Content area
Purpose
Artificial intelligence (AI) is a significant game changer in human resource development (HRD). The launch of ChatGPT has accelerated its progress and amplified its impact on organizations and employees. This study aims to review and examine literature on AI in HRD, using a bibliometric approach.
Design/methodology/approach
This study is a bibliometric review. Scopus was used to identify studies in the field. In total, 236 papers published in the past 10 years were examined using the VOSviewer program.
Findings
The obtained results showed that most cited documents and authors are mainly from computer sciences, emphasizing machine learning over human learning. While it was expected that HRD authors and studies would have a more substantial presence, the lesser prominence suggests several interesting avenues for explorations.
Practical implications
This study provides insights and recommendations for researchers, managers, HRD practitioners and policymakers. Prioritizing the development of both humans and machines becomes crucial, as an exclusive focus on machines may pose a risk to the sustainability of employees' skills and long-term career prospects.
Originality/value
There is a dearth of bibliometric studies examining AI in HRD. Hence, this study proposes a relatively unexplored approach to examine this topic. It provides a visual and structured overview of this topic. Also, it highlights areas of research concentration and areas that are overlooked. Shedding light on the presence of more research originating from computer sciences and focusing on machine learning over human learning represent an important contribution of this study, which may foster interdisciplinary collaboration with experts from diverse fields, broadening the scope of research on technologies and learning in workplaces.
1. Introduction
Artificial intelligence (AI) is an interdisciplinary field of study that combines computer science, engineering, mathematics, psychology, linguistics and other related areas (Dwivedi et al., 2023). It involves several cognitive technologies, which vary based on their complexity, such as robotic process automation and machine learning (Yorks et al., 2020). Its primary aim revolves around the creation of intelligent machines capable of emulating and executing tasks akin to human intelligence, encompassing domains such as visual perception, speech recognition, decision-making and language translation (Lund et al., 2023). The landscape of AI is in a perpetual state of evolution, consistently giving rise to novel subdomains like machine learning, natural language processing, robotics and computer vision (Ogurlu et al., 2021). It has rapidly revolutionized business and society (Malik et al., 2022; Hamouche, 2021). Within organizations, AI's role lies in the automation of routine tasks and processes (Wang et al., 2023). It analyses massive data sets faster and more accurately than humans (George and George, 2023), which improves decision-making (Rodgers et al., 2023) and enhances the overall efficiency and effectiveness of business operations (Chowdhury et al., 2023).
More recently, the launch of ChatGPT, i.e. generative pre-trained transformer, the latest in the stream of AI developments, generated a media buzz, due to its ability to human-like answers to almost any question asked (Budhwar et al., 2022). It significantly reduces the distance between AI and humans using it. Besides shedding light on the fast progression of AI, which represents a significant step forward in bringing it to the public's attention, ChatGPT raised questions and concerns about its potential impacts on employees’ jobs and skills. The fast progress of this new AI technology generated a collective call to pause its development to fully examine and consider these impacts, which are still unclear and uncertain (Budhwar et al., 2022). According to the International Labour Organization (ILO, 2023), Generative AI (ChatGPT) will not necessarily destroy jobs, but it will rather complement them. Meanwhile, practitioners suggest that the evolving landscape of artificial intelligence is reshaping the skillsets sought by employers from their workforce (Gurchiek, 2023). According to the survey of Salesforce (2023), on a sample of 11,035 working adults across 11 countries, such as Australia, France, Germany, Italy, The Netherlands and Singapore, 98% of leaders believe the shift to skills-based hiring will benefit business; however, only one in ten of workers declare that they have AI skills, while they are the most required digital skills. Therefore, AI is a significant game changer in human resource development (HRD). Hence, it is important to develop research that will provide the necessary support for organizations, specifically managers and HRD practitioners.
Most studies examine the impact of artificial intelligence on human resource management (Palos-Sánchez et al., 2022; Kaur et al., 2021; Kaushal and Ghalawat, 2023); however, very few addressed it in HRD. Indeed, the distinction between HRM and HRD is blurring (Alagaraja, 2013) and the general public often makes little distinction between them (Werner, 2014); however, these two fields differ in terms of the definition of HR and its role in the organization (Alagaraja, 2013). HRD focuses on training design, delivery, and evaluation (Alagaraja, 2013). It is “a process for developing and unleashing human expertise through organization development and personnel training and development for the purpose of improving performance” (Swanson, 1995). Whereas, HRM has a strong managerial orientation (Alagaraja, 2013). It refers to “the design and management of human resource systems based on employment policy, comprising a set of policies designed to maximize organizational integration, employee commitment, flexibility, and quality of work” (Alagaraja, 2013). Despite this difference, these two fields are closely related (Harrison et al., 2021) and are complementary (Alagaraja, 2013). Both address human resources in an organization (Alagaraja, 2013). Given the complexity of the organizational context (Shamout et al., 2022), concentrating primarily on AI and HRM, with a limited volume of research dedicated to AI in HRD might not help fully capture and understand how AI advances can influence HR in organizations. Furthermore, considering that AI will profoundly alter jobs and skills, including those of HR professionals (Budhwar et al., 2022), it is imperative to gain a deeper understanding of the knowledge structure of this field and how it integrates AI. According to Budhwar et al. (2022), the growing adoption of AI is likely to generate a shift in skills and competencies expected from HR professionals.
