dimensional feature matrix, where the reduced data retains the main information of the original data | Intrinsic | Global | Model-specific | |
| Asset and Risk Management | ||||||
| 100 | Knowledge Distillation and Rule Extraction | Development of deep learning Markov chain approximation method (MCAM) | Intrinsic | Global | Model-specific | |
| 101 | Attention Mechanism | Tuning of the NN | Intrinsic | Local | Model-specific | |
| 102 | Knowledge Distillation and Rule Extraction | MCAM to estimate the initial guess of the NN | Intrinsic | Global | Model-specific | |
| 103 | Knowledge Distillation and Rule Extraction | Approximation of representative portfolio groups to then nest in NN | Post hoc | Local | Model-specific | |
Appendix A. XAI Variables
Key XAI variables and criteria used both in the systematic review and throughout this paper are briefly outlined below as foundation for the paper’s results and discussion. These XAI groupings are derived from
Appendix A.1. Intrinsic vs. Post hoc Interpretability
The main differentiating aspect between an intrinsic and a post hoc interpretable explanation is whether interpretability is achieved through imposing constraints on the complexity of the model (intrinsic) or whether the models’ explainability was analysed after training (post hoc) (
Appendix A.2. Local vs. Global Interpretability
Local explanations primarily reveal the impact of input features on the overall model’s prediction, while local explanations inspect model concepts to describe how the model works (
Appendix A.3. Model-Specific vs. Model-Agnostic Interpretation
Both model-specific and model-agnostic interpretation methods are derived from the above intrinsic vs. post hoc explainability criteria. As the name suggests, model-specific interpretation methods are limited to specific model classes as each method is based on a specific model’s internals (
Association between XAI Interpretability Criteria where In-model and Post-model interpretability are defined using XAI variables.
| In-Model | Intrinsic | Model-specific |
| Post-Model | Post hoc | Model-agnostic |
Appendix B. Database of Reviewed Articles
Appendix B.1. Journal Articles Included in the Systematic Review
| Reference | Title | Lead Author | Year | Source | Volume | Issue Number |
| Automating the underwriting of insurance applications | Aggour | 2006 | AI Magazine | 27 | 3 | |
| The value of vehicle telematics data in insurance risk selection processes | Baecke | 2017 | Decision Support Systems | 98 | ||
| A machine learning approach for individual claims reserving in insurance | Baudry | 2019 | Applied Stochastic Models in Business and Industry | 35 | 5 | |
| A model for the detection of insurance fraud | Belhadji | 2000 | The Geneva Papers on Risk and Insurance-Issues and Practice | 25 | 4 | |
| Identifying Key Fraud Indicators in the Automobile Insurance Industry Using SQL Server Analysis Services | Benedek | 2019 | Studia Universitatis Babes-Bolyai | 64 | 2 | |
| A Bayesian dichotomous model with asymmetric link for fraud in insurance | Bermúdez | 2008 | Insurance: Mathematics and Economics | 42 | 2 | |
| Risk prediction in life insurance industry using supervised learning algorithms | Boodhun | 2018 | Complex & Intelligent Systems | 4 | 2 | |
| A “pay-how-you-drive” car insurance approach through cluster analysis | Carfora | 2019 | Soft Computing | 23 | 9 | |
| A Neural Network-Based Approach in Predicting Consumers’ Intentions of Purchasing Insurance Policies | Chang | 2021 | Acta Informatica Pragensia | 10 | 2 | |
| Decision making for contractor insurance deductible using the evolutionary support vector machines inference model | Cheng | 2011 | Expert Systems with Applications | 38 | 6 | |
| Optimal insurance strategies: A hybrid deep learning Markov chain approximation approach | Cheng | 2020 | ASTIN Bulletin: The Journal of the IAA | 50 | 2 | |
| An approach to model complex high–dimensional insurance data | Christmann | 2004 | Allgemeines Statistisches Archiv | 88 | 4 | |
| Auto insurance premium calculation using generalized linear models | David | 2015 | Procedia Economics and Finance | 20 | ||
| Neural networks for the joint development of individual payments and claim incurred | Delong | 2020 | Risks | 8 | 2 | |
| Non-life rate-making with Bayesian GAMs | Denuit | 2004 | Insurance: Mathematics and Economics | 35 | 3 | |
| Machine learning techniques for mortality modeling | Deprez | 2017 | European Actuarial Journal | 7 | 2 | |
| Acquiring Insurance Customer: The CHAID Way | Desik | 2012 | IUP Journal of Knowledge Management | 10 | 3 | |
| Segmentation-Based Predictive Modeling Approach in Insurance Marketing Strategy | Desik | 2016 | IUP Journal of Business Strategy | 13 | 2 | |
| Sparse regression with multi-type regularized feature modeling | Devriendt | 2021 | Insurance: Mathematics and Economics | 96 | ||
| Individual loss reserving using a gradient boosting-based approach | Duval | 2019 | Risks | 7 | 3 | |
| Customer profitability forecasting using Big Data analytics: A case study of the insurance industry | Fang | 2016 | Computers & Industrial Engineering | 101 | ||
| Hierarchical insurance claims modeling | Frees | 2008 | Journal of the American Statistical Association | 103 | 484 | |
| An individual claims reserving model for reported claims | Gabrielli | 2021 | European Actuarial Journal | 11 | 2 | |
| Application of data clustering and machine learning in variable annuity valuation | Gan | 2013 | Journal of the American Statistical Association | 53 | 3 | |
| Regression modeling for the valuation of large variable annuity portfolios | Gan | 2018 | North American Actuarial Journal | 22 | 1 | |
| Fraud detection in automobile insurance using a data mining based approach | Ghorbani | 2018 | International Journal of Mechatronics, Elektrical and Computer Technology (IJMEC) | 8 | 27 | |
| Why to buy insurance? An Explainable Artificial Intelligence Approach | Gramegna | 2020 | Risks | 8 | 4 | |
| Gradient boosting trees for auto insurance loss cost modeling and prediction | Guelman | 2012 | Expert Systems with Applications | 39 | 3 | |
| An effective bias-corrected bagging method for the valuation of large variable annuity portfolios | Gweon | 2020 | ASTIN Bulletin: The Journal of the IAA | 50 | 3 | |
| The detection of medicare fraud using machine learning methods with excluded provider labels | Herland | 2018 | Journal of Big Data | 5 | 1 | |
| Automobile insurance classification ratemaking based on telematics driving data | Huang | 2019 | Decision Support Systems | 127 | ||
| Artificial neural network model for predicting insurance insolvency | Ibiwoye | 2012 | International Journal of Management and Business Research | 2 | 1 | |
| Assessing risk in life insurance using ensemble learning | Jain | 2019 | Journal of Intelligent & Fuzzy Systems | 37 | 2 | |
| Association rules for understanding policyholder lapses | Jeong | 2018 | Risks | 6 | 3 | |
| Cost-sensitive parallel learning framework for insurance intelligence operation | Jiang | 2018 | IEEE Transactions on Industrial Electronics | 66 | 12 | |
| A hybrid deep learning method for optimal insurance strategies: Algorithms and convergence analysis | Jin | 2021 | Insurance: Mathematics and Economics | 96 | ||
| Medicare fraud detection using neural networks | Johnson | 2019 | Journal of Big Data | 6 | 1 | |
| A knowledge-based system for life insurance underwriting | Joram | 2017 | International Journal of Information Technology and Computer Science | 3 | ||
| Using the clustering algorithms and rule-based of data mining to identify affecting factors in the profit and loss of third party insurance, insurance company auto | Karamizadeh | 2016 | Indian Journal of science and Technology | 9 | 7 | |
| A nonparametric data mining approach for risk prediction in car insurance: a case study from the Montenegrin market | Kašćelan | 2016 | Economic research-Ekonomska istraživanja | 29 | 1 | |
| Driving Behavior Classification Based on Oversampled Signals of Smartphone Embedded Sensors Using an Optimized Stacked-LSTM Neural Networks | Khodairy | 2021 | IEEE Access | 9 | ||
| Grouping of contracts in insurance using neural networks | Kiermayer | 2021 | Scandinavian Actuarial Journal | 2021 | 4 | |
| An interactive machine-learning-based electronic fraud and abuse detection system in healthcare insurance | Kose | 2015 | Applied Soft Computing | 36 | ||
| Driver Identification Based on Wavelet Transform Using Driving Patterns | Kwak | 2020 | IEEE Transactions on Industrial Informatics | 17 | 4 | |
| Predicting customer retention and profitability by using random forests and regression forests techniques | Lariviere | 2005 | Expert systems with applications | 29 | 2 | |
| Actuarial applications of word embedding models | Lee | 2020 | ASTIN Bulletin: The Journal of the IAA | 50 | 1 | |
| A principle component analysis-based random forest with the potential nearest neighbor method for automobile insurance fraud identification | Li | 2018 | Applied Soft Computing | 70 | ||
| Using neural networks as a support tool in the decision making for insurance industry | Lin | 2009 | Expert Systems with Applications | 36 | 3 | |
| An ensemble random forest algorithm for insurance big data analysis | Lin | 2017 | IEEE Access | 5 | ||
| Using multi-class AdaBoost tree for prediction frequency of auto insurance | Liu | 2014 | Journal of Applied Finance and Banking | 4 | 5 | |
| Sequence Mining and Prediction-Based Healthcare Fraud Detection Methodology | Matloob | 2020 | IEEE Access | 8 | ||
| Machine Learning-Based Predictions of Customers’ Decisions in Car Insurance | Neumann | 2019 | Applied Artificial Intelligence | 33 | 9 | |
| A fuzzy-based algorithm for auditors to detect elements of fraud in settled insurance claims | Pathak | 2005 | Managerial Auditing Journal | 20 | 6 | |
| Fuzzy formal concept analysis based opinion mining for CRM in financial services | Ravi | 2017 | Applied Soft Computing | 60 | ||
| Cancel-for-Any-Reason Insurance Recommendation Using Customer Transaction-Based Clustering | Sadreddini | 2021 | IEEE Access | 9 | ||
| Artificial intelligence for estimation of future claim frequency in non-life insurance | Sakthivel | 2017 | Global Journal of Pure and Applied Mathematics | 13 | 6 | |
