An adaptive virtual reality game for programming education using fuzzy cognitive maps and pedagogical models

Abstract

Virtual Reality has proven to be highly promising within the field of learning. Most VR learning methods do not effectively implement pedagogical models or adapt to the individual's learning style. This research aims to bridge this gap by integrating Fuzzy Cognitive Maps (FCMs), Flow Theory and Gamification within an educational Virtual Reality video game to introduce and teach learners to Java programming. This new integration offers real-time accommodation to the learners' performance through dynamically balancing challenges and competencies (Flow Theory) and personalized, data-driven feedback (FCMs) and motivational stimulation through interactive gamified mechanisms (Gamification). With the use of FCMs to enable real-time personalization, this approach offers the ideal balance between competence and challenge to ensure deeper understanding of the subject matter. A comprehensive analysis verified significant improvements to the task performance, knowledge outcomes, along with the reduction of errors, validating the effectiveness of this adaptive VR method. The future of a more efficient and adaptive learning VR is made possible through this research that offers new knowledge about the integration of cognitive engagement, motivational aspects, and adaptive AI-powered learning.

Introduction

In the past ten years, Virtual Reality technology has shown great potential. The field related to technological advancements made its commercial availability possible since several technological companies currently offer high-quality products for wide distribution.

However, VR had to overcome numerous obstacles preventing it from being established. One significant obstacle was the prohibitively high gear cost, which prevented people from using VR. The low computational power of the hardware, which resulted in low resolution, unrealistic visuals with low latency and low frame rates and a bad user experience that caused motion sickness, pain and disorientation, were additional obstacles (Stauffert et al., 2020). Also, the equipment needed to be less spacious, lighter and more comfortable to use at home (Wang et al., 2022).

Many Head Mounted Devices (HMD) manufacturers are now researching and developing cutting-edge devices and services that will shape the future of VR. To mention a few, there are companies like Meta with its Oculus Quest Series, HTC with its Vive series, Sony with PlayStation VR for PlayStation consoles and PCs, Valve with its Valve Index and Pico Interactive with its Pico series.

VR is already being used by industries such as manufacturing and construction (Rokooei et al., 2022), education (Marougkas et al., 2023a, b), design (Liao et al., 2023), travel and tourism (Godovykh et al., 2022), entertainment (Onderdijk et al., 2023). There are many diverse applications for VR in the field of education and can let learners access Virtual Learning Environments (VLEs) that they otherwise wouldn't be able to interact with or experience safely (Stefan et al., 2023).

VR can be also used in medical training (Wu and Ho, 2023), engineering (Memarsadeghi & Varshney, 2020), history (Remolar et al., 2021), archaeology (Korniejenko et al., 2024), astronomy (Kersting et al., 2023), military (Kot et al., 2018), art (Hutson & Olsen., 2021), language learning (Chen et al., 2022), physical education (Pérez-Muñoz et al., 2024) and more.

VR can support programming education by assisting students to understand complex coding concepts through interactive experiences. VR might help learners experience an enjoyable learning process that is tailored to diverse needs and preferences (Papakostas et al., 2022).

VR has revolutionized learning through the delivery of personalized learning experiences that are dynamically adapted to the learning requirements of the individual learners in real time. Unlike static content-dependent traditional models of e-learning, VR learning environments enable real-time adaptabilities based on AI-driven models, sensor-based inputs, and real-time performance tracking. Adaptive VR has, for instance, been designed to enhance interventions to support children with Autism Spectrum Disorder (ASD) through real-time personalized tasking that improves social and emotional competencies (Maddalon et al., 2024).

Medical training has witnessed the utilization of cloud-based adaptive learning approaches that make use of Virtual Reality (VR) and Augmented Reality (AR) to simulate clinical scenarios where nursing students can train to undertake procedures under safe conditions to enhance clinical competencies (Oyekunle et al., 2024). This means that VR offers a more interactive and adaptable learning process compared to the utilization of the common digital methods.

Even though VR in education has been increasingly employed, many gaps can be filled in the integration of pedagogical theories and the personalization of learning experiences (Marougkas et al., 2023a, b). Similar trends have been observed in the adoption of online education, where models such as the Technology Acceptance Model (TAM) highlight key factors affecting in a positive manner learners'engagement with the use of digital educational platforms (Abuhassna et al., 2023).

The integration of adaptive learning within VR has gained recognition as an effective component to facilitate successful learning interventions. Adaptive mechanisms like AI models deliver a structured approach of personalizing learning through the monitoring of student performance and the adjusting of learning pathways accordingly. Combined with the use of Flow Theory that offers an ideal balance between challenge and skill levels, the learning experience through VR becomes more efficient capturing and maintaining the interest of the learners.

For example, an EEG frequency-based adaptive VR system has been developed to balance external and internal attention states to optimize the performance of the working memory performance during complex tasks (Chiossi et al., 2025). VR-based adaptive learning methods have been proposed within the engineering sector to tailor instructional information to the learning style of the individual to enhance understanding and memorizing complex information (Lin et al., 2022). In the study carried out by Scavarelli et al. (2021) the necessity of advanced adaptive systems within VR environments is highlighted to address effectively the diverse needs of learners. The majority of the VR applications are not employing learning theories like Constructivism (Hein et al., 1991) or Flow Theory (Csikszentmihalyi, 1997a, b) in order to deliver better learning outcomes. In the study of Oyelere et al. (2020), it is evident that most of the current frameworks lack a holistic approach that effectively integrates technological advancement with educational strategies.