The objective of this study is to present a complete overview of artificial intelligence in HRD, using a review based on bibliometric analyses. To the best of our knowledge, there are no bibliometric studies on this topic. In fact, the existing review studies are limited in scope, primarily falling into either general literature reviews or integrative reviews with a specific focus. For example, the general literature review by Ardichvili (2022) focused on knowledge-intensive profession and accounting, while the integrative literature reviews by Chang and Ke (2023) examined socially responsible AI in the context of public awareness and HRD, using sustainability concepts. Despite their significant and valuable contribution, being focused on specific aspects of AI and HRD, existing review studies do not provide a holistic overview of the entire landscape of the examined topic. For instance, the existing reviews lack insights into the keywords and trends associated with AI in HRD. Hence, a bibliometric analysis is needed to gain a more thorough understanding and bridge these gaps. This methodological approach enables a more in-depth exploration and synthesis of the existing research, offering insights beyond the limitations of existing review studies. Bibliometric analyses are recognized for their ability to provide a macroscopic overview of several pieces of academic literature, which is not possible to get through traditional literature review (Guo, 2022). Therefore, this study makes a significant contribution not only by expanding research on AI and human resources in organizations and adding evidence to it but also by providing additional evidence to existing studies on AI in HRD and broadening their scope through bibliometric analysis. This method allows for a more comprehensive exploration of this topic within the field of HRD and a deeper understanding of its knowledge structure.
In this context, the following research questions are examined in alignment with the purpose of this study:
The article is organized into four sections. The current Section 1 presented is an introduction. Section 2 explains the bibliometric methodology used in this paper. Section 3 presents the obtained results. Finally, Section 4 is a discussion of the main findings, theoretical and practical implications, limitations and future research recommendations.
2. Methodology
This study is a review using bibliometric analyses, which gained traction across various scientific fields in recent times (Yilmaz et al., 2020), due to their capacity to facilitate the exploration of large sets of aggregated bibliographic data, including published journal articles and their corresponding citations. They enable the identification of article keyword occurrences and co-citations showing the structure of the examined topic, as well as topic clusters and their relationship with each other, by mapping them visually (Van Eck and Waltman, 2010). Bibliometric analyses are used to discover the intellectual structure of a research field and topic, through authorship analysis, journal analysis, keyword occurrences and co-citations (García-Lillo et al., 2017).
Search strategy and data collection
Data analyzed were collected from the Scopus collection database, which is widely adopted in bibliometric analyses. Scopus stands as the world's most extensive multidisciplinary curated abstract and citation (Thelwall and Sud, 2022). It covers 240 disciplines, with more than 84 million records and content from more than 7,000 publishers (Elsevier, 2023). Pranckutė (2021) and Singh et al. (2021) offer a more comprehensive overview of Scopus and its points of comparison with other analogous databases.
After determining the literature scope. The following keywords were identified for data collection: The search words were (“artificial intelligence” OR “AI” OR “ChatGPT” OR “generative artificial intelligence” OR “generative AI”) AND (“human resource development” OR “HRD” OR “employee development” OR “employee training” OR “skills development” OR “competencies development” OR “Personnel training” OR “workforce training and development” OR “Staff training and development” OR “Employee training and development” OR “Staff development” OR “Employee development” OR “Talent development” OR “employee skilling” OR “employee upskilling” OR “employee reskilling” OR “staff skilling” OR “staff upskilling” OR “staff reskilling” OR “worker skilling” OR “worker upskilling” OR “worker reskilling” OR “employee skills development”) in the search field (Article Title, Abstract and Keywords) for the period between 2013and 2023, to captures the most up-to-date research output. Given that AI is a field that spans multiple disciplines, including computer science, psychology and management, the selected subject areas were computer science; business, management and accounting; social sciences; decision sciences; and psychology. Other subject areas were excluded, such are physics and astronomy, medicine, chemistry and energy. The search strategy was limited to articles, excluding conference papers, books and book papers. In this case, articles written in English were retained to cover a large number of publications. Peer-reviewed journal articles are recognized to be best suitable for bibliometric analysis, and specifically citation analysis, rather than other types of publications, such as books, book chapters and conference proceedings (De Bellis, 2009; Kate et al., 2019), for which bibliometric data are often not available or incomplete. A total of 509 papers were displayed. Then a filter was applied based on the aforementioned selected areas and language, to refine the search of relevant papers, resulting in a total of 379 articles, which were exported from the Scopus database in their respective formats into an Excel Table file. Another round of refinement has been performed to examine the relevance of selected papers based on the title and abstract, resulting in the inclusion of 236 papers that were analyzed in this study. Figure 1 summarizes the process of identification of articles in Scopus database.