| Risk Assessment for Accounting Professional Liability Insurance | Sevim | 2016 | Sosyoekonomi | 24 | 29 | |
| Mortality forecasting using neural networks and an application to cause-specific data for insurance purposes | Shah | 2009 | Journal of Forecasting | 28 | 6 | |
| Semi-autonomous vehicle motor insurance: A Bayesian Network risk transfer approach | Sheehan | 2017 | Transportation Research Part C: Emerging Technologies | 82 | ||
| A mobile telematics pattern recognition framework for driving behavior extraction | Siami | 2020 | IEEE Transactions on Intelligent Transportation Systems | 22 | 3 | |
| An analysis of customer retention and insurance claim patterns using data mining: A case study | Smith | 2000 | Journal of the Operational Research Society | 51 | 5 | |
| Fitting Tweedie’s compound Poisson model to insurance claims data: dispersion modelling | Smyth | 2002 | ASTIN Bulletin: The Journal of the IAA | 32 | 1 | |
| Abnormal group-based joint medical fraud detection | Sun | 2018 | IEEE Access | 7 | ||
| How to separate the wheat from the chaff: Improved variable selection for new customer acquisition | Tillmanns | 2017 | Journal of Marketing | 81 | 2 | |
| A holistic fuzzy approach to create competitive advantage via quality management in services industry (case study: life-insurance services) | Vaziri | 2016 | Management decision | 54 | 8 | |
| Auto claim fraud detection using Bayesian learning neural networks | Viaene | 2002 | Expert Systems with Applications | 29 | 3 | |
| A case study of applying boosting Naive Bayes to claim fraud diagnosis | Viaene | 2004 | Journal of Risk and Insurance | 69 | 3 | |
| A case study of applying boosting Naive Bayes to claim fraud diagnosis | Viaene | 2005 | IEEE Transactions on Knowledge and Data Engineering | 16 | 5 | |
| Research on the Features of Car Insurance Data Based on Machine Learning | Wang | 2020 | Procedia Computer Science | 166 | ||
| Leveraging deep learning with LDA-based text analytics to detect automobile insurance fraud | Wang | 2018 | Decision Support Systems | 105 | ||
| Market fluctuation and agricultural insurance forecasting model based on machine learning algorithm of parameter optimization | Wei | 2019 | Journal of Intelligent & Fuzzy Systems | 37 | 5 | |
| Bias regularization in neural network models for general insurance pricing | Wüthrich | 2020 | European Actuarial Journal | 10 | 1 | |
| Research on the UBI Car Insurance Rate Determination Model Based on the CNN-HVSVM Algorithm | Yan | 2020 | IEEE Access | 8 | ||
| Improved adaptive genetic algorithm for the vehicle Insurance Fraud Identification Model based on a BP Neural Network | Yan | 2020 | Theoretical Computer Science | 817 | ||
| Extracting actionable knowledge from decision trees | Yang | 2006 | IEEE Transactions on Knowledge and data Engineering | 19 | 1 | |
| Insurance premium prediction via gradient tree-boosted Tweedie compound Poisson models | Yang | 2018 | Journal of Business & Economic Statistics | 36 | 3 | |
| A mathematical programming approach to optimise insurance premium pricing within a data mining framework | Yeo | 2002 | Journal of the Operational research Society | 53 | 11 |
Appendix B.2. Conference Papers Included in the Systematic Review
| Reference | Title | Lead Author | Year | Source |
| Predicting car insurance policies using random forest | Alshamsi | 2014 | 2014 10th International Conference on Innovations in Information Technology (IIT) | |
| Good drivers pay less: A study of usage-based vehicle insurance models | Bian | 2018 | Transportation research part A: policy and practice | |
| Transportation research part A: policy and practice | Biddle | 2018 | Australasian Database Conference | |
| Evolutionary optimization of fuzzy decision systems for automated insurance underwriting | Bonissone | 2002 | 2002 IEEE World Congress on Computational Intelligence. 2002 IEEE International Conference on Fuzzy Systems | |
| Contextualising local explanations for non-expert users: an XAI pricing interface for insurance | Bove | 2021 | IUI Workshops | |
| Using PCA to improve the detection of medical insurance fraud in SOFM neural networks | Cao | 2019 | Proceedings of the 2019 3rd International Conference on Management Engineering, Software Engineering and Service Sciences | |
| Extreme gradient boosting machine learning algorithm for safe auto insurance operations | Dhieb | 2019 | 2019 IEEE International Conference on Vehicular Electronics and Safety (ICVES) | |
| A data mining framework for valuing large portfolios of variable annuities | Gan | 2017 | Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining | |
| Interactive learning for efficiently detecting errors in insurance claims | Ghani | 2011 | Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining | |
| Distinguishing trajectories from different drivers using incompletely labeled trajectories | Kieu | 2018 | Proceedings of the 27th ACM international conference on information and knowledge management | |
| Predicting fraudulent claims in automobile insurance | Kowshalya | 2018 | 2018 Second International Conference on Inventive Communication and Computational Technologies (ICICCT) | |
| Data mining to predict and prevent errors in health insurance claims processing | Kumar | 2010 | Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining | |
| Car Damage Detection and Classification | Kyu | 2020 | Proceedings of the 11th International Conference on Advances in Information Technology | |
| Mine your business—A novel application of association rules for insurance claims analytics | Lau | 2011 | CAS E-Forum. Arlington: Casualty Actuarial Society | |
| Application of evolutionary data mining algorithms to insurance fraud prediction | Liu | 2012 | Proceedings of 2012 4th International Conference on Machine Learning and Computing IPCSIT | |
| End-user access to multiple sources-Incorporating knowledge discovery into knowledge management | Morik | 2002 | International Conference on Practical Aspects of Knowledge Management | |
| ICD-9 tagging of clinical notes using topical word embedding | Samonte | 2018 | Proceedings of the 2018 International Conference on Internet and e-Business | |
| Feature importance analysis for customer management of insurance products | Sohail | 2021 | 2021 International Joint Conference on Neural Networks (ICJNN) | |
| Robust fuzzy rule based technique to detect frauds in vehicle insurance | Supraja | 2017 | 2017 International Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS) | |
| Insurance fraud identification research based on fuzzy support vector machine with dual membership | Tao | 2012 | 2012 International Conference on Information Management, Innovation Management and Industrial Engineering | |
| Computational intelligence approach for estimation of vehicle insurance risk level | Vassiljeva | 2017 | 2017 International Joint Conference on Neural Networks (IJCNN) | |
| Fraud detection and frequent pattern matching in insurance claims using data mining techniques | Verma | 2017 | 2017 Tenth International Conference on Contemporary Computing (IC3) | |
| Random rough subspace based neural network ensemble for insurance fraud detection | Xu | 2011 | 2011 Fourth International Joint Conference on Computational Sciences and Optimization | |
| Designing a Neural Network Decision System for Automated Insurance Underwriting | Yan | 2006 | Insurance Studies | |
| Clustering of the population benefiting from health insurance using k-means | Zahi | 2019 | Proceedings of the 4th International Conference on Smart City Applications | |
| Dynamic estimation model of insurance product recommendation based on Naive Bayesian model | Zhang | 2020 | Proceedings of the 2020 International Conference on Cyberspace Innovation of Advanced Technologies |
References
Adadi, Amina; Berrada, Mohammed. Peeking inside the black-box: A survey on explainable artificial intelligence (XAI). IEEE Access; 2018; 6, pp. 52138-60. [DOI: https://dx.doi.org/10.1109/ACCESS.2018.2870052]
Aggour, Kareem S.; Bonissone, Piero P.; Cheetham, William E.; Messmer, Richard P. Automating the underwriting of insurance applications. AI Magazine; 2006; 27, pp. 36-36.
Alshamsi, Asma S. Predicting car insurance policies using random forest. Paper presented at the 2014 10th International Conference on Innovations in Information Technology (IIT); Al Ain, United Arab Emirates, November 9–11; 2014.
Al-Shedivat, Maruan; Dubey, Avinava; Xing, Eric P. Contextual Explanation Networks. Journal of Machine Learning Research; 2020; 21, pp. 194:1-94:44.
Andrew, Jane; Baker, Max. The general data protection regulation in the age of surveillance capitalism. Journal of Business Ethics; 2021; 168, pp. 565-78. [DOI: https://dx.doi.org/10.1007/s10551-019-04239-z]
Anjomshoae, Sule; Najjar, Amro; Calvaresi, Davide; Främling, Kary. Explainable agents and robots: Results from a systematic literature review. Paper presented at the 18th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2019); Montreal, Canada, May 13–17; 2019.
Antoniadi, Anna Markella; Du, Yuhan; Guendouz, Yasmine; Wei, Lan; Mazo, Claudia; Becker, Brett A; Mooney, Catherine. Current challenges and future opportunities for XAI in machine learning-based clinical decision support systems: A systematic review. Applied Sciences; 2021; 11, 5088. [DOI: https://dx.doi.org/10.3390/app11115088]
Arrieta, Alejandro Barredo; Díaz-Rodríguez, Natalia; Ser, Javier Del; Bennetot, Adrien; Tabik, Siham; Barbado, Alberto; García, Salvador; Gil-López, Sergio; Molina, Daniel; Benjamins, Richard. Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information Fusion; 2020; 58, pp. 82-115. [DOI: https://dx.doi.org/10.1016/j.inffus.2019.12.012]
Baecke, Philippe; Bocca, Lorenzo. The value of vehicle telematics data in insurance risk selection processes. Decision Support Systems; 2017; 98, pp. 69-79. [DOI: https://dx.doi.org/10.1016/j.dss.2017.04.009]
Baehrens, David; Schroeter, Timon; Harmeling, Stefan; Kawanabe, Motoaki; Hansen, Katja; Müller, Klaus-Robert. How to explain individual classification decisions. The Journal of Machine Learning Research; 2010; 11, pp. 1803-31.
Balasubramanian, Ramnath; Libarikian, Ari; McElhaney, Doug. Insurance 2030—The Impact of AI on the Future of Insurance; McKinsey & Company: New York, 2018.