Fuzzy Cognitive Maps (FCMs) (Kosko, 1986) combined with Flow Theory and gamification (Lee & Hammer, 2011) work to bring in adaptive and enjoyable learning experience in an effective manner. It can dynamically balance the learner's skills and challenges to fit educational requirements in a satisfying way and keep with the students'unique competencies (Troussas et al., 2020). Flow Theory keeps the learners more engaged and focused on the process by letting them get fully involved, while Gamification (Krouska, et al., 2022) develops motivation for increased interest in learning while at the same time transforming the learning process into an enjoyable and rewarding experience (Strousopoulos et al., 2024).

Therefore, this study aims to contribute to the discussion of personalized education environments in VR settings. The presented system provides a way to deliver an individually tailored educational experience blended with Artificial Intelligence (AI) Mechanisms by considering challenging factors like differing levels of skills, low engagement and the need for instant feedback.

To guide the design and evaluation of the study, the following research questions (RQ) were formulated:

• RQ1: Does the use of FCMs in a VR educational game lead to significantly higher conceptual understanding of programming constructs compared to a linear, non-adaptive hint system?

• RQ2: How does an adaptive VR system utilizing FCMs influence learners’ motivation, engagement, and sense of presence compared to a non-adaptive system?

• RQ3: Is knowledge retention improved when students interact with an FCM-adaptive VR system as opposed to a static system?

• RQ4: How do learners perceive AI-driven personalized feedback in terms of usefulness and support during the learning process?

In Sect."Literature review", research relevant to VR-based education in learning environments is discussed, with emphasis on gamification and personalization approaches taken in VR learning environments. In Sect."Methodology", the methodology of the study is presented, focusing on the design and implementation of the VR game and the development of the FCM. Sect."Results"focuses on delivering the results of the study. This includes an analysis of the user's experience, performance, engagement, learning outcomes and the learning effectiveness of the developed VR application. Sect."Discussion"discusses the findings of the study and finally, Sect."Conclusion"concludes this study by summarizing the key contributions of the study and suggesting directions for future research.

Literature review

Virtual reality (VR) in education

VR is already transforming education as an effective tool for delivering complex practical curricula by simulating real-world situations within the virtual environment. Immersion and presence are key dimensions of VR as technology and as a medium. Bowman and McMahan's (2007) study present immersion as an objective metric in which the VR system can replicate auditory and visual concepts similar to those found in the real world. Presence is the user's subjective sense of"being there"in the virtual environment and depends on individual and contextual factors (Witmer & Singer, 1998).

Students can engage with virtual environments to explore complex concepts through intuitive interactions using VR controllers or hand-tracking technology (Marougkas et al., 2024; Buckingham, 2021).

On the other hand, VR can contribute to programming, which is a basic discipline within informatics. The students can use the virtual environment to develop and enhance their coding skills (Klochko et al., 2022). In this way, the students interact with the virtual environment, which influences their cognitive processes and, in that way, they learn something new or enhance the knowledge that has already been acquired (Adnan et al., 2020). Problem-solving concepts in learning can improve students'perceptions of educational outcomes (Monita & Ikhsan, 2020), increase their understanding of the subject matter (Parmar et al., 2016), improve their academic achievement (Mariscal et al., 2020), improve their memory retention (Sanchez et al., 2019), enhance learning effectiveness (Chang et al., 2018), improve their perception of the subject (Bennie et al., 2019), improve their learning experience and stimulate their creativity in the classroom through this hands-on, immersive approach (Rychkova et al., 2020).

Students can, hence, through the system, receive immediate feedback to learn effectively from any possible mistakes while developing new coding strategies or improving old ones.

Integrating VR into the courses for programming education shows that ways can be fostered to turn towards new and innovative methods away from the traditional ones. Blended learning approaches, which combine digital and traditional educational strategies, have been shown to enhance engagement and effectiveness in similar educational contexts (Abuhassna et al., 2022a). Srimadhaven et al. (2020) discuss VR mobile applications used in teaching a Python programming language. As such, this software uses a cognitively effective method of learning and problem-solving scenarios by employing a gamified approach, which is fun and engaging for a user. This is an effective approach because the method personalizes learning, whereby learners study at their own pace and at a pace that considers their cognitive needs and diversified educational backgrounds.

Segura et al. (2020) also introduced, in their study, how VR was used in programming to develop the VR-OCKS that was used in teaching fundamental concepts of programming. The VR application, VR-OCKS, uses gamification, which introduces enjoyable dynamics in curriculum development to provide users with a complete and entertaining experience. Children and adults can solve puzzles as they experience VR by interacting with virtual objects and concepts. They would therefore be able to learn about loops, conditionals and sequences. The findings of this study were able to illustrate how logical thinking, spatial awareness and problem-solving skills are enhanced among the users. This research also indicates that the users enhance their motivation and engagement with this experience, understanding programming concepts, hence showing the potentiality of VR for complex and modern educational scenarios (Segura et al., 2020).

The role that VR plays helps students to interact with the world, hence practicing and providing the ability to enhance such a skill quite easily, all while comprehending or interpreting the curricula. The learners, therefore, get to give their game as certain situations or scenarios throw them into real action wherein their coding skills come into play due to knowledge and filling gaps.

Virtual Reality (VR) applications have been employed in prior research in various fields such as medical training, engineering, history, archaeology, and programming education. Despite the broad range of topics and applications, most do not employ adaptive methods to deliver a tailored learning experience.

Research has shown that while VR has been shown to enhance engagement and immersion, pedagogic models have not been satisfactorily integrated with VR, and dynamic adaptation to learners is usually absent. This study fills this gap by employing Fuzzy Cognitive Maps and Gamification in conjunction with Flow Theory to design and deliver an adaptive VR-based learning system that provides personalized learning content in real-time. By making dynamic adjustments to challenge levels and delivering personalization in response to learners'progress, this work enhances engagement and makes VR-based learning more effective and tailored to learners.