Data analysis
In this study, data analyses were performed with the VOSviewer program, The VOSviewer program was used to analyze the data. Besides providing a visual network, this program helps generate maps based on citation and co-citation analyses (Yilmaz et al., 2020). Also, it encompasses a robust graphical interface that enables the creation of maps representing the connections of each analysis unit (Feng et al., 2017).
3. Findings
Thematic analysis: keywords co-occurrence
VOSviewer software was used to analyze extensively the keywords mostly used in the dataset. The two options “co-occurrence” and “all keywords” were selected. The minimum number of occurrences of a keyword is 10. A total of 29 keywords were displayed. This analysis provides a broader view of global research trends and patterns within a specific field or domain. It can help identify emerging topics, interdisciplinary connections and dominant themes (Zhang and Eichmann-Kalwara, 2019). Table 1 provides the list of co-occurrence keywords and their total link strength, i.e. the cumulative strength of connections between keywords within a data set, which helps quantify the frequency and strength of associations or relationships between these keywords based on their co-occurrence in a set of documents. Figure 2 provides keyword co-occurrence information in VOSviewer to determine the hotspots of AI and HRD.
Keyword co-occurrence analysis resulted in the following four clusters:
Cluster 1. The keywords in Cluster 1 with the highest occurrence are “artificial intelligence” and “personnel training”, which also have many links to keywords in other clusters, suggesting that keywords function like bridges between this and other clusters. Other Keywords of this cluster are “big data”, “decision-making”, “decision support system”, “Human resource management”, “learning algorithms”, “learning systems”, “machine learning” and “optimization”. These keywords suggest a strong emphasis on the integration of advanced technologies, particularly artificial intelligence (AI), big data, machine learning and learning algorithms to enhance training and learning experiences for employees or personnel within organizations, as well as human resource management, with a focus on data-driven decision-making processes within HRM, where AI and machine learning techniques are likely used to optimize HR-related decisions.
In recent years, HRD has been interconnected with artificial intelligence programs, leading to the advent and development of new relationships between humans and smart machines (Su et al., 2021). AI has the potential to improve learning and development in organizations (Bhatt and Muduli, 2022) and make employee training more individualized and effective by using adaptive learning algorithms tailored to the specific requirements and preferences of each employee (Huang et al., 2023). AI algorithms have the capability to identify knowledge gaps among employees and subsequently offer tailored training programs to address them (McWhorter, 2023). Machine learning and natural language processing can automate HR tasks, analyze massive employee data, optimize decision-making processes and personalize employees’ learning and development. Furthermore, AI-based chatbots and virtual assistants can help employees learn at their own pace and meet their development needs (Božić, 2023). Hence, these technologies can make HRD more agile and responsive, empowering employees to grow professionally (Dirani et al., 2020). AI can also influence coaching practices (Graßmann and Schermuly, 2021; Bridgeman and Giraldez-Hayes, 2023). Graßmann and Schermuly (2021) referred to AI coaching, i.e. a machine-assisted, systematic process that helps clients fix objectives and find solutions to achieve them based on AI learning that does not require human guidance. In this case, machine learning will not be limited to coaching process data but also to the interaction with the same client (Graßmann and Schermuly, 2021).
The use of AI in the training and development process can improve the perception related to the time spent on training and reduce the lack of attention in training compared to traditional methods (Kambur and Akar, 2022). However, other studies suggest that automation can result in the loss of expertise because it reduces opportunities to learn from it (Ardichvili, 2022). Consequently, HRD practitioners should create and develop alternative individual development opportunities as well as an organizational culture that enhances expertise development in a human-machine interaction context. AI-powered software can also report on organizational processes (Chowdhury et al., 2023). This helps managers make data-driven decisions (Kollmann et al., 2023).