Barocas, Solon; Selbst, Andrew D. Big data’s disparate impact. California Law Review; 2016; 104, 671. [DOI: https://dx.doi.org/10.2139/ssrn.2477899]
Barry, Laurence; Charpentier, Arthur. Personalization as a promise: Can Big Data change the practice of insurance?. Big Data & Society; 2020; 7, 2053951720935143.
Baser, Furkan; Apaydin, Aysen. Calculating insurance claim reserves with hybrid fuzzy least squares regression analysis. Gazi University Journal of Science; 2010; 23, pp. 163-70.
Baudry, Maximilien; Robert, Christian Y. A machine learning approach for individual claims reserving in insurance. Applied Stochastic Models in Business and Industry; 2019; 35, pp. 1127-55. [DOI: https://dx.doi.org/10.1002/asmb.2455]
Bayamlıoğlu, Emre. The right to contest automated decisions under the General Data Protection Regulation: Beyond the so-called “right to explanation”. Regulation & Governance; 2021; 16, pp. 1058-78.
Bean, Randy. Transforming the Insurance Industry with Big Data, Machine Learning and AI. Forbes; 6 July 2021; Available online: https://www.forbes.com/sites/randybean/2021/07/06/transforming-the-insurance-industry-with-big-data-machine-learning-and-ai/?sh=4004a662f8a6 (accessed on 11 August 2021).
Beck, Hall P.; Dzindolet, Mary T.; Pierce, Linda G. Operators’ automation usage decisions and the sources of misuse and disuse. Advances in Human Performance and Cognitive Engineering Research; Emerald Group Publishing Limited: Bingley, 2002.
Belhadji, El Bachir; Dionne, George; Tarkhani, Faouzi. A model for the detection of insurance fraud. The Geneva Papers on Risk and Insurance-Issues and Practice; 2000; 25, pp. 517-38. [DOI: https://dx.doi.org/10.1111/1468-0440.00080]
Benedek, Botond; László, Ede. Identifying Key Fraud Indicators in the Automobile Insurance Industry Using SQL Server Analysis Services. Studia Universitatis Babes-Bolyai; 2019; 64, pp. 53-71. [DOI: https://dx.doi.org/10.2478/subboec-2019-0009]
Bermúdez, Lluís; Pérez, José María; Ayuso, Mercedes; Gómez, Esther; Vázquez, Francisco. J. A Bayesian dichotomous model with asymmetric link for fraud in insurance. Insurance: Mathematics and Economics; 2008; 42, pp. 779-86. [DOI: https://dx.doi.org/10.1016/j.insmatheco.2007.08.002]
Bian, Yiyang; Yang, Chen; Zhao, J. Leon; Liang, Liang. Good drivers pay less: A study of usage-based vehicle insurance models. Transportation Research Part A: Policy and Practice; 2018; 107, pp. 20-34. [DOI: https://dx.doi.org/10.1016/j.tra.2017.10.018]
Biddle, Rhys; Liu, Shaowu; Xu, Guandong. Automated Underwriting in Life Insurance: Predictions and Optimisation (Industry Track). Paper presented at Australasian Database Conference; Gold Coast, QLD, Australia, May 24–27; 2018.
Biecek, Przemysław; Chlebus, Marcin; Gajda, Janusz; Gosiewska, Alicja; Kozak, Anna; Ogonowski, Dominik; Sztachelski, Jakub; Wojewnik, Piotr. Enabling Machine Learning Algorithms for Credit Scoring—Explainable Artificial Intelligence (XAI) methods for clear understanding complex predictive models. arXiv; 2021; arXiv: 2104.06735
Biran, Or; Cotton, Courtenay. Explanation and justification in machine learning: A survey. Paper presented at the IJCAI-17 Workshop on Explainable AI (XAI); Melbourne, VIC, Australia, August 19–21; 2017.
Blier-Wong, Christopher; Cossette, Hélène; Lamontagne, Luc; Marceau, Etienne. Machine Learning in P&C Insurance: A Review for Pricing and Reserving. Risks; 2021; 9, 4.
Bonissone, Piero. P.; Subbu, Raj; Aggour, Kareem S. Evolutionary optimization of fuzzy decision systems for automated insurance underwriting. Paper presented at the 2002 IEEE World Congress on Computational Intelligence, 2002 IEEE International Conference on Fuzzy Systems, FUZZ-IEEE’02. Proceedings (Cat. No. 02CH37291); Honolulu, HI, USA, May 12–17; 2002.
Boodhun, Noorhannah; Jayabalan, Manoj. Risk prediction in life insurance industry using supervised learning algorithms. Complex & Intelligent Systems; 2018; 4, pp. 145-54.
Bove, Clara; Aigrain, Jonathan; Lesot, Marie-Jeanne; Tijus, Charles; Detyniecki, Marcin. Contextualising local explanations for non-expert users: An XAI pricing interface for insurance. Paper presented at the IUI Workshops; College Station, TX, USA, April 13–17; 2021.
Burgt, Joost van der. Explainable AI in banking. Journal of Digital Banking; 2020; 4, pp. 344-50.
Burrell, Jenna. How the machine ‘thinks’: Understanding opacity in machine learning algorithms. Big Data & Society; 2016; 3, 2053951715622512.
Bussmann, Niklas; Giudici, Paolo; Marinelli, Dimitri; Papenbrock, Jochen. Explainable ai in fintech risk management. Frontiers in Artificial Intelligence; 2020; 3, 26. [DOI: https://dx.doi.org/10.3389/frai.2020.00026] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33733145]
Cao, Hongfei; Zhang, Runtong. Using PCA to improve the detection of medical insurance fraud in SOFM neural networks. Paper presented at the 2019 3rd International Conference on Management Engineering, Software Engineering and Service Sciences; Wuhan, China, January 12–14; 2019.
Carabantes, Manuel. Black-box artificial intelligence: An epistemological and critical analysis. AI & Society; 2020; 35, pp. 309-17.
Carfora, Maria Francesca; Martinelli, Fabio; Mercaldo, Francesco; Nardone, Vittoria; Orlando, Albina; Santone, Antonella; Vaglini, Gigliola. A “pay-how-you-drive” car insurance approach through cluster analysis. Soft Computing; 2019; 23, pp. 2863-75. [DOI: https://dx.doi.org/10.1007/s00500-018-3274-y]
Carvalho, Diogo V.; Pereira, Eduardo M.; Cardoso, Jaime S. Machine learning interpretability: A survey on methods and metrics. Electronics; 2019; 8, 832. [DOI: https://dx.doi.org/10.3390/electronics8080832]
Cevolini, Alberto; Esposito, Elena. From pool to profile: Social consequences of algorithmic prediction in insurance. Big Data & Society; 2020; 7, 2053951720939228.
Chang, Wen Teng; Lai, Kee Huong. A Neural Network-Based Approach in Predicting Consumers’ Intentions of Purchasing Insurance Policies. Acta Informatica Pragensia; 2021; 10, pp. 138-54. [DOI: https://dx.doi.org/10.18267/j.aip.152]
Chen, Irene Y.; Szolovits, Peter; Ghassemi, Marzyeh. Can AI help reduce disparities in general medical and mental health care?. AMA Journal of Ethics; 2019; 21, pp. 167-79.
Cheng, Min-Yuan; Peng, Hsien-Sheng; Wu, Yu-Wei; Liao, Yi-Hung. Decision making for contractor insurance deductible using the evolutionary support vector machines inference model. Expert Systems with Applications; 2011; 38, pp. 6547-55. [DOI: https://dx.doi.org/10.1016/j.eswa.2010.11.084]
Cheng, Xiang; Jin, Zhuo; Yang, Hailiang. Optimal insurance strategies: A hybrid deep learning Markov chain approximation approach. ASTIN Bulletin: The Journal of the IAA; 2020; 50, pp. 449-77. [DOI: https://dx.doi.org/10.1017/asb.2020.9]
Chi, Oscar Hengxuan; Denton, Gregory; Gursoy, Dogan. Artificially intelligent device use in service delivery: A systematic review, synthesis, and research agenda. Journal of Hospitality Marketing & Management; 2020; 29, pp. 757-86.
Christmann, Andreas. An approach to model complex high–dimensional insurance data. Allgemeines Statistisches Archiv; 2004; 88, pp. 375-96. [DOI: https://dx.doi.org/10.1007/s101820400178]
Clinciu, Miruna-Adriana; Hastie, Helen. A survey of explainable AI terminology. Paper presented at the 1st Workshop on Interactive Natural Language Technology for Explainable Artificial Intelligence (NL4XAI 2019); Tokyo, Japan, October 29–November 1; 2019.
Coalition Against Insurance Fraud. Artificial Intelligence & Insurance Fraud; Coalition Against Insurance Fraud: Washington, DC, 2020; Available online: https://insurancefraud.org/wp-content/uploads/Artificial-Intelligence-and-Insurance-Fraud-2020.pdf (accessed on 2 May 2021).
Colaner, Nathan. Is explainable artificial intelligence intrinsically valuable?. AI & Society; 2022; 37, pp. 231-38.
Confalonieri, Roberto; Coba, Ludovik; Wagner, Benedikt; Besold, Tarek R. A historical perspective of explainable Artificial Intelligence. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery; 2021; 11, e1391. [DOI: https://dx.doi.org/10.1002/widm.1391]
Daniels, Norman. The ethics of health reform: Why we should care about who is missing coverage. Connecticut Law Review; 2011; 44, 1057.