Flow theory and gamification in learning

The psychological state of total immersion and focused motivation in which a person fully engages in action is known as Flow, as defined by Mihaly Csikszentmihalyi (1997a, b). Flow is also appropriate for VR environments since it allows the user to engage in an educational setting where they can be motivated and fully engaged, resulting in an effective learning process. Because of this, challenges in VR educational games have to maintain a balance between focusing on the player's skills while maintaining interest throughout the whole process to prevent learners from losing interest or getting bored when enduring an experience that is either too easy or too difficult, which eventually leads to a less engaging experience resulting in frustration.

In their study, Akman and Çakır (2019) presented an educational VR game to teach fractions in the fourth-grade mathematics curriculum. The study focused on the effectiveness of the developed VR game using Flow Theory, focusing on the balance between challenge and skills, concentration and educational goals. The study's results indicated that the game delivered a positive flow experience. The integration of Flow theory made the learning process enjoyable and effective for the students.

Gamification, when combined with Flow Theory, provides more layers of incentive and reward to capture learners'interest throughout the learning experience. It is through the integration of Fow Theory and gamification that VR educational games are able to provide very captivating contexts where learners are driven to learn and apply new concepts in a context that seems so rewarding and enjoyable.

Liu and Tarng (2021), in their study related to ecological awareness, developed and presented a game-based VR educational application for students. Following the Gamification of learning the educational game, they incorporated feedback mechanisms, missions, and Role-Playing Game (RPG) elements. This immersive learning experience using a Head-Mounted Device (HMD) was compared to a desktop VR version of the curricula, indicating that the HMD version as an immersive experience led to improved learning effectiveness and a better sense of presence, decreasing the levels of students’ anxiety.

Gamification adds a whole new dimension to the learning process by transforming non-gaming contexts into gamified experiences, assisting Flow Theory with the elements of motivation techniques such as rewarding systems (Abuhammad et al., 2021), gamified scenarios (Surer et al., 2020), scoring systems (Johnson-Glenberg et al., 2020), gaming levels (Johnson-Glenberg et al., 2020), multiplayer and collaborative modes (Rychkova et al., 2020), different levels of difficulty (Chen et al., 2005), items inventory (Akman & Çakır, 2019), timers (Abuhammad et al., 2021; Liu et al., 2021), health point system (Sureret al., 2020), badges (Chen, 2020), leaderboards (Chen, 2020), Non Playable Characters(NPCs’) (Surer et al., 2020), real-time feedback (Bendeck et al., 2020) are just a few ways to capture learner’s interest and engage them into the immersive learning experience.

Fuzzy cognitive maps (FCMs) in educational technology

Fuzzy Cognitive Maps (FCMs), introduced by Bart Kosko is a technique for depicting complex systems. The components of an FCM are displayed in a graph as concepts in the form of nodes that consist of the entire system and their interactions. Nodes are connected depending on their relationship, with different weights and influences on each other, allowing simulations to examine different real-life situation scenarios (Felix et al., 2019). FCMs are also used in education to present concepts related to the learning process. Recent studies have highlighted the role of FCMs in analyzing learning readiness and structuring personalized interventions based on cognitive modeling (Abuhassna et al., 2022b). These can involve using a dynamic AI technique and in real-time adjusting elements like motivation, feedback, personalization, engagement, cognitive load, collaboration, retention of knowledge, difficulty and learning outcomes.

Cai et al. (2010), introduced a version of FCMs that aims to enhance users’ immersive experience in serious games. Their version of FCMs, called Evolutionary Fuzzy Cognitive Maps (E-FCMs), is an extension of FCMs that incorporate both fuzzy logic and probabilistic causal relationships. The real-time dynamic modeling of E-FCMs provided more complex interactions between game characters and virtual environments resulting in a personalized and enhanced engagement of the users within the virtual environment. This approach delivered a more effective learning experience within virtual environments incorporating gamified techniques.

FCMs have proven to be a useful approach to developing learning environments (Hossain & Brooks, 2008). They can dynamically change their elements to help students have a more personalized, effective and enjoyable learning experience. They can also transform a learning process into a more challenging experience for the learner. FCMs can take into consideration a variety of factors, such as prior knowledge, learning preferences, performance, learning pace and more. In this way, the learner can be held motivated and excited, resulting in a more effective, enjoyable and more challenging learning experience.

Addressing personalization and pedagogical gaps in VR learning with FCMs, flow theory and gamification

Recent research highlights the need for intelligent adaptive systems to integrate AI-based models with VR to optimize learning personalization. AI-adaptive learning systems such as FCM have been shown to have promise in adapting learning content in real-time to student performance, delivering a personalized and effective learning experience (Kosko, 1986). With the integration with established learning theories such as Gamification and Flow Theory, VR can personalize content and optimize engagement and motivation to deliver a learning experience that puts students in a productive and engaged state of learning (Csikszentmihalyi, 1997a).

With the incorporation of FCMs, Flow Theory and Gamification in VR-based learning, this research aims to provide a comprehensive adaptive learning system that modifies challenges, material difficulty and feedback in real-time in line with learners'ability and requirement. Shifts from theoretical models to practical applications in e-learning systems have been extensively explored, with an emphasis on how technological interventions influence student performance and learning outcomes (Abuhassna, 2024). The following sections explore how this integration enables enhanced adaptive learning, in programming education.

Addressing gaps in literature: the need for a holistic adaptive learning model in VR

Despite the fact that VR’s use in education is rapidly growing, existing VR-based adaptive learning frameworks lack adequate integration of pedagogic theories with AI-based personalization. Although game-based learning and adaptive AI-based systems have been explored in research (Zawacki-Richter et al., 2019), a comprehensive framework that integrates Flow Theory, Gamification and FCMs in VR remains underexplored.