Cluster 2. Cluster 2 is the other dominant cluster. The keywords “students” and “teaching” dominate this cluster. Followed by other keywords, “active learning”, “curricula”, “e-learning”, “educational computing”, “engineering education” and “teachers”, indicating that this cluster is likely centered on the use of technology in pedagogical approaches, and online learning. In addition to a focus on innovative teaching methods and curriculum development. This cluster underscores the pivotal shift in the way organizations perceive and implement HRD strategies, heavily influenced by AI technologies. It suggests that traditional HRD methods are evolving rapidly to meet the demands of the contemporary workforce. The wide adoption of online learning affects training both in academic institutions and organizations, in different sectors, such as education and engineering (Zemliansky, 2021). The transition to online environments can be significantly affected by AI (Bennett and McWhorter, 2022). This transition requires shifting the teaching philosophies and methods as well as the development of new skills necessary for instructors to be effective online. According to Magana et al. (2022), learning strategies can enhance the success of e-learning. In this case, active learning can be an effective strategy (Zemliansky, 2021). Active learning is a pedagogical approach that fosters learners’ involvement and engagement with subject matter through activities like experimentation, group discussions and role-play (Lughofer, 2017). Literature suggests that learning strategies are not limited to humans, they are also used for machines and robots (Tseng et al., 2017). In fact, AI operates through machine learning, i.e. the ability of the system to improve performance without the need for humans. Hence, the pedagogical approach of active learning has been transferred to the machine learning and data mining community, aiming to give control to the model, identifying the data samples from which further learning can be derived for the machine/robots’ self-improvement (Lughofer, 2017; Tseng et al., 2017). In this case, robots can be able to personalize their interactions with humans and learn human responses from real-world interactions (Tseng et al., 2017; Zhang, 2023).
Cluster 3. The keywords “deep learning” and “support vector machine” are most dominant in this cluster. Other keywords are included, such as job analysis, task analysis, training and virtual reality. This cluster appears to be oriented toward advanced machine-learning techniques. Keywords like “deep learning” and “support vector machine” point to sophisticated learning algorithms and their applications. While “Virtual reality” may suggest a connection to immersive learning experiences and training. In addition, “Job analysis” and “task analysis” might indicate the use of data-driven approaches for job-related assessments and training. Deep learning feeds automatic systems, which can effectively monitor learners' experience, such as their emotional reactions to a learning process, which can assist instructors with the emotional management of their classes to maximize learners’ motivation and engagement (Fernández Herrero et al., 2023). While support vector machines can be used to improve how a computer learns (training data) (Liu et al., 2017). In this case, task analysis can refer both to employees and machines. For example, some studies (Yu et al., 2018) examined how reinforcement learning can help computer programs to learn how to do tasks on their own through interaction with their environment. Job analysis can assist AI by providing the necessary data and insights to understand job roles, tasks, and requirements, which are essential for developing AI-driven automation (Gonzalez et al., 2019). Moreover, other studies suggest that technologies such as virtual reality (VR) and augmented reality (AR) have the potential to deliver training experiences that are both immersive and interactive (Al-Ansi et al., 2023; Jacobsen et al., 2022). AR refers to a technology that overlays digital information, such as images, videos or 3 D models, onto the real-world environment. It combines elements of the physical world with computer-generated data to enhance the user's perception and interaction with their surroundings (Al-Ansi et al., 2023). While VR is a technology that immerses users in a simulated, computer-generated environment that can be similar to or entirely different from the real world. It involves wearing a head-mounted display (HMD) or using other sensory input devices like gloves or motion trackers (Al-Ansi et al., 2023).
Cluster 4. The most dominant keywords of Cluster 4 are “article” and “education.” Other keywords included in this cluster are “human,” “humans” and “learning.” This cluster is more general, encompassing various aspects of learning and human behavior. the prevalence of the keyword “article,” which may indicate a focus on literature related to these topics. Authors who consider that AI can enhance human welfare and learning recommend the evolving of AI into human-centered AI, especially in education, to enhance human learning and productivity (Yang et al., 2021). In this case, precision education is recommended (Yang, 2021), involving a process where learning behaviors, environment and strategies can be analyzed through diagnosis, predictions, treatment and prevention to determine solutions (Yang, 2021), which will be facilitated by a human-centered AI (Yang et al., 2021). Learning and education of humans are key drivers of personal growth, professional development and societal progress. These processes empower individuals to acquire new knowledge, skills and perspectives, enabling them to adapt to changing environments, make informed decisions and contribute meaningfully to their communities and the world at large (Hennessy et al., 2022). The pursuit of knowledge and lifelong learning is central to human advancement.