Danilevsky, Marina; Qian, Kun; Aharonov, Ranit; Katsis, Yannis; Kawas, Ban; Sen, Prithviraj. A survey of the state of explainable AI for natural language processing. arXiv; 2020; arXiv: 2010.00711
Das, Arun; Rad, Paul. Opportunities and challenges in explainable artificial intelligence (xai): A survey. arXiv; 2020; arXiv: 2006.11371
David, Mihaela. Auto insurance premium calculation using generalized linear models. Procedia Economics and Finance; 2015; 20, pp. 147-56. [DOI: https://dx.doi.org/10.1016/S2212-5671(15)00059-3]
Delong, Łukasz; Wüthrich, Mario V. Neural networks for the joint development of individual payments and claim incurred. Risks; 2020; 8, 33. [DOI: https://dx.doi.org/10.3390/risks8020033]
Demajo, Lara Marie; Vella, Vince; Dingli, Alexiei. Explainable AI for interpretable credit scoring. arXiv; 2020; arXiv: 2012.03749
Denuit, Michel; Lang, Stefan. Non-life rate-making with Bayesian GAMs. Insurance: Mathematics and Economics; 2004; 35, pp. 627-47. [DOI: https://dx.doi.org/10.1016/j.insmatheco.2004.08.001]
Deprez, Philippe; Shevchenko, Pavel V.; Wüthrich, Mario V. Machine learning techniques for mortality modeling. European Actuarial Journal; 2017; 7, pp. 337-52. [DOI: https://dx.doi.org/10.1007/s13385-017-0152-4]
Desik, P. H. Anantha; Behera, Samarendra; Soma, Prashanth; Sundari, Nirmala. Segmentation-Based Predictive Modeling Approach in Insurance Marketing Strategy. IUP Journal of Business Strategy; 2016; 13, pp. 35-45.
Desik, P. H. Anantha; Behera, Samarendra. Acquiring Insurance Customer: The CHAID Way. IUP Journal of Knowledge Management; 2012; 10, pp. 7-13.
Devriendt, Sander; Antonio, Katrien; Reynkens, Tom; Verbelen, Roel. Sparse regression with multi-type regularized feature modeling. Insurance: Mathematics and Economics; 2021; 96, pp. 248-61. [DOI: https://dx.doi.org/10.1016/j.insmatheco.2020.11.010]
Dhieb, Najmeddine; Ghazzai, Hakim; Besbes, Hichem; Massoud, Yehia. Extreme gradient boosting machine learning algorithm for safe auto insurance operations. Paper presented at the 2019 IEEE International Conference on Vehicular Electronics and Safety (ICVES); Cairo, Egypt, September 4–6; 2019.
Diprose, William K.; Buist, Nicholas; Hua, Ning; Thurier, Quentin; Shand, George; Robinson, Reece. Physician understanding, explainability, and trust in a hypothetical machine learning risk calculator. Journal of the American Medical Informatics Association; 2020; 27, pp. 592-600. [DOI: https://dx.doi.org/10.1093/jamia/ocz229]
Doshi-Velez, Finale; Kim, Been. Towards a rigorous science of interpretable machine learning. arXiv; 2017; arXiv: 1702.08608
Došilović, Filip Karlo; Brčić, Mario; Hlupić, Nikica. Explainable artificial intelligence: A survey. Paper presented at the 2018 41st International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO); Opatija, Croatia, May 21–25; 2018.
Du, Mengnan; Liu, Ninghao; Hu, Xia. Techniques for interpretable machine learning. Communications of the ACM; 2019; 63, pp. 68-77. [DOI: https://dx.doi.org/10.1145/3359786]
Duval, Francis; Pigeon, Mathieu. Individual loss reserving using a gradient boosting-based approach. Risks; 2019; 7, 79.
Eckert, Theresa; Hüsig, Stefan. Innovation portfolio management: A systematic review and research agenda in regards to digital service innovations. Management Review Quarterly; 2021; 72, pp. 187-230. [DOI: https://dx.doi.org/10.1007/s11301-020-00208-3]
EIOPA. Artificial Intelligence Governance Principles: Towards Ethical and Trustworthy Artificial Intelligence in the European Insurance Sector; European Insurance and Occupational Pensions Authority (EIOPA): Luxembourg, 2021.
Eling, Martin; Nuessle, Davide; Staubli, Julian. The impact of artificial intelligence along the insurance value chain and on the insurability of risks. The Geneva Papers on Risk and Insurance-Issues and Practice; 2021; 47, pp. 205-41. [DOI: https://dx.doi.org/10.1057/s41288-020-00201-7]
EU. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data, and Repealing Directive 95/46/EC (General Data Protection Regulation). 2016; Available online: https://eur-lex.europa.eu/eli/reg/2016/679/oj (accessed on 27 June 2020).
Fang, Kuangnan; Jiang, Yefei; Song, Malin. Customer profitability forecasting using Big Data analytics: A case study of the insurance industry. Computers & Industrial Engineering; 2016; 101, pp. 554-64.
Felzmann, Heike; Villaronga, Eduard Fosch; Lutz, Christoph; Tamò-Larrieux, Aurelia. Transparency you can trust: Transparency requirements for artificial intelligence between legal norms and contextual concerns. Big Data & Society; 2019; 6, 2053951719860542.
Ferguson, Niall. The Ascent of Money: A Financial History of the World; Penguin: London, 2008.
Floridi, Luciano; Cowls, Josh; Beltrametti, Monica; Chatila, Raja; Chazerand, Patrice; Dignum, Virginia; Luetge, Christoph; Madelin, Robert; Pagallo, Ugo; Rossi, Francesca. AI4People—An ethical framework for a good AI society: Opportunities, risks, principles, and recommendations. Minds and Machines; 2018; 28, pp. 689-707. [DOI: https://dx.doi.org/10.1007/s11023-018-9482-5]
Ford, Martin. Architects of Intelligence: The Truth about AI from the People Building it; Packt Publishing Ltd.: Birmingham, 2018.
Fox, Maria; Long, Derek; Magazzeni, Daniele. Explainable planning. arXiv; 2017; arXiv: 1709.10256
Frees, Edward W.; Valdez, Emiliano A. Hierarchical insurance claims modeling. Journal of the American Statistical Association; 2008; 103, pp. 1457-69. [DOI: https://dx.doi.org/10.1198/016214508000000823]
Freitas, Alex A. Comprehensible classification models: A position paper. ACM SIGKDD Explorations Newsletter; 2014; 15, pp. 1-10. [DOI: https://dx.doi.org/10.1145/2594473.2594475]
Gabrielli, Andrea. An individual claims reserving model for reported claims. European Actuarial Journal; 2021; 11, pp. 541-77. [DOI: https://dx.doi.org/10.1007/s13385-021-00271-4]
Gade, Krishna; Geyik, Sahin Cem; Kenthapadi, Krishnaram; Mithal, Varun; Taly, Ankur. Explainable AI in industry. Paper presented at the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining; Anchorage, AK, USA, August 4–8; 2019.
Gan, Guojun; Valdez, Emiliano A. Valuation of large variable annuity portfolios: Monte Carlo simulation and synthetic datasets. Dependence Modeling; 2017; 5, pp. 354-74. [DOI: https://dx.doi.org/10.1515/demo-2017-0021]
Gan, Guojun; Huang, Jimmy Xiangji. A data mining framework for valuing large portfolios of variable annuities. Paper presented at the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; Halifax, NS, Canada, August 13–17; 2017.
Gan, Guojun. Application of data clustering and machine learning in variable annuity valuation. Insurance: Mathematics and Economics; 2013; 53, pp. 795-801.
Ghani, Rayid; Kumar, Mohit. Interactive learning for efficiently detecting errors in insurance claims. Paper presented at the 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; San Diego, CA, USA, August 21–24; 2011.
Ghorbani, Ali; Farzai, Sara. Fraud detection in automobile insurance using a data mining based approach. International Journal of Mechatronics, Elektrical and Computer Technology (IJMEC); 2018; 8, pp. 3764-71.
GlobalData. Artificial Intelligence (AI) in Insurance—Thematic Research; GlobalData: London, 2021.
Goddard, Michelle. The EU General Data Protection Regulation (GDPR): European regulation that has a global impact. International Journal of Market Research; 2017; 59, pp. 703-5. [DOI: https://dx.doi.org/10.2501/IJMR-2017-050]
Goodman, Bryce; Flaxman, Seth. European Union regulations on algorithmic decision-making and a “right to explanation”. AI Magazine; 2017; 38, pp. 50-57. [DOI: https://dx.doi.org/10.1609/aimag.v38i3.2741]
Gramegna, Alex; Giudici, Paolo. Why to Buy Insurance? An Explainable Artificial Intelligence Approach. Risks; 2020; 8, 137. [DOI: https://dx.doi.org/10.3390/risks8040137]
Gramespacher, Thomas; Posth, Jan-Alexander. Employing explainable AI to optimize the return target function of a loan portfolio. Frontiers in Artificial Intelligence; 2021; 4, 693022. [DOI: https://dx.doi.org/10.3389/frai.2021.693022]
Grant, Eric. The Social and Economic Value of Insurance; The Geneva Association (The International Association for the Study of Insurance Economics): Geneva, 2012; Available online: https://www.genevaassociation.org/sites/default/files/research-topics-docu-ment-type/pdf_public/ga2012-the_social_and_economic_value_of_insurance.pdf (accessed on 3 July 2020).
Grize, Yves-Laurent; Fischer, Wolfram; Lützelschwab, Christian. Machine learning applications in nonlife insurance. Applied Stochastic Models in Business and Industry; 2020; 36, pp. 523-37. [DOI: https://dx.doi.org/10.1002/asmb.2543]
Guelman, Leo. Gradient boosting trees for auto insurance loss cost modeling and prediction. Expert Systems with Applications; 2012; 39, pp. 3659-67. [DOI: https://dx.doi.org/10.1016/j.eswa.2011.09.058]
Gweon, Hyukjun; Li, Shu; Mamon, Rogemar. An effective bias-corrected bagging method for the valuation of large variable annuity portfolios. ASTIN Bulletin: The Journal of the IAA; 2020; 50, pp. 853-71. [DOI: https://dx.doi.org/10.1017/asb.2020.28]
Hadji Misheva, Branka; Hirsa, Ali; Osterrieder, Joerg; Kulkarni, Onkar; Lin, Stephen Fung. Explainable AI in Credit Risk Management. Credit Risk Management; 1 March 2021.
Hawley, Katherine. Trust, distrust and commitment. Noûs; 2014; 48, pp. 1-20. [DOI: https://dx.doi.org/10.1111/nous.12000]
Henckaerts, Roel; Antonio, Katrien; Côté, Marie-Pier. Model-Agnostic Interpretable and Data-driven suRRogates suited for highly regulated industries. Stat; 2020; 1050, 14.