Lack of pedagogical framework integration in VR-based learning

The majority of VR educational applications are centered on engagement but lack a solid pedagogic background. It has been determined that Flow Theory, which ensures learners are in a state of optimal engagement through a balance between challenge and skill, has not been fully utilized in adaptive VR learning (Csikszentmihalyi, 1997a). Similarly, Gamification elements such as rewards, feedback loops and progression systems are applied inconsistently and thus fail to maintain motivation.

FCMs as a solution for personalized and adaptive VR learning

Conventional systems for adaptive learning apply pre-existing rules and static adaptations to content. However, FCMs present a dynamic and flexible approach to adapting VR learning via analyzing learners'real-time actual performance and adapting instructional strategies accordingly (Papageorgiou and Groumpos, 2005). Though FCMs have been proved to be effective in decision-making and AI, their applicability in real-time adaptive VR-based learning is yet to be fully explored.

Bridging the gap: a novel framework for adaptive VR learning

This work addresses such gaps by employing an adaptive learning framework that integrates FCMs, Flow Theory and Gamification to deliver a personalized VR learning experience. Whereas previous research has focused on either engagement (Gamification) or adaptability (AI-based models) in isolation, this work demonstrates how all three can be blended to deliver a more effective, engaging and adaptive learning experience.

Personalized learning experiences and the integration of key pedagogical frameworks like Flow Theory and Gamification are frequently absent from current VR systems. To meet the different requirements of learners, research emphasizes the necessity of AI-driven solutions and adaptive mechanisms (Oyelere et al., 2020; Scavarelli et al., 2021). By using Fuzzy Cognitive Maps (FCMs) to develop a VR educational game that adapts to each learner's unique characteristics, this study fills these gaps and provides an engaging and effective educational experience.

FCMs, Flow Theory and Gamification in VR-based education environments can be combined to develop adaptive, engaging and effective learning environments. FCMs as an approach can assist the game in real-time by balancing the learner’s skills and challenge absorbing the learner in an immersive experience. According to Flow Theory, the learner is completely immersed and focused on this state, and the tasks are neither too easy nor too difficult, which maintains the engagement with optimal learning. Gamification also is an approach that transforms a learning environment into an engaging and exciting experience enhancing learners’ motivation and attracting students’ interest in the learning process.

Combining these three distinct methods enhances learning outcomes and aids students in understanding the subject matter while attracting their curiosity and preventing boredom, which keeps students engaged and motivated throughout the learning process. This will likely contribute to an improved learning approach for students to retain the knowledge they have learned so they can apply it in scenarios and contexts that they are going to experience in the real world (Table 1).

Table 1. Roles of learning approaches in adaptive learning environments

Component	Role in adaptive learning environment
FCMs	Dynamic adaptation of content and difficulty
Flow theory	Maintaining user engagement through challenge-skill balance
Gamification	Motivating and rewarding user participation

Figure 1 shows how Gamification and FCMs are combined along with Flow Theory within the educational VR environment. FCMs adjust in a dynamic manner, challenge and content continuously according to learners'real-time performance to deliver direct personalized feedback and greater engagement. Flow Theory provides balance between challenge and skills to prevent boredom and frustration. Gamified features such as reward systems, interactive tasks and progressive adaptation enhance learners'engagement and motivation. Synergy between them provides a fully adaptive and engaging experience in VR, tailored to all students’ skills and competencies.

[See PDF for image]

Fig. 1

Interactions between components in the VR environment

In particular, the role of these components is the following:

• FCMs enable adaptive learning content based on real-time inputs. The learning materials must be in line with the player's performance in the game to achieve that.

• Flow Theory is in charge of establishing a balance between user skills and challenges to keep student engagement high and prevent boredom, annoyance, or even quitting the VR educational experience.

• Gamification Theory provides interactive mechanisms, including adaptive feedback, immersive challenges, non-player characters (NPCs), reward systems, storyline integration, visual and audio enhancements, tools and interactions, error-based learning, gamified feedback system, gaming levels and progression and difficulty levels that are intended to increase player motivation.

As such, the developed VR educational game provides the following functionalities incorporating innovative approaches:

• Dynamic Feedback This module handles user interactions to deliver real-time updates to the FCMs, supporting continuous adaptation.

• Learning Content The module is customized through FCMs and forms the main element of the VR-based educational experience.

• User Interaction This module enables key mechanisms, such as feedback, engagement and gamification rewards, delivering an interactive and immersive experience.

Methodology

VR game design and structure

This VR game provides an interactive and immersive way for university students to learn Java programming. For this game, FCMs, Flow Theory and Gamification were chosen to design, implement, and deliver this educational experience so the users will engage in this adaptive learning process.

This VR game was implemented in Unity 3D. Unity 3D is among the most adopted and spread development platforms, which has broadened possibilities in developing applications such as games (Strousopoulos et al., 2023), simulations (Esposito et al., 2020), Virtual Reality (Innocenti et al., 2019), Augmented Reality (Papakostas et al., 2021) and Extended Reality (Isik et al., 2024). It provides a complete rendering, physics, animation, photorealism and scripting toolset that allows developers to design and develop their content on a lot of kinds of platforms: PCs, game consoles, cell phones, VR, AR, XR devices and many more. It also has a wide community that supports the system by offering 3D models, plugins and ready-to-use scripts, making the platform easy to use and developing their unique ideas and concepts from students to huge organizations.

Some of the virtual objects and animations that were employed within the game were downloaded either from sketchfab.com or Unity Asset Store. Sketchfab.com and Unity Asset Store (provided by Unity Technologies) are popular websites for developers, artists and designers. They allow users to publish and interact with 3D models, scripts, animations and related tools via their web browser. They also function as marketplaces where users can buy or sell their work.