Citation analyses
The most cited documents.
To identify the most cited documents, the options “citation” and “document” were chosen, and the minimum citation number was set to 50. The total number displayed was 28. The list of top 10 obtained documents is presented in Table 2. According to the obtained results, The most cited documents were those of Youyou et al. (2015) (584 citations); Vapnik and Izmailov (2015) (252 citations); Jiang et al. (2017) (158 citations); Chen et al. (2019) (133); Sousa and Rocha (2019) (130 citations); Manita et al. (2020) (94 citations); Liu et al. (2017) (84 citations); Rana et al. (2022) (79 citations); Santos (2016) (54 citations); and Peng et al. (2019) (53 citations).
4.2.6 Author co-citation analysis. Co-citation analysis helps to identify influential articles and their authors (White and McCain, 1998). It reveals the frequency of two documents cited together by other documents. In this case, the strength increases when the frequency of citation of these documents together gets higher (Small, 1973). With a minimum number of co-citations set to 10, a total of 154 authors was displayed and visualized in Figure 3, which exhibits intellectual connections among authors. Based on the obtained results, with 51citations and 1,912 total link strength, Bengio is the most cited author, followed by Wang (47 citations, 1,382 total link strength); Zhang (42 citations, 1,139 total link strength); and Li (39 citations, 1,021 total link strength). These authors are mainly from computer science and engineering fields.
The most cited journals.
Another citation analysis was performed on VOSviewer to identify the most cited sources (journals, with 2 as the minimum number of documents, and 10 as the minimum number of citations per source. A total of 20 sources were displayed. Table 3 presents the list of the top ten cited journals. According to the citation analysis results, the most cited journals were the Ieee Access (303 citations) and Ieee Internet of Things Journal (160 citations) with the highest scores. Additionally, Ieee Transactions on Pattern Analysis and machine intelligence (78 citations); International Journal of Artificial Intelligence in Education (76 citations); Ieee Transactions on Neural Networks and Learning Systems (73 citations); Computers in Human Behavior (60 citations); International Journal of Manpower (47 citations); Human Factors (45 citations); Human Resource Development Review (16 citations); Advances in Developing Human Resources (13 citations); were the most cited journals.
The most cited institutions/organizations.
With a minimum of 2 documents and 50 citations set for the institutions/organizations’ citation analysis. A total of 30 organizations were displayed. Table 4 presents the top 10 institutions with the highest number of citations in AI and HRD studies. The findings reveal that Stanford University (USA) and the University of Cambridge (UK) held the first position with 584 citations followed by Columbia University (USA) (252 citations); Hong Kong Polytechnic University (Hong Kong) (158 citations); Jjiangnan University (China) (158 citations); Beijing University of posts and telecommunications (China) (133 citations); University of Electronic Science and Technology of China (China) (133 citations); Universidade Do Algarve (Portugal) (130 citations); Ecole de management Léonard De Vinci (France) (94 citations), and University of New Brunswick (Canada) (84 citations).
The most cited countries.
With a minimum of 5 documents and 10 citations set for the country citation analysis. A total of 14 countries were displayed. Table 5 presents the top 10 most productive countries in AI and HRD studies. The findings reveal that the USA was the most productive with 44 documents and 1,453 citations followed by the UK (21 documents,969 citations), China (60 documents, 783 citations); France (10 documents, 317 citations); Hong Kong (7 documents, 257 citations); Canada (8 documents, 241 citations); Italy (13 documents, 205 citations); India (13 documents, 198 citations); Australia (10 documents, 164 citations); and Spain (6 documents, 106 citations).
4. Discussion and conclusions
This study is a systematic literature review, using bibliometric analysis. It aims to examine studies related to the topic of artificial intelligence in HRD and to gain a deeper understanding of the knowledge structure of the integration of AI in HRD. The obtained results are discussed in the following sections highlighting their theoretical and practical implications.
Theoretical implications
Taking the results presented above, this study offers several theoretical implications. First, it broadens the scope of research on HRD and addresses the gap of the limited number of research on AI and HRD. Second, given that HRD and HRM are closely related (Harrison et al., 2021) and are complementary (Alagaraja, 2013), this study extends research on AI and human resources in organizations and adds evidence to studies examining AI in HRM. Considering the complexity of the organizational context, concentrating primarily on AI and HRM, with a limited volume of research dedicated to AI in HRD might not help to fully capture and understand how AI advances can influence HR in organizations. Thus, this study helps address this gap. Furthermore, considering that AI will profoundly alter jobs and skills, including those of HR professionals (Budhwar et al., 2022), it is imperative to gain a deeper understanding of the knowledge structure of this field and how it integrates AI. This study provides a compelling exploration and a comprehensive overview of AI in HRD based on the literature. It examines an extensive range of publications, identifying patterns in research and indicating future trajectories.