Henckaerts, Roel; Côté, Marie-Pier; Antonio, Katrien; Verbelen, Roel. Boosting insights in insurance tariff plans with tree-based machine learning methods. North American Actuarial Journal; 2021; 25, pp. 255-85. [DOI: https://dx.doi.org/10.1080/10920277.2020.1745656]
Herland, Matthew; Khoshgoftaar, Taghi M.; Bauder, Richard A. Big data fraud detection using multiple medicare data sources. Journal of Big Data; 2018; 5, pp. 1-21. [DOI: https://dx.doi.org/10.1186/s40537-018-0138-3]
Hinton, Geoffrey; Vinyals, Oriol; Dean, Jeff. Distilling the knowledge in a neural network. arXiv; 2015; arXiv: 1503.02531
Hoffman, Robert R. A taxonomy of emergent trusting in the human–machine relationship. Cognitive Systems Engineering: The Future for a Changing World; CRC Press: Boca Raton, 2017; pp. 137-64.
Hoffman, Robert R.; Mueller, Shane T.; Klein, Gary; Litman, Jordan. Metrics for explainable AI: Challenges and prospects. arXiv; 2018; arXiv: 1812.04608
Hollis, Aidan; Strauss, Jason. Privacy, Driving Data and Automobile Insurance: An Economic Analysis; University Library of Munich: Munich, 2007.
Honegger, Milo. Shedding light on black box machine learning algorithms: Development of an axiomatic framework to assess the quality of methods that explain individual predictions. arXiv; 2018; arXiv: 1808.05054
Huang, Yifan; Meng, Shengwang. Automobile insurance classification ratemaking based on telematics driving data. Decision Support Systems; 2019; 127, 113156. [DOI: https://dx.doi.org/10.1016/j.dss.2019.113156]
Ibiwoye, Ade; Ajibola, Olawale Olaniyi; Sogunro, Ashim Babatunde. Artificial neural network model for predicting insurance insolvency. International Journal of Management and Business Research; 2012; 2, pp. 59-68.
Islam, Sheikh Rabiul; Eberle, William; Ghafoor, Sheikh K. Towards quantification of explainability in explainable artificial intelligence methods. Paper presented at the Thirty-Third International Flairs Conference; North Miami Beach, FL, USA, May 17–20; 2020.
Jacovi, Alon; Marasović, Ana; Miller, Tim; Goldberg, Yoav. Formalizing trust in artificial intelligence: Prerequisites, causes and goals of human trust in AI. Paper presented at the 2021 ACM Conference on Fairness, Accountability, and Transparency; Toronto, ON, Canada, March 3–10; 2021.
Jain, Rachna; Alzubi, Jafar A.; Jain, Nikita; Joshi, Pawan. Assessing risk in life insurance using ensemble learning. Journal of Intelligent & Fuzzy Systems; 2019; 37, pp. 2969-80.
Jeong, Himchan; Gan, Guojun; Valdez, Emiliano A. Association rules for understanding policyholder lapses. Risks; 2018; 6, 69. [DOI: https://dx.doi.org/10.3390/risks6030069]
Jiang, Xinxin; Pan, Shirui; Long, Guodong; Xiong, Fei; Jiang, Jing; Zhang, Chengqi. Cost-sensitive parallel learning framework for insurance intelligence operation. IEEE Transactions on Industrial Electronics; 2018; 66, pp. 9713-23. [DOI: https://dx.doi.org/10.1109/TIE.2018.2873526]
Jin, Zhuo; Yang, Hailiang; Yin, George. A hybrid deep learning method for optimal insurance strategies: Algorithms and convergence analysis. Insurance: Mathematics and Economics; 2021; 96, pp. 262-75. [DOI: https://dx.doi.org/10.1016/j.insmatheco.2020.11.012]
Johnson, Justin M.; Khoshgoftaar, Taghi M. Medicare fraud detection using neural networks. Journal of Big Data; 2019; 6, pp. 1-35. [DOI: https://dx.doi.org/10.1186/s40537-019-0225-0]
Joram, Mutai K.; Harrison, Bii K.; Joseph, Kiplang’at N. A knowledge-based system for life insurance underwriting. International Journal of Information Technology and Computer Science; 2017; 3, pp. 40-49. [DOI: https://dx.doi.org/10.5815/ijitcs.2017.03.05]
Josephson, John R.; Josephson, Susan G. Abductive Inference: Computation, Philosophy, Technology; Cambridge University Press: Cambridge, 1996.
Karamizadeh, Faramarz; Zolfagharifar, Seyed Ahad. Using the clustering algorithms and rule-based of data mining to identify affecting factors in the profit and loss of third party insurance, insurance company auto. Indian Journal of Science and Technology; 2016; 9, pp. 1-9. [DOI: https://dx.doi.org/10.17485/ijst/2016/v9i7/87846]
Kašćelan, Vladimir; Kašćelan, Ljiljana; Burić, Milijana Novović. A nonparametric data mining approach for risk prediction in car insurance: A case study from the Montenegrin market. Economic Research-Ekonomska Istraživanja; 2016; 29, pp. 545-58. [DOI: https://dx.doi.org/10.1080/1331677X.2016.1175729]
Keller, Benno; Eling, Martin; Schmeiser, Hato; Christen, Markus; Loi, Michele. Big Data and Insurance: Implications for Innovation, Competition and Privacy; Geneva Association-International Association for the Study of Insurance: Geneva, 2018.
Kelley, Kevin H.; Fontanetta, Lisa M.; Heintzman, Mark; Pereira, Nikki. Artificial intelligence: Implications for social inflation and insurance. Risk Management and Insurance Review; 2018; 21, pp. 373-87. [DOI: https://dx.doi.org/10.1111/rmir.12111]
Khodairy, Moayed A.; Abosamra, Gibrael. Driving Behavior Classification Based on Oversampled Signals of Smartphone Embedded Sensors Using an Optimized Stacked-LSTM Neural Networks. IEEE Access; 2021; 9, pp. 4957-72. [DOI: https://dx.doi.org/10.1109/ACCESS.2020.3048915]
Khuong, Mai Ngoc; Tuan, Tran Manh. A new neuro-fuzzy inference system for insurance forecasting. Paper presented at the International Conference on Advances in Information and Communication Technology; Bikaner, India, August 12–13; 2016.
Kiermayer, Mark; Weiß, Christian. Grouping of contracts in insurance using neural networks. Scandinavian Actuarial Journal; 2021; 2021, pp. 295-322. [DOI: https://dx.doi.org/10.1080/03461238.2020.1836676]
Kieu, Tung; Yang, Bin; Guo, Chenjuan; Jensen, Christian S. Distinguishing trajectories from different drivers using incompletely labeled trajectories. Paper presented at the 27th ACM International Conference on Information and Knowledge Management; Torino, Italy, October 22–26; 2018.
Kim, Hyong; Gardner, Errol. The Science of Winning in Financial Services-Competing on Analytics: Opportunities to Unlock the Power of Data. Journal of Financial Perspectives; 2015; 3, pp. 1-34.
Kopitar, Leon; Cilar, Leona; Kocbek, Primoz; Stiglic, Gregor. Local vs. global interpretability of machine learning models in type 2 diabetes mellitus screening. Artificial Intelligence in Medicine: Knowledge Representation and Transparent and Explainable Systems; Springer: Berlin, 2019; pp. 108-19.
Kose, Ilker; Gokturk, Mehmet; Kilic, Kemal. An interactive machine-learning-based electronic fraud and abuse detection system in healthcare insurance. Applied Soft Computing; 2015; 36, pp. 283-99. [DOI: https://dx.doi.org/10.1016/j.asoc.2015.07.018]
Koster, Harold. Towards better implementation of the European Union’s anti-money laundering and countering the financing of terrorism framework. Journal of Money Laundering Control; 2020; 23, pp. 379-86. [DOI: https://dx.doi.org/10.1108/JMLC-09-2019-0073]
Koster, Olivier; Kosman, Ruud; Visser, Joost. A Checklist for Explainable AI in the Insurance Domain. Paper presented at the International Conference on the Quality of Information and Communications Technology; Algarve, Portugal, September 8–11; 2021.
Kowshalya, G.; Nandhini, M. Predicting fraudulent claims in automobile insurance. Paper presented at the 2018 Second International Conference on Inventive Communication and Computational Technologies (ICICCT); Coimbatore, India, April 20–21; 2018.
Krafft, Peaks; Young, Meg; Katell, Michael; Huang, Karen; Bugingo, Ghislain. Defining AI in policy versus practice. Paper presented at the AAAI/ACM Conference on AI, Ethics, and Society; New York, NY, USA, February 7–8; 2020.
Kumar, Akshi; Dikshit, Shubham; Albuquerque, Victor Hugo C. Explainable Artificial Intelligence for Sarcasm Detection in Dialogues. Wireless Communications and Mobile Computing; 2021; 2021, 2939334. [DOI: https://dx.doi.org/10.1155/2021/2939334]
Kumar, Mohit; Ghani, Rayid; Mei, Zhu-Song. Data mining to predict and prevent errors in health insurance claims processing. Paper presented at the 16th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; Washington, DC, USA, July 24–28; 2010.
Kuo, Kevin; Lupton, Daniel. Towards Explainability of Machine Learning Models in Insurance Pricing. arXiv; 2020; arXiv: 2003.10674
Kute, Dattatray V.; Pradhan, Biswajeet; Shukla, Nagesh; Alamri, Abdullah. Deep learning and explainable artificial intelligence techniques applied for detecting money laundering—A critical review. IEEE Access; 2021; 9, pp. 82300-17. [DOI: https://dx.doi.org/10.1109/ACCESS.2021.3086230]
Kwak, Byung Il; Han, Mee Lan; Kim, Huy Kang. Driver Identification Based on Wavelet Transform Using Driving Patterns. IEEE Transactions on Industrial Informatics; 2020; 17, pp. 2400-10. [DOI: https://dx.doi.org/10.1109/TII.2020.2999911]
Kyu, Phyu Mar; Woraratpanya, Kuntpong. Car Damage Detection and Classification. Paper presented at the 11th International Conference on Advances in Information Technology; Bangkok, Thailand, July 1–3; 2020.