Meta Quest 2 is the device that was chosen as the development platform. Meta Quest 2 is an HMD produced by Meta (formerly known as Oculus Quest 2) and it comes with 2 controllers that allow users to interact within the virtual environment, with virtual objects and mechanisms in an intuitive manner.

Meta Quest 2 was selected for this study as the primary HMD because it is affordable, easily accessible and due to its functions, which makes it a great choice for educational applications. Meta Quest 2 comes with lightweight, ergonomic and user-friendly controllers that allow intuitive interactions between the user and the VR environment.

This HMD is compatible with Unity3D and supported by the Oculus Integration SDK, which provides access to essential components, prefabs, scripts and APIs to streamline VR development. Additionally, Meta Quest 2 has a dual role in that it can be utilized as a stand-alone device or paired with a PC to boost graphic processing. Standalone mode eliminates dependency on high-priced gaming PCs, reducing cost and making it more affordable for educational institutions and providing wireless freedom and unrestricted mobility. All these factors made Meta Quest 2 the best device for this research to deliver an immersive and flexible VR learning experience.

The VR educational game of this study aims to provide an interactive and immersive learning experience for students learning the Object-Oriented Programming language Java. This game is designed based on Flow Theory and Gamification within the context of FCMs for delivering adaptive learning related to the needs and competencies of the students. The game consists of three levels of gaming, each with a different approach and progressive building of students'knowledge and skills in the subject loops of Java programming.

Game mechanics and learning activities

Game structure overview

Landing screen

Upon launching the application, the first screen “transfers” the user into a landing screen that features information through the user interface and includes a “Start” button. Pressing the button transitions the user to the next scene of the game.

Game Scenario

The second scene of the game “transfers” the user inside a library hall, which provides information in the form of text on a transparent board. The information presents the game scenario, which is taking place in the virtual world of “Codearia” where players seek the “Java Codex”, an ancient book hidden in the forgotten floating island of the “Forgotten Algorithms” and filled with unmatched programming knowledge to become the “Master Programmer” of the realm. The user must choose and click the"Continue"button in the middle of the screen in order to move on to the next scene of"Course material".

Course material

This scene contains the material that the user has to read in order to be prepared before proceeding to the gaming levels that will face challenges related to the subject matter. For this reason, the loops section was chosen from Java Object-Oriented Programming. The scene appears in the same virtual space as the"Game Scenario"scene and a transparent board holds the text that provides information about the loops (Fig. 2).

[See PDF for image]

Fig. 2

Course material

Level 1: Basic Syntax and Constructs

At the beginning of the first level, the player is placed in the square of a virtual village. The player can roam around with the help of controllers provided by Meta Quest 2. The user has to win each challenge in all three difficult levels of the game in order to unlock the next level. In each level, the way is being blocked at every stage of the game by giant rocks. The first gaming level requires the player to interact with an NPC (Non-Player Character) by answering questions once the NPC has been located within the virtual space. This level makes use of the Fuzzy Cognitive Maps (FCMs) method and dynamically assists the player by providing tailored hints that are adjusted to the player’s understanding and current performance. FCMs adjust and provide hints depending on the level of the player's knowledge assessed dynamically. All hints provided will be personalized and efficient to enhance learning outcomes and bring greater understanding to players'achievements. The player here, besides getting support through challenges, maintains a sense of accomplishment and interest and thereby will get a feeling of involvement with confidence in the skills to progress.

This approach is one of the most efficient adaptive mechanics in a game to help the player not be caught in any sense of discomfort or boredom. The more the player moves around, interacting with the environment and the NPC, the system adjusts the subsequent hints based on answers and errors. In this way, the learning environment will become responsive and effectively engaging with the player, whereas the player will improve his or her comprehension of Java programming principles, which will make the process of learning more personalized and effective.

Figure 3 shows an example of the Level 1 quiz with the FCM-driven adaptive hint mechanism activated. The system detects user struggles and provides contextual hints based on real-time interactions, helping learners progress without excessive frustration.

[See PDF for image]

Fig. 3

Level 1 quiz with hint assistance

Level 2: Logical Sequencing and Code Construction

In the second level of the game, the player enters a whole new level in which is required to apply the newly acquired knowledge of the subject by activating four torches that represent programming statements. At the start of the stage, the torches are not lit, so each time the player activates one, it lights up independently. Activating four torches, which represent programming statements, is how the player applies the knowledge they have gained to a new level in the game. Each torch lights up on its own whenever the player activates it because they aren't lit at the beginning of the stage. The statement is illustrated on a sign located on the front of each torch. To choose the statement of their choice and to approach each torch, players must use Meta Quest 2 controllers, which are required to cast a beam inside the game. In order to form a valid line of code, the player has to activate these four torches in the correct sequence. This level also focuses on loops and conditional statements. The mechanics behind this level also make use of Fuzzy Cognitive Maps, as in the first gaming level, and assist the player on activating the torches in the right sequence. The feedback mechanism not only resets and adjusts the torches appropriately but also provides the player with a hint on how to proceed with the challenge if any of the four are activated in the incorrect order. The learning environment becomes adaptive in this way, allowing the player to develop critical thinking and develop from errors that they make.

Level 3: Debugging and Code Optimization

Placing four rupees in four stones is one of the challenges in the third and final gaming level. To complete and successfully solve the challenge, each rupee has a different programming statement attached to it in the form of a text on top. Each rupee has to be matched with its corresponding stone. A sign is placed above each stone, with a line of code that needs to match the statement of each rupee. Similarly, as in the previous levels of the game, players will need to walk around in the level using their Meta Quest 2 controllers while matching and transferring rupees from their initial position to the stone by pointing at the beams cast throughout the virtual environment. The challenge of the level will be based on the logic of debugging and code optimization, teaching the player to find errors within the code and their correction for efficient coding. Additionally, at this level, the hint system offers support based on the user's errors in the form of failed attempts.