Third, this study enhances the existing literature by unveiling prevalent subjects within HRD and AI research using the most frequent keyword analysis. The obtained results shed light on research trends and guide researchers to keywords that are widely used, such as artificial intelligence, personnel training, deep learning and virtual reality, which represent research trends. The use of these keywords can be beneficial both for HRD research and practitioners who will be able to identify the relevant articles, needed in their field.
Fourth, the analyses of citation and co-citations can help researchers identify distinct research avenues that shape the intellectual structure of the examined topic, which might enable new and future researchers to develop a comprehensive grasp of the themes and foundational knowledge. The citation analysis results provide valuable insights into the most cited authors and documents. Researchers can draw upon the work of these highly cited authors and refer to the most frequently cited documents to inform and guide their future research. The obtained results of citation and co-citations analyses showed that most cited documents and authors are mainly from computer science and engineering fields. While it was expected that HRD authors would have a more substantial presence, the lesser prominence suggests several interesting avenues for explorations. It may indicate that AI in HRD is still a relatively emerging field where contributions from HRD researchers are gaining traction but have not yet reached the level of recognition seen in more established areas. Alternatively, it could signify that AI in HRD research benefits from interdisciplinary collaboration, with experts from diverse fields, including computer science and AI, contributing significantly. This finding underscores the interdisciplinary nature of AI in HRD and highlights the potential for further collaboration and knowledge exchange between HRD and AI researchers to advance this burgeoning field. This collaboration can enhance the integration of technology-related theories such as the human–agent and human–robot interaction theory in HRD research (Krämer et al., 2012). Also, the integration also of Reinforcement Learning Theory (Sutton and Barto, 2018) in the field of AI, with an application that covers both humans and machines/robots (Tseng et al., 2017). Future research could examine the reasons behind the observed pattern and identify strategies to encourage greater involvement of HRD experts in developing the future of AI in HRD research.
Moreover, the results of citation analyses provided insights into the journals, institutions and countries with the highest number of citations. The findings of this study underscore the dominance of computer science journals in the field of AI and HRD research, with journals like IEEE Access, IEEE Internet of Things Journal and IEEE Transactions on Pattern Analysis and Machine Intelligence emerging as top citation leaders. Notably, management journals hold a more limited presence within the top ten journals, with only three journals, namely, the International Journal of Manpower, Human Resource Development Review and Advances in Developing Human Resources. These findings suggest a distinct pattern in the citation landscape, possibly reflecting the interdisciplinary nature of AI in HRD research. This highlights an opportunity for greater collaboration and knowledge exchange between the fields of computer science and management to further advance the application of AI in HRD and bridge the gap between these domains. Future research should explore strategies to enhance the visibility of HRD-focused AI research in management journals and uncover emerging areas for interdisciplinary exploration in this evolving field.
In addition, this study shed light on institutions with the highest citations, mainly Stanford University, University of Cambridge, Columbia University and Hong Kong Polytechnic University. These institutions serve as beacons of research excellence, with their extensive citation records underscoring their global impact in AI and HRD research. Knowing about these universities can help future researchers develop a fertile ground for interdisciplinary collaboration and global knowledge exchange, through the creation of international research partnerships, facilitating collective efforts to address complex global challenges related to AI in HRD. In essence, this knowledge empowers researchers to tap into the expertise and networks offered by these institutions, catalyzing progress and innovation in the field.”
Furthermore, the obtained results revealed the top countries with high citations, underlining a global distribution led by the USA, UK, China and France. However, it's noteworthy that some emerging countries are notably absent from the list of high-citation countries. This absence highlights potential disparities in research infrastructure and collaboration networks that warrant further investigation.
Practical implications for organizations and policymakers
This study has practical implications for organizations (mainly managers and HRD practitioners), and policymakers.