Larivière, Bart; Van den Poel, Dirk. Predicting customer retention and profitability by using random forests and regression forests techniques. Expert Systems with Applications; 2005; 29, pp. 472-84. [DOI: https://dx.doi.org/10.1016/j.eswa.2005.04.043]
Lau, Lucas; Tripathi, Arun. Mine your business—A novel application of association rules for insurance claims analytics. CAS E-Forum; Casualty Actuarial Society: Arlington, 2011.
Lee, Gee Y.; Manski, Scott; Maiti, Tapabrata. Actuarial applications of word embedding models. ASTIN Bulletin: The Journal of the IAA; 2020; 50, pp. 1-24. [DOI: https://dx.doi.org/10.1017/asb.2019.28]
Li, Yaqi; Yan, Chun; Liu, Wei; Li, Maozhen. A principle component analysis-based random forest with the potential nearest neighbor method for automobile insurance fraud identification. Applied Soft Computing; 2018; 70, pp. 1000-9. [DOI: https://dx.doi.org/10.1016/j.asoc.2017.07.027]
Liao, Shu-Hsien; Chu, Pei-Hui; Hsiao, Pei-Yuan. Data mining techniques and applications–A decade review from 2000 to 2011. Expert Systems with Applications; 2012; 39, pp. 11303-11. [DOI: https://dx.doi.org/10.1016/j.eswa.2012.02.063]
Lin, Chaohsin. Using neural networks as a support tool in the decision making for insurance industry. Expert Systems with Applications; 2009; 36, pp. 6914-17. [DOI: https://dx.doi.org/10.1016/j.eswa.2008.08.060]
Lin, Justin; Chang, Ha-Joon. Should Industrial Policy in developing countries conform to comparative advantage or defy it? A debate between Justin Lin and Ha-Joon Chang. Development Policy Review; 2009; 27, pp. 483-502. [DOI: https://dx.doi.org/10.1111/j.1467-7679.2009.00456.x]
Lin, Weiwei; Wu, Ziming; Lin, Longxin; Wen, Angzhan; Li, Jin. An ensemble random forest algorithm for insurance big data analysis. IEEE Access; 2017; 5, pp. 16568-75. [DOI: https://dx.doi.org/10.1109/ACCESS.2017.2738069]
Lipton, Zachary C. The Mythos of Model Interpretability: In machine learning, the concept of interpretability is both important and slippery. Queue; 2018; 16, pp. 31-57. [DOI: https://dx.doi.org/10.1145/3236386.3241340]
Liu, Jenn-Long; Chen, Chien-Liang. Application of evolutionary data mining algorithms to insurance fraud prediction. Paper presented at the 4th International Conference on Machine Learning and Computing IPCSIT; Hong Kong, China, March 10–11; 2012.
Liu, Qing; Pitt, David; Wu, Xueyuan. On the prediction of claim duration for income protection insurance policyholders. Annals of Actuarial Science; 2014; 8, pp. 42-62. [DOI: https://dx.doi.org/10.1017/S1748499513000134]
Lopez, Olivier; Milhaud, Xavier. Individual reserving and nonparametric estimation of claim amounts subject to large reporting delays. Scandinavian Actuarial Journal; 2021; 2021, pp. 34-53. [DOI: https://dx.doi.org/10.1080/03461238.2020.1793218]
Lou, Yin; Caruana, Rich; Gehrke, Johannes; Hooker, Giles. Accurate intelligible models with pairwise interactions. Paper presented at the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; Chicago, IL, USA, August 11–14; 2013.
Lundberg, Scott M; Lee, Su-In. A unified approach to interpreting model predictions. Paper presented at the 31st International Conference on Neural Information Processing Systems; Long Beach, CA, USA, December 4–9; 2017.
Lundberg, Scott M; Erion, Gabriel; Chen, Hugh; DeGrave, Alex; Prutkin, Jordan M; Nair, Bala; Katz, Ronit; Himmelfarb, Jonathan; Bansal, Nisha; Lee, Su-In. From local explanations to global understanding with explainable AI for trees. Nature Machine Intelligence; 2020; 2, pp. 56-67. [DOI: https://dx.doi.org/10.1038/s42256-019-0138-9]
Ma, Yu-Luen; Zhu, Xiaoyu; Hu, Xianbiao; Chiu, Yi-Chang. The use of context-sensitive insurance telematics data in auto insurance rate making. Transportation Research Part A: Policy and Practice; 2018; 113, pp. 243-58. [DOI: https://dx.doi.org/10.1016/j.tra.2018.04.013]
Mascharka, David; Tran, Philip; Soklaski, Ryan; Majumdar, Arjun. Transparency by design: Closing the gap between performance and interpretability in visual reasoning. Paper presented at the IEEE Conference on Computer Vision and Pattern Recognition; Salt Lake City, UT, USA, June 18–23; 2018.
Matloob, Irum; Khan, Shoab Ahmed; Rahman, Habib Ur. Sequence Mining and Prediction-Based Healthcare Fraud Detection Methodology. IEEE Access; 2020; 8, pp. 143256-73. [DOI: https://dx.doi.org/10.1109/ACCESS.2020.3013962]
Mayer, Roger C.; Davis, James H.; Schoorman, F. David. An integrative model of organizational trust. Academy of Management Review; 1995; 20, pp. 709-34. [DOI: https://dx.doi.org/10.2307/258792]
Maynard, Trevor; Baldassarre, Luca; de Montjoye, Yves-Alexandre; McFall, Liz; Óskarsdóttir, María. AI: Coming of age?. Annals of Actuarial Science; 2022; 16, pp. 1-5. [DOI: https://dx.doi.org/10.1017/S1748499521000245]
McFall, Liz; Meyers, Gert; Hoyweghen, Ine Van. The personalisation of insurance: Data, behaviour and innovation. Big Data & Society; 2020; 7, 2053951720973707.
McKnight, D. Harrison; Choudhury, Vivek; Kacmar, Charles. Developing and validating trust measures for e-commerce: An integrative typology. Information Systems Research; 2002; 13, pp. 334-59. [DOI: https://dx.doi.org/10.1287/isre.13.3.334.81]
Mehdiyev, Nijat; Houy, Constantin; Gutermuth, Oliver; Mayer, Lea; Fettke, Peter. Explainable Artificial Intelligence (XAI) Supporting Public Administration Processes–On the Potential of XAI in Tax Audit Processes; Springer: Cham, 2021.
Meyerson, Debra; Weick, Karl E.; Kramer, Roderick M. Swift trust and temporary groups. Trust in Organizations: Frontiers of Theory and Research; 1996; 166, 195.
Mizgier, Kamil J.; Kocsis, Otto; Wagner, Stephan M. Zurich Insurance uses data analytics to leverage the BI insurance proposition. Interfaces; 2018; 48, pp. 94-107. [DOI: https://dx.doi.org/10.1287/inte.2017.0928]
Mohamadloo, Azam; Ramezankhani, Ali; Zarein-Dolab, Saeed; Salamzadeh, Jamshid; Mohamadloo, Fatemeh. A systematic review of main factors leading to irrational prescription of medicine. Iranian Journal of Psychiatry and Behavioral Sciences; 2017; 11, e10242. [DOI: https://dx.doi.org/10.5812/ijpbs.10242]
Molnar, Christoph. Interpretable Machine Learning; Lulu Press: Morrisville, 2019.
Moradi, Milad; Samwald, Matthias. Post-hoc explanation of black-box classifiers using confident itemsets. Expert Systems with Applications; 2021; 165, 113941. [DOI: https://dx.doi.org/10.1016/j.eswa.2020.113941]
Morik, Katharina; Hüppej, Christian; Unterstein, Klaus. End-user access to multiple sources-Incorporating knowledge discovery into knowledge management. Paper presented at the International Conference on Practical Aspects of Knowledge Management; Vienna, Austria, December 2–3; 2002.
Motoda, Hiroshi; Liu, Huan. Feature selection, extraction and construction. Communication of IICM (Institute of Information and Computing Machinery, Taiwan); 2002; 5, 2.
Mueller, Shane T.; Hoffman, Robert R.; Clancey, William; Emrey, Abigail; Klein, Gary. Explanation in Human-AI Systems: A literature meta-review, synopsis of key ideas and publications, and bibliography for Explainable AI. arXiv; 2019; arXiv: 1902.01876
Mullins, Martin; Holland, Christopher P.; Cunneen, Martin. Creating ethics guidelines for artificial intelligence and big data analytics customers: The case of the consumer European insurance market. Patterns; 2021; 2, 100362. [DOI: https://dx.doi.org/10.1016/j.patter.2021.100362]
NallamReddy, Sundari; Behera, Samarandra; Karadagi, Sanjeev; Desik, A. Application of multiple random centroid (MRC) based k-means clustering algorithm in insurance—A review article. Operations Research and Applications: An International Journal; 2014; 1, pp. 15-21.
Naylor, Michael. Insurance Transformed: Technological Disruption; Springer: Berlin, 2017.
Neumann, Łukasz; Nowak, Robert M.; Okuniewski, Rafał; Wawrzyński, Paweł. Machine Learning-Based Predictions of Customers’ Decisions in Car Insurance. Applied Artificial Intelligence; 2019; 33, pp. 817-28. [DOI: https://dx.doi.org/10.1080/08839514.2019.1630151]
Ngai, Eric W. T.; Hu, Yong; Wong, Yiu Hing; Chen, Yijun; Sun, Xin. The application of data mining techniques in financial fraud detection: A classification framework and an academic review of literature. Decision Support Systems; 2011; 50, pp. 559-69. [DOI: https://dx.doi.org/10.1016/j.dss.2010.08.006]
Ntoutsi, Eirini; Fafalios, Pavlos; Gadiraju, Ujwal; Iosifidis, Vasileios; Nejdl, Wolfgang; Vidal, Maria-Esther; Ruggieri, Salvatore; Turini, Franco; Papadopoulos, Symeon; Krasanakis, Emmanouil. Bias in data-driven artificial intelligence systems—An introductory survey. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery; 2020; 10, e1356. [DOI: https://dx.doi.org/10.1002/widm.1356]
OECD, Organisation for Economic Co-operation and Development. The Impact of Big Data and Artificial Intelligence (AI) in the Insurance Sector; OECD: Paris, 2020; Available online: https://www.oecd.org/finance/Impact-Big-Data-AI-in-the-Insurance-Sector.pdf (accessed on 1 September 2021).