Ending scene

Upon completion of Level 3, the player at the end of the game is rewarded when he finally acquires the"Java Codex": an ancient book hidden in the lost floating island of “Forgotten Algorithms”. The book contains legendary knowledge in programming on how to become the"Master Programmer"of the realm.

The ambient music is carefully integrated into the landing screen, game scenario and course material screens to assist players in focusing on the projected content. Other background sounds for the game stages include mystical music, the chirping of birds and the ambiance of the forest. The audio design contributes to the level of immersion that the VR experience offers, enabling the user to feel like a part of this virtual world. Furthermore, on these screens, the player will not be as free to move around in the virtual environment as during the three following game stages. In this respect, in order not to cause motion sickness and vertigo while improving reading comprehension, the user is kept static, being able to move only the head. This means that the role of the controllers will be limited to selecting and clicking buttons.

Adaptive learning with fuzzy cognitive maps

For modeling interaction among concepts like Player Knowledge, Engagement Level, Game Difficulty, and Learning Outcome in a learning process, FCMs was used as one method. FCM is an important component that assists this game in delivering adaptability and adjusting to the players'requirements and competencies, aiming to facilitate effective learning.

Concept identification

Before FCMs are used, the concepts need to be listed and assessed first. These concepts will be integrated into the in-game mechanisms. The degree of influence on their learning outcomes and the level at which the system delivers the engagement mechanisms needed to capture the player's interests is a crucial concern that a system must consider.

The following concepts were selected:

• C1: Player Knowledge—This concept illustrates the player’s comprehension of Java programming-based concepts, especially loops. It is fundamental to the game's FCM as it directly affects other concepts, such as Engagement Level (C3), Learning Outcomes (C5), and Game Difficulty (C4) while being influenced by Hint Effectiveness (C2) and Feedback Timing (C8).

• C2: Hint Effectiveness – Measures the effectiveness of the hint system in supporting players during the learning process. Effective hints are expected to improve Player Knowledge (C1), increase Engagement Level (C3), and reduce Error Rate (C7).

• C3: Engagement Level – Illustrates the player's level of commitment and interest in the virtual experience. The level of engagement to sustain learning has to remain high, and it shall be determined by the balance between the levels of Player Knowledge (C1), Game Difficulty (C4), and other dynamic factors.

• C4: Game Difficulty – Represents the degree of challenge in the game. It adapts based on Player Knowledge (C1) and Error Rate (C7) to sustain an optimal challenge level, thereby influencing Engagement Level (C3) and Learning Outcomes (C5).

• C5: Learning Outcomes – Refers to the learning achievements made through the game, such as understanding and applying Java programming concepts. Learning Outcomes are influenced by Player Knowledge (C1), Engagement Level (C3), and Hint Effectiveness (C2) and directly impact Player Satisfaction (C6).

• C6: Player Satisfaction – This concept reflects the overall enjoyment and satisfaction of the player while playing the game. It is influenced by Learning Outcomes (C5), Engagement Level (C3), and Game Difficulty (C4), ensuring the player is being kept motivated throughout the whole experience.

• C7: Error Rate – Delivers the frequency of player's mistakes in performing certain tasks. A high error rate could indicate that a game is too hard, while a low error rate may indicate increased proficiency. Error Rate is influenced by Hint Effectiveness (C2) and Player Knowledge (C1) and affects Game Difficulty (C4).

• C8: Feedback Timing – Refers to the speed and relevance at which the players receive their feedback. Feedback Timing influences Hint Effectiveness (C2), Player Knowledge (C1), and Error Rate (C7), and ensures that players receive timely support to improve their performance.

• C9: Task Completion Time – Represents how much time players spend completing tasks. It reflects on the understanding and efficiency of players, which may be affected by Error Rate (C7), Game Difficulty (C4), and Feedback Timing (C8).

• C10: Adaptability of Game – This reflects that the game dynamically changes configurations based on the performance of each player to ensure a personalized challenging and engaging educational experience. Game Adaptability interacts in all other concepts to balance difficulty, engagement and learning outcomes.

Table 2 illustrates the key concepts in the FCM-based adaptive system and how different factors, such as player knowledge, level of involvement, error rate and difficulty level, interact to improve the learning experience.

Table 2. Conceptual framework of adaptive learning factors

Concept number	Concept name	Influence from other concepts	Influences on other concepts	Additional notes
1	Player knowledge	Hint effectiveness, Game difficulty, Feedback timing	Engagement level, Game difficulty, Learning outcomes	Core to adapting game difficulty and content delivery
2	Hint effectiveness	Player knowledge, Feedback timing	Player knowledge, Engagement level	Critical for providing timely and contextually relevant guidance
3	Engagement Level	Player knowledge, Hint effectiveness, Game difficulty	Learning outcomes, Player satisfaction	Affects how absorbed players are in the game
4	Game difficulty	Player knowledge, Error rate, Task completion time	Player knowledge, Engagement level, Error rate	Adjusts to suit player’s ability, keeping them in the flow state
5	Learning outcomes	Player knowledge, Engagement level	Player satisfaction	Measures educational effectiveness
6	Player satisfaction	Engagement level, Learning outcomes	–	Influenced by the overall game experience
7	Error rate	Player knowledge, Game difficulty, Feedback timing	Game difficulty, Task completion time	Indicator of player challenges and misunderstanding
8	Feedback timing	Error Rate, Task Completion Time	Hint effectiveness, Error rate	Timeliness and relevance of feedback impacts learning
9	Task completion time	Game difficulty, Error rate, Feedback timing	Feedback timing, Error rate	Reflects efficiency and understanding of game tasks
10	Adaptability of game	All other concepts	All other concepts	Allows the game to adjust dynamically to player performance

Dynamic relationships between the key components in FCM-driven learning VR environment are illustrated in Fig. 4. Through a system of interconnected weighted relationships, each component (Player Knowledge, Engagement Level, Game Difficulty, Learning Outcomes, and Adaptive Feedback) impacts the others. FCMs provide personalized learning by continuously modifying challenge levels and feedback systems based on user activity. This figure illustrates how real-time information drives the adaptive VR system to enhance student engagement and comprehension of programming concepts.