From an organizational perspective, this study highlights the available research that can be used as a reference regarding the integration of AI in HRD in workplaces. Managers and HRD practitioners can use this bibliometric analysis to identify emerging trends in AI and HRD research, which might help them to stay ahead and anticipate changes or advancements related to workforce development. HRD practitioners can also leverage AI-based strategies based on the identified trends to enhance employee engagement and learning outcomes and drive innovation in HRD practices, which will enhance the organization's competitiveness. In this case, developing new organizational structures that leverage AI-centered design can be relevant (Yorks et al., 2020). Moreover, managers and HRD practitioners can use this study to support strategic HRD initiatives by aligning them with evidence-based research. In this case, the emphasis can shift towards crafting strategies and interventions aimed at achieving a harmonious equilibrium between human and machine training. Prioritizing the development of both humans and machines becomes crucial, as an exclusive focus on machines may pose a risk to the sustainability of employees' skills and long-term career prospects.
Finally, this study can provide insights to policymakers to create an environment that fosters research and innovation in AI and HRD. They can use these bibliometric insights to inform the development of policies related to AI adoption and its impact on the workforce. This may include regulations on upskilling initiatives and more focus on the training and development of people than machines. Furthermore, listing the most productive countries in this bibliometric analysis can help policymakers identify research gaps and needs in the country, allocate research funding more strategically to support areas of research related to AI and HRD, as well as identify international collaboration opportunities, which will facilitate partnerships to promote global HRD solutions for employees. Moreover, this study can help policymakers evaluate the impact of the existing policies by measuring the way they influenced research trends and outcomes.
Limitations and avenue for future research
This study has some limitations that need further investigation in future research. First, focusing on bibliometrics represents a significant contribution to this study, given the robustness and recognized reliability of this type of analysis. Despite the refinement performed to assess the relevance of selected papers based on their titles and abstracts, it is essential to acknowledge that the quantitative nature of bibliometric analyses may limit the capture of full qualitative aspects. These aspects include contextual nuances, theoretical frameworks and narrative elements present in the literature.
To obtain more qualitative insights from the analysis, future studies should go beyond the bibliometric analysis to identify any missing links that the bibliometric software was unable to capture. For this purpose, combining bibliometric analysis with other techniques, such as content analysis, can help identify contextual details and offer insights into themes, theories and narrative elements in the literature. Additionally, using case studies, interviews with experts and research surveys can provide a broader and more nuanced perspective on AI in HRD. Furthermore, future studies can use bibliometric analysis with altmetrics, a qualitative approach that measures the online attention and impact of research using metrics such as Altmetric Attention Score, Mendeley reads and social media mentions.
Second, while a bibliometric approach can provide an overview of knowledge mapping of studies on AI in HRD, this approach does not facilitate the empirical examination of relationships between variables. Future studies can leverage the outcomes of this study to empirically examine associations between variables and identify potential effects. This can be achieved through cross-sectional or longitudinal approaches, such as examining the effects of AI on employee training, the impact of learning algorithms on HRD effectiveness, or the effects of human interaction on machine and human learning effectiveness, using relevant theories. For example, social cognitive theory (Bandura, 2001) and Technology Acceptance Model (TAM) (Davis, 1985) can be used as theoretical frameworks to examine these effects by providing insights into the cognitive processes, observational learning and user acceptance that shape employees' responses to the use of AI in HRD in organizations. Additionally, future studies can investigate organizational readiness to implement AI for enhancing employee training and development, using for example the theory of organizational readiness (Weiner, 2009), which can offer a relevant theoretical framework that can guide research on organizational readiness to implement AI in HRD.
Third, in this study Scopus was used as the only database for data collection. Only articles published in Scopus were examined. Indeed, Scopus is a very large database, but it might not include all publications related to AI in HRD. So, some relevant articles, indexed in other databases may have been missed. Future research should consider the adoption of a holistic approach and include other databases such as Web of Science, Google Scholar and EBSCO. Finally, this study limits its analysis to articles as a source type, excluding books, book chapters, research notes and proceedings papers. These other source types should be included in future studies to attain a more comprehensive coverage of the field.
Conflict of interest: The authors state that there is no conflict of interest.
Figure 1.PRISMA flow diagram
Figure 2.The network of co-occurring keywords
Figure 3.Co-citation analysis for authors
Table 1.