Page, Matthew J.; Moher, David. Evaluations of the uptake and impact of the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) Statement and extensions: A scoping review. Systematic Reviews; 2017; 6, pp. 1-14. [DOI: https://dx.doi.org/10.1186/s13643-017-0663-8] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29258593]
Palacio, Sebastian; Lucieri, Adriano; Munir, Mohsin; Hees, Jörn; Ahmed, Sheraz; Dengel, Andreas. XAI Handbook: Towards a Unified Framework for Explainable AI. arXiv; 2021; arXiv: 2105.06677
Paruchuri, Harish. The Impact of Machine Learning on the Future of Insurance Industry. American Journal of Trade and Policy; 2020; 7, pp. 85-90. [DOI: https://dx.doi.org/10.18034/ajtp.v7i3.537]
Pathak, Jagdish; Vidyarthi, Navneet; Summers, Scott L. A fuzzy-based algorithm for auditors to detect elements of fraud in settled insurance claims. Managerial Auditing Journal; 2005; 20, pp. 632-44. [DOI: https://dx.doi.org/10.1108/02686900510606119]
Payrovnaziri, Seyedeh Neelufar; Chen, Zhaoyi; Rengifo-Moreno, Pablo; Miller, Tim; Bian, Jiang; Chen, Jonathan H.; Liu, Xiuwen; He, Zhe. Explainable artificial intelligence models using real-world electronic health record data: A systematic scoping review. Journal of the American Medical Informatics Association; 2020; 27, pp. 1173-85. [DOI: https://dx.doi.org/10.1093/jamia/ocaa053]
Pieters, Wolter. Explanation and trust: What to tell the user in security and AI?. Ethics and Information Technology; 2011; 13, pp. 53-64. [DOI: https://dx.doi.org/10.1007/s10676-010-9253-3]
Putnam, Vanessa; Conati, Cristina. Exploring the Need for Explainable Artificial Intelligence (XAI) in Intelligent Tutoring Systems (ITS). Paper presented at the IUI Workshops; Los Angeles, CA, USA, March 16–20; 2019.
Quan, Zhiyu; Valdez, Emiliano A. Predictive analytics of insurance claims using multivariate decision trees. Dependence Modeling; 2018; 6, pp. 377-407. [DOI: https://dx.doi.org/10.1515/demo-2018-0022]
Ravi, Kumar; Ravi, Vadlamani; Prasad, P. Sree Rama Krishna. Fuzzy formal concept analysis based opinion mining for CRM in financial services. Applied Soft Computing; 2017; 60, pp. 786-807. [DOI: https://dx.doi.org/10.1016/j.asoc.2017.05.028]
Ribeiro, Marco Tulio; Singh, Sameer; Guestrin, Carlos. “Why should i trust you?” Explaining the predictions of any classifier. Paper presented at the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; San Francisco, CA, USA, August 13–17; 2016.
Rieder, Gernot; Simon, Judith. Big data: A new empiricism and its epistemic and socio-political consequences. Berechenbarkeit der Welt?; Springer: Berlin, 2017; pp. 85-105.
Riikkinen, Mikko; Saarijärvi, Hannu; Sarlin, Peter; Lähteenmäki, Ilkka. Using artificial intelligence to create value in insurance. International Journal of Bank Marketing; 2018; 36, pp. 1145-68. [DOI: https://dx.doi.org/10.1108/IJBM-01-2017-0015]
Robinson, Stephen Cory. Trust, transparency, and openness: How inclusion of cultural values shapes Nordic national public policy strategies for artificial intelligence (AI). Technology in Society; 2020; 63, 101421. [DOI: https://dx.doi.org/10.1016/j.techsoc.2020.101421]
Rosenfeld, Avi. Better metrics for evaluating explainable artificial intelligence. Paper presented at the 20th International Conference on Autonomous Agents and Multiagent Systems; London, UK, May 3–7; 2021.
Rudin, Cynthia. Please stop explaining black box models for high stakes decisions. Stat; 2018; 1050, 26.
Sadreddini, Zhaleh; Donmez, Ilknur; Yanikomeroglu, Halim. Cancel-for-Any-Reason Insurance Recommendation Using Customer Transaction-Based Clustering. IEEE Access; 2021; 9, pp. 39363-74. [DOI: https://dx.doi.org/10.1109/ACCESS.2021.3064929]
Sakthivel, K. M.; Rajitha, C. S. Artificial intelligence for estimation of future claim frequency in non-life insurance. Global Journal of Pure and Applied Mathematics; 2017; 13, pp. 1701-10.
Samonte, Mary Jane C.; Gerardo, Bobby D.; Fajardo, Arnel C.; Medina, Ruji P. ICD-9 tagging of clinical notes using topical word embedding. Paper presented at the 2018 International Conference on Internet and e-Business; Singapore, April 25–27; 2018.
Sarkar, Abhineet. Disrupting the Insurance Value Chain. The AI Book: The Artificial Intelligence Handbook for Investors, Entrepreneurs and FinTech Visionaries; Wiley: New York, 2020; pp. 89-91.
Sevim, Şerafettin; Yildiz, Birol; Dalkiliç, Nilüfer. Risk Assessment for Accounting Professional Liability Insurance. Sosyoekonomi; 2016; 24, pp. 93-112. [DOI: https://dx.doi.org/10.17233/se.2016.06.004]
Shah, Paras; Guez, Allon. Mortality forecasting using neural networks and an application to cause-specific data for insurance purposes. Journal of Forecasting; 2009; 28, pp. 535-48. [DOI: https://dx.doi.org/10.1002/for.1111]
Shapiro, Arnold F. An overview of insurance uses of fuzzy logic. Computational Intelligence in Economics and Finance; Springer: Berlin, 2007; pp. 25-61.
Sheehan, Barry; Murphy, Finbarr; Ryan, Cian; Mullins, Martin; Liu, Hai Yue. Semi-autonomous vehicle motor insurance: A Bayesian Network risk transfer approach. Transportation Research Part C: Emerging Technologies; 2017; 82, pp. 124-37. [DOI: https://dx.doi.org/10.1016/j.trc.2017.06.015]
Siami, Mohammad; Naderpour, Mohsen; Lu, Jie. A mobile telematics pattern recognition framework for driving behavior extraction. IEEE Transactions on Intelligent Transportation Systems; 2020; 22, pp. 1459-72. [DOI: https://dx.doi.org/10.1109/TITS.2020.2971214]
Siau, Keng; Wang, Weiyu. Building trust in artificial intelligence, machine learning, and robotics. Cutter Business Technology Journal; 2018; 31, pp. 47-53.
Siegel, Magdalena; Assenmacher, Constanze; Meuwly, Nathalie; Zemp, Martina. The legal vulnerability model for same-sex parent families: A mixed methods systematic review and theoretical integration. Frontiers in Psychology; 2021; 12, 683. [DOI: https://dx.doi.org/10.3389/fpsyg.2021.644258] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33796052]
Sithic, H. Lookman; Balasubramanian, T. Survey of insurance fraud detection using data mining techniques. arXiv; 2013; arXiv: 1309.0806
Smith, Kate A.; Willis, Robert J.; Brooks, Malcolm. An analysis of customer retention and insurance claim patterns using data mining: A case study. Journal of the Operational Research Society; 2000; 51, pp. 532-41. [DOI: https://dx.doi.org/10.1057/palgrave.jors.2600941]
Smyth, Gordon K.; Jørgensen, Bent. Fitting Tweedie’s compound Poisson model to insurance claims data: Dispersion modelling. ASTIN Bulletin: The Journal of the IAA; 2002; 32, pp. 143-57. [DOI: https://dx.doi.org/10.2143/AST.32.1.1020]
Sohail, Misbah; Peres, Pedro; Li, Yuhua. Feature importance analysis for customer management of insurance products. Paper presented at the 2021 International Joint Conference on Neural Networks (IJCNN); Shenzhen, China, July 18–22; 2021.
Srihari, Sargur. Explainable Artificial Intelligence: An Overview. Journal of the Washington Academy of Sciences; 2020.
Stovold, Elizabeth; Beecher, Deirdre; Foxlee, Ruth; Noel-Storr, Anna. Study flow diagrams in Cochrane systematic review updates: An adapted PRISMA flow diagram. Systematic Reviews; 2014; 3, pp. 1-5. [DOI: https://dx.doi.org/10.1186/2046-4053-3-54]
Štrumbelj, Erik; Kononenko, Igor. Explaining prediction models and individual predictions with feature contributions. Knowledge and Information Systems; 2014; 41, pp. 647-65. [DOI: https://dx.doi.org/10.1007/s10115-013-0679-x]
Sun, Chenfei; Yan, Zhongmin; Li, Qingzhong; Zheng, Yongqing; Lu, Xudong; Cui, Lizhen. Abnormal group-based joint medical fraud detection. IEEE Access; 2018; 7, pp. 13589-96. [DOI: https://dx.doi.org/10.1109/ACCESS.2018.2887119]
Supraja, K.; Saritha, S. Jessica. Robust fuzzy rule based technique to detect frauds in vehicle insurance. Paper presented at the 2017 International Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS); Chennai, India, August 1–2; 2017.
Tallant, Jonathan. Commitment in cases of trust and distrust. Thought; 2017; 6, pp. 261-267. [DOI: https://dx.doi.org/10.1002/tht3.259]
Tanninen, Maiju. Contested technology: Social scientific perspectives of behaviour-based insurance. Big Data & Society; 2020; 7, 2053951720942536.