[See PDF for image]

Fig. 4

Dynamic interactions in the FCM-based learning system

A graphical representation of FCM-based learning system is illustrated in Fig. 4. The nodes show the key factors of FCMs and how they affect learning (e.g., performance, engagement, and errors) and weighted edges indicate each factor's influence in real-time adaptive decision making.

FCM weight assignments and calibration

In this section, the process of assigning and calibrating weights within the Fuzzy Cognitive Map (FCM) model is presented, highlighting how interactions among key learning concepts in the adaptive VR educational environment are effectively represented and adjusted.

Step 1: mapping relationships and assigning weights

Furthermore, high Game Difficulty (C9) rates may result in higher Error Rates (C6), but it may also improve Player Knowledge (C1) as the player gains experience in effectively overcoming in-game challenges.

A theoretical approach was taken to determine how the weights of relationships will be distributed across nodes interconnecting concepts. Testing and feedback provided by players after early interventions further strengthened these value validities. Seven experts also participated in setting up such weights. Some of them are more focused on FCMs in educational contexts, while all the experts have substantial experience in intelligent and adaptive systems in educational technologies. They are all affiliated with the Department of Informatics and Computer Engineering at the University of West Attica (UniWA). Their contribution provides robustness and reliability to weight distributions. FCMs in the VR system were calibrated through iterative validation with a mix of expert judgment and empirical testing. Weights were initially established using theoretical principles and expert judgment to ensure principal learning parameters like engagement, knowledge retention, and difficulty were modeled correctly. To refine these initial weights, a user study was conducted in early stages with a pilot set of students, who interacted with the VR environment while tracking levels of engagement, task completion time, and use of hints. Validation of FCM weight assignments was performed through empirical method involving expert initialization and iterative fine-tuning in real-time interactions with students. Domain experts estimated the initial weights from prior related studies in adaptive learning models. The weights were fine-tuned with pilot testing, in which students'performance metrics—such as variations in engagement, response accuracy and time to finish—were analyzed. The system dynamically adjusted these weights using a reinforcement mechanism, ensuring the learning pathways remained aligned with the principles of Flow Theory and personalized learning needs (Papageorgiou & Groumpos, 2005). These interactions provided real-time feedback, allowing for incremental adjustments to FCM weights to ensure the system dynamically adapted to learner performance. The final calibration ensured that the FCM-based adaptation mechanism aligned with empirical learning behaviors, improving both engagement and knowledge retention.

Calibration of FCM weights

Supportive relationships are displayed by positive weights (e.g., effective hints enhance player knowledge) and opposing relationships are represented by negative weights (e.g., high error rates and lower player satisfaction).

Concepts are denoted as:$$\text{C}=\left[\text{C1},\text{C2},\text{C3},\dots ..\text{C1}0\right]$$where:

• $C1$ represents Player Knowledge

• $C2$ represents Hint Effectiveness

• $C3$ represents Engagement Level

• $C4$ represents Game Difficulty

• $C5$ represents Learning Outcomes

• $C6$ represents Player Satisfaction

• $C7$ represents Error Rate

• $C8$ represents Feedback Timing

• $C9$ represents Task Completion Time

• $C10$ represents Adaptability of the Game

Dynamic adaptation example using fuzzy cognitive maps (FCMs)

FCMs modify the learning environment dynamically while students interact with the VR environment, turning the educational experience into a personalized one.

For example, a student is constantly making errors in an activity that requires identifying the correct syntax of Java Loops. In this case, Error Rate (C7) enters in a high activation mode after the same errors occur several times. This scenario will trigger the interaction by the system:

A high Error Rate (C7) will decrease the activation level of the Player Knowledge node (C1), thus indicating a knowledge gap. The system increases the impact of the Hint Effectiveness (C2) node, the Feedback Timing (C8) node based on the weights within the FCM. Thus, the game supports the player with hints related and linked to the type of error the player has committed.

The Game Difficulty (C4) node adjusts to make the current task less complicated if the error keeps occurring. In this way, the students enhance their engagement level while improving their skills.

Impact on learning variables

These adjustments affect the learning experience of the student significantly:

As the student effectively completes the tasks, the Error Rate (C7) concept decreases. Student motivation is maintained as an outcome of increased levels of engagement (Engagement Level (C3)) caused by enhancements in Player Knowledge (C1). The Learning Outcomes (C5) are enhanced by adaptive hints and real-time feedback, which assist learners to comprehend the content more effectively.

Visualization of adaptive interactions

The following diagram illustrates the dynamic interactions within the FCM:

Hint Effectiveness (C2) has a positive effect on Player Knowledge (C1), while Error Rate (C7) has a negative effect. Loops for real-time feedback that adapt the game difficulty and hint effectiveness according to user performance. A stimulating learning process and continuous adaptation are provided by the interlinked system.

Figure 5 shows the learning flow within the VR-based adaptive system. It illustrates how the system dynamically adjusts difficulty levels, provides real-time feedback, and modifies learning paths based on the learner’s performance. The integration of Fuzzy Cognitive Maps (FCMs) ensures personalized adaptation, enhancing engagement and knowledge retention.