List of co-occurring keywords
| # | Keyword | Occurrences | Total link strength |
|---|---|---|---|
| 1 | Personnel training | 166 | 560 |
| 2 | Artificial intelligence | 144 | 457 |
| 3 | Learning systems | 54 | 245 |
| 4 | Students | 37 | 176 |
| 5 | Machine learning | 27 | 86 |
| 6 | Human | 25 | 143 |
| 7 | Teaching | 25 | 134 |
| 8 | Human resource management | 21 | 66 |
| 9 | E-learning | 20 | 85 |
| 10 | Deep learning | 19 | 82 |
| 11 | Article | 18 | 98 |
| 12 | Decision making | 18 | 63 |
| 13 | Education | 18 | 81 |
| 14 | Engineering education | 18 | 104 |
| 15 | Learning algorithms | 17 | 79 |
| 16 | Active learning | 16 | 75 |
| 17 | Curricula | 16 | 81 |
| 18 | Humans | 16 | 100 |
| 19 | Support vector machines | 16 | 76 |
| 20 | Teachers' | 16 | 75 |
| 21 | Virtual reality | 16 | 54 |
| 22 | Task analysis | 15 | 76 |
| 23 | Big data | 14 | 49 |
| 24 | Education computing | 14 | 84 |
| 25 | Learning | 14 | 62 |
| 26 | Training | 14 | 63 |
| 27 | Decision support systems | 12 | 48 |
| 28 | Job analysis | 12 | 53 |
| 29 | Optimization | 11 | 47 |
Source: Developed by the authors
Table 2.
The list of top ten cited documents
| Rank | Article | Authors | Citations |
|---|---|---|---|
| 1 | Computer-based personality judgments are more accurate than those made by humans | Youyou et al. (2015) | 584 |
| 2 | Learning using privileged information: Similarity control and knowledge transfer | Vapnik and Izmailov (2015) | 252 |
| 3 | Seizure classification from EEG signals using transfer learning, semi-supervised learning and TSK fuzzy system | Jiang et al. (2017) | 158 |
| 4 | IRAF: a deep reinforcement learning approach for collaborative mobile edge computing IoT networks | Chen et al. (2019) | 133 |
| 5 | Digital learning: developing skills for the digital transformation of organizations | Sousa and Rocha (2019) | 130 |
| 6 | The digital transformation of external audit and its impact on corporate governance | Manita et al. (2020) | 94 |
| 7 | Ensemble transfer learning algorithm | Liu et al. (2017) | 84 |
| 8 | Understanding the dark side of artificial intelligence (AI) integrated business analytics: assessing the firm’s operational inefficiency and competitiveness | Rana et al. (2022) | 79 |
| 9 | Training the body: the potential of AIED to support personalized motor skills learning | Santos (2016) | 54 |
| 10 | A self-learning dynamic path planning method for evacuation in large public buildings based on neural networks | Peng et al. (2019) | 53 |
Source: Developed by the authors
Table 3.
The list of top ten cited journals
| Rank | Source | Documents | Citations |
|---|---|---|---|
| 1 | Ieee Access | 19 | 303 |
| 2 | Ieee Internet of Things Journal | 3 | 160 |
| 3 | Ieee Transactions on Pattern Analysis and Machine Intelligence | 5 | 78 |
| 4 | International Journal of Artificial Intelligence in Education | 3 | 76 |
| 5 | Ieee Transactions on Neural Networks and Learning Systems | 6 | 73 |
| 6 | Computers in Human Behavior | 4 | 60 |
| 7 | International Journal of Manpower | 2 | 47 |
| 8 | Human Factors | 2 | 45 |
| 9 | Human Resource Development Review | 2 | 16 |
| 10 | Advances in Developing Human Resources | 3 | 13 |
Source: Developed by the authors
Table 4.
The list of top ten cited organizations
| Rank | Organization | Documents | Citations |
|---|---|---|---|
| 1 | Stanford University (USA) | 1 | 584 |
| 2 | University of Cambridge, Cambridge (UK) | 1 | 584 |
| 3 | Columbia University (USA) | 1 | 252 |
| 4 | Hong Kong Polytechnic University (Hong Kong) | 1 | 158 |
| 5 | Jjiangnan university (China) | 1 | 158 |
| 6 | Beijing University of Posts and Telecommunications (China) | 1 | 133 |
| 7 | University of Electronic Science and Technology of China (China) | 1 | 133 |
| 8 | Universidade Do Algarve (Portugal) | 1 | 130 |
| 9 | Ecole de management Léonard De Vinci (France) | 1 | 94 |
| 10 | University of New Brunswick (Canada) | 1 | 84 |
Source: Developed by the authors
Table 5.
The list of top ten cited countries
| Rank | Country | Documents | Citations |
|---|---|---|---|
| 1 | USA | 44 | 1,453 |
| 2 | UK | 21 | 969 |
| 3 | China | 60 | 783 |
| 4 | France | 10 | 317 |
| 5 | Hong Kong | 7 | 257 |
| 6 | Canada | 8 | 241 |
| 7 | Italy | 13 | 205 |
| 8 | India | 13 | 198 |
| 9 | Australia | 10 | 164 |
| 10 | Spain | 6 | 106 |
Source: Developed by the authors
© Emerald Publishing Limited.