Tao, Han; Zhixin, Liu; Xiaodong, Song. Insurance fraud identification research based on fuzzy support vector machine with dual membership. Paper presented at the 2012 International Conference on Information Management, Innovation Management and Industrial Engineering; Sanya, China, October 20–21; 2012.
Taylor, Linnet. What is data justice? The case for connecting digital rights and freedoms globally. Big Data & Society; 2017; 4, 2053951717736335.
Tekaya, Balkiss; Feki, Sirine El; Tekaya, Tasnim; Masri, Hela. Recent applications of big data in finance. Paper presented at the 2nd International Conference on Digital Tools & Uses Congress; Virtual Event, October 15–17; 2020.
Tillmanns, Sebastian; Hofstede, Frenkel Ter; Krafft, Manfred; Goetz, Oliver. How to separate the wheat from the chaff: Improved variable selection for new customer acquisition. Journal of Marketing; 2017; 81, pp. 99-113. [DOI: https://dx.doi.org/10.1509/jm.15.0398]
Tonekaboni, Sana; Joshi, Shalmali; McCradden, Melissa D; Goldenberg, Anna. What clinicians want: Contextualizing explainable machine learning for clinical end use. Paper presented at the Machine Learning for Healthcare Conference; Ann Arbor, MI, USA, August 9–10; 2019.
Toreini, Ehsan; Aitken, Mhairi; Coopamootoo, Kovila; Elliott, Karen; Zelaya, Carlos Gonzalez; Moorsel, Aad Van. The relationship between trust in AI and trustworthy machine learning technologies. Paper presented at the 2020 Conference on Fairness, Accountability, and Transparency; Barcelona, Spain, January 27–30; 2020.
Umamaheswari, K.; Janakiraman, S. Role of data mining in insurance industry. Int J Adv Comput Technol; 2014; 3, pp. 961-66.
Ungur, Cristina. Socio-economic valences of insurance. Revista Economia Contemporană; 2017; 2, pp. 112-18.
van den Boom, Freyja. Regulating Telematics Insurance. Insurance Distribution Directive; Springer: Berlin, 2021; pp. 293-325.
Vassiljeva, Kristina; Tepljakov, Aleksei; Petlenkov, Eduard; Netšajev, Eduard. Computational intelligence approach for estimation of vehicle insurance risk level. Paper presented at the 2017 International Joint Conference on Neural Networks (IJCNN); Anchorage, AK, USA, May 14–19; 2017.
Vaziri, Jalil; Beheshtinia, Mohammad Ali. A holistic fuzzy approach to create competitive advantage via quality management in services industry (case study: Life-insurance services). Management Decision; 2016; 54, pp. 2035-62. [DOI: https://dx.doi.org/10.1108/MD-11-2015-0535]
Verma, Aayushi; Taneja, Anu; Arora, Anuja. Fraud detection and frequent pattern matching in insurance claims using data mining techniques. Paper presented at the 2017 Tenth International Conference on Contemporary Computing (IC3); Noida, India, August 10–12; 2017.
Viaene, Stijn; Dedene, Guido; Derrig, Richard A. Auto claim fraud detection using Bayesian learning neural networks. Expert Systems with Applications; 2005; 29, pp. 653-66. [DOI: https://dx.doi.org/10.1016/j.eswa.2005.04.030]
Viaene, Stijn; Derrig, Richard A.; Dedene, Guido. A case study of applying boosting Naive Bayes to claim fraud diagnosis. IEEE Transactions on Knowledge and Data Engineering; 2004; 16, pp. 612-20. [DOI: https://dx.doi.org/10.1109/TKDE.2004.1277822]
Viaene, Stijn; Derrig, Richard A.; Baesens, Bart; Dedene, Guido. A comparison of state-of-the-art classification techniques for expert automobile insurance claim fraud detection. Journal of Risk and Insurance; 2002; 69, pp. 373-421. [DOI: https://dx.doi.org/10.1111/1539-6975.00023]
Vilone, Giulia; Longo, Luca. Explainable artificial intelligence: A systematic review. arXiv; 2020; arXiv: 2006.00093
von Eschenbach, Warren J. Transparency and the black box problem: Why we do not trust AI. Philosophy & Technology; 2021; 34, pp. 1607-22.
Walsh, Nigel; Taylor, Mike. Cutting to the Chase: Mapping AI to the Real-World Insurance Value Chain. The AI Book: The Artificial Intelligence Handbook for Investors, Entrepreneurs and FinTech Visionaries; Wiley: New York, 2020; pp. 92-97.
Wang, Hui Dong. Research on the Features of Car Insurance Data Based on Machine Learning. Procedia Computer Science; 2020; 166, pp. 582-87. [DOI: https://dx.doi.org/10.1016/j.procs.2020.02.016]
Wang, Yibo; Xu, Wei. Leveraging deep learning with LDA-based text analytics to detect automobile insurance fraud. Decision Support Systems; 2018; 105, pp. 87-95. [DOI: https://dx.doi.org/10.1016/j.dss.2017.11.001]
Wei, Cheng; Dan, Li. Market fluctuation and agricultural insurance forecasting model based on machine learning algorithm of parameter optimization. Journal of Intelligent & Fuzzy Systems; 2019; 37, pp. 6217-28.
Wulf, Alexander J.; Seizov, Ognyan. “Please understand we cannot provide further information”: Evaluating content and transparency of GDPR-mandated AI disclosures. AI & Society; 2022; pp. 1-22. [DOI: https://dx.doi.org/10.1007/s00146-022-01424-z]
Wüthrich, Mario V. Machine learning in individual claims reserving. Scandinavian Actuarial Journal; 2018; 2018, pp. 465-80. [DOI: https://dx.doi.org/10.1080/03461238.2018.1428681]
Wüthrich, Mario V. Bias regularization in neural network models for general insurance pricing. European Actuarial Journal; 2020; 10, pp. 179-202. [DOI: https://dx.doi.org/10.1007/s13385-019-00215-z]
Xiao, Bo; Benbasat, Izak. E-commerce product recommendation agents: Use, characteristics, and impact. MIS Quarterly; 2007; 31, pp. 137-209. [DOI: https://dx.doi.org/10.2307/25148784]
Xie, Ning; Ras, Gabrielle; van Gerven, Marcel; Doran, Derek. Explainable deep learning: A field guide for the uninitiated. arXiv; 2020; arXiv: 2004.14545
Xu, Wei; Wang, Shengnan; Zhang, Dailing; Yang, Bo. Random rough subspace based neural network ensemble for insurance fraud detection. Paper presented at the 2011 Fourth International Joint Conference on Computational Sciences and Optimization; Kunming, China, April 15–19; 2011.
Yan, Chun; Li, Meixuan; Liu, Wei; Qi, Man. Improved adaptive genetic algorithm for the vehicle Insurance Fraud Identification Model based on a BP Neural Network. Theoretical Computer Science; 2020a; 817, pp. 12-23. [DOI: https://dx.doi.org/10.1016/j.tcs.2019.06.025]
Yan, Chun; Wang, Xindong; Liu, Xinhong; Liu, Wei; Liu, Jiahui. Research on the UBI Car Insurance Rate Determination Model Based on the CNN-HVSVM Algorithm. IEEE Access; 2020b; 8, pp. 160762-73. [DOI: https://dx.doi.org/10.1109/ACCESS.2020.3021062]
Yan, Weizhong; Bonissone, Piero P. Designing a Neural Network Decision System for Automated Insurance Underwriting. Paper presented at the 2006 IEEE International Joint Conference on Neural Network Proceedings; Vancouver, BC, Canada, July 16–21; 2006.
Yang, Qiang; Yin, Jie; Ling, Charles; Pan, Rong. Extracting actionable knowledge from decision trees. IEEE Transactions on Knowledge and Data Engineering; 2006; 19, pp. 43-56. [DOI: https://dx.doi.org/10.1109/TKDE.2007.250584]
Yang, Yi; Qian, Wei; Zou, Hui. Insurance premium prediction via gradient tree-boosted Tweedie compound Poisson models. Journal of Business & Economic Statistics; 2018; 36, pp. 456-70.
Yeo, Ai Cheo; Smith, Kate A.; Willis, Robert J.; Brooks, Malcolm. A mathematical programming approach to optimise insurance premium pricing within a data mining framework. Journal of the Operational Research Society; 2002; 53, pp. 1197-203. [DOI: https://dx.doi.org/10.1057/palgrave.jors.2601413]
Yeung, Karen; Howes, Andrew; Pogrebna, Ganna. AI governance by human rights-centred design, deliberation and oversight: An end to ethics washing. The Oxford Handbook of AI Ethics; Oxford University Press: Oxford, 2019.
Zahi, Sara; Achchab, Boujemâa. Clustering of the population benefiting from health insurance using K-means. Paper presented at the 4th International Conference on Smart City Applications; Casablanca, Morocco, October 2–4; 2019.
Zarifis, Alex; Holland, Christopher P.; Milne, Alistair. Evaluating the impact of AI on insurance: The four emerging AI-and data-driven business models. Emerald Open Research; 2019; 1, 15. [DOI: https://dx.doi.org/10.35241/emeraldopenres.13249.1]
Zhang, Bo; Kong, Dehua. Dynamic estimation model of insurance product recommendation based on Naive Bayesian model. Paper presented at the 2020 International Conference on Cyberspace Innovation of Advanced Technologies; Guangzhou, China, December 4–6; 2020.
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Sheehan, Barry 1
; Mullins, Martin 1 ; Cunneen, Martin 1 ; Ressel, Juliane 2 ; Castignani, German 3 1 Department of Accounting and Finance, University of Limerick, V94 PH93 Limerick, Ireland
2 Department of Accounting and Finance, University of Limerick, V94 PH93 Limerick, Ireland; Research Center for the Insurance Market, Institute for Insurance Studies, TH Köln, 50968 Cologne, Germany
3 Motion-S S.A., Avenue des Bains 4, Mondorf-les-Bains, L-5610 Luxembourg, Luxembourg; Faculty of Science, Technology and Medicine (FSTM), University of Luxembourg, Esch-sur-Alzette, L-4365 Luxembourg, Luxembourg