[See PDF for image]

Fig. 5

Dynamic interactions in the FCM-based learning system

Step 2: initial concept values and weights

Through weight-connected relationships, each concept influences every other concept. Each concept is represented in the following weight matrix by the degree and direction to which it influences another concept. A positive weight means a positive influence and a negative weight means a negative influence. Each element $\mathit{wij}$ e. i. the impact of concept Ci on concept Cj and thus, the weights are organized in the matrix.

The initial activation values for each concept $\mathit{Ci}$ at time $t$=0:$$C\left(0\right)=\left[0.5,,,0.3,,,0.4,,,0.6,,,0.2,,,0.7,,,0.3,,,0.4,,,0.5,,,0.6\right]$$

The following matrix represents the corresponding Weight Matrix. $W$ illustrating the influence of one concept on another:

Table 3 provides the weight matrix used in the FCM model, illustrating how different learning factors influence one another in the adaptive VR educational system. Positive weights indicate supportive relationships, while negative weights represent opposing influences.

Table 3. Weight Matrix of Conceptual Influences

Concept	C1	C2	C3	C4	C5	C6	C7	C8	C9	C10
C1	0,00	0,30	0,40	− 0,20	0,50	0,30	− 0,30	0,20	− 0,10	0,40
C2	0,20	0,00	0,40	0,00	0,20	0,10	− 0,10	0,50	0,00	0,30
C3	0,30	0,50	0,00	0,40	0,50	0,40	− 0,20	0,10	0,30	0,40
C4	− 0,30	0,20	0,40	0,00	0,10	0,20	− 0,40	0,10	0,20	0,50
C5	0,50	0,20	0,40	0,30	0,00	0,40	− 0,20	0,10	0,20	0,40
C6	0,30	0,20	0,50	0,30	0,40	0,00	− 0,30	0,10	0,20	0,40
C7	− 0,40	− 0,10	− 0,20	0,30	− 0,30	− 0,30	0,00	− 0,20	− 0,30	0,10
C8	0,30	0,50	0,40	0,20	0,20	0,30	− 0,20	0,00	0,10	0,40
C9	− 0,20	0,10	0,20	0,30	0,10	0,20	− 0,30	0,20	0,00	0,30
C10	0,50	0,30	0,40	0,40	0,50	0,40	0,10	0,30	0,20	0,00

$\mathit{Wij}$ is the influence of the concept $\mathit{Ci}$ on concept $\mathit{Cj}$.

Step 3: mathematical formulation of FCM

The updated values for each concept are then calculated using the influences illustrated in the weight matrix.

The following formula provides the update value for each concept $\mathit{Ci}$ at time t + 1:$$C_i^{\left\{\left(t,+,1\right)\right\}}=f({\sum }_j{w}_{\left\{i,j\right\}{C}_j^{\left\{\left(t\right)\right\}}})$$

• $C_i^{(t+1)}$ provides the updated value of the concept $\mathit{Ci}$,

• $\mathit{wij}$​ is the weight which illustrates the influence of the concept $\mathit{Cj}$​ on concept $\mathit{Ci}$,

• $C_J^{(t)}$ provides the value of the concept $\mathit{Cj}$ at the current iteration $t$,

• f(x) is a nonlinear transfer function that guarantees the concept values fall within a desired range between 0 and 1. In this case, the sigmoid function is being applied:$$f\left(x\right)=\frac{1}{1+e^{-x}}$$

Step 4: first iteration

As the first iteration takes place the new values of each concept separately have to be calculated. For this reason, the above equation has to be applied. Based on the initial values of the concepts and the weight matrix which projects the weight distribution, the values will be updated.

This results in:$$C^{\left\{\left(1\right)\right\}}=f\left(W,·,C^{(0)}\right)$$

After the replacement of the values from $W$ and $C^{(0)}$, the new values for $C$ as produced after the first iteration are:

$C^{(1)}$ = [0.6502, 0.6726, 0.7841, 0.6341, 0.7773, 0.7231, 0.3612, 0.7427, 0.6271, 0.7974].

Step 5: iterative process

Following the first iteration, the same process has to be followed. Each concept’s values have to be recalculated repeatedly each time for each iteration stage until the equilibrium state is reached. Equilibrium occurs right when the difference between the values in consecutive iterations is sufficiently small (e.g., $\mathrm{\Delta }C$ < 0.001).

Mathematical expressions in detail

After the first iteration, the value of C1 (Player Knowledge) is:$$C_1^{\left(t,+,1\right)}=f(0·C_1^{\left(t\right)}+0.3·C_2^{\left(t\right)}+0.4·C_3^{\left(t\right)}-0.2·C_4^{\left(t\right)}+0.5·C_5^{\left(t\right)}+0.3·C_6^{\left(t\right)}-0.3·C_7^{\left(t\right)}+0.2·C_8^{\left(t\right)}-0.1·C_9^{\left(t\right)}+0.4·C_{10}^{\left(t\right)})$$

Replacing the initial values for $\mathit{Ci}$ (from $C^{\left(0\right)}$)$$\begin{array}{c}{C}_1^{\left(1\right)}=f\left(0+0.3·0.3+0.4·0.4-0.2·0.6+0.5·0.2+0.3·0.7-0.3·0.3+0.2·0.4-0.1·0.5+0.4·0.6\right)\\ C_1^{\left(1\right)}=f\left(0+0.09+0.16-0.12+0.1+0.21-0.09+0.08-0.05+0.24\right)\\ C_1^{\left(1\right)}=f\left(0.62\right)\\ f\left(x\right)=\frac{1}{1+e^{-0.62}}\approx 0.65\\ \end{array}$$

After the first iteration is completed, the new values for all the concepts are:$$C_1^{\left(1\right)}=[\text{0.6502,0.6726,0.7841,0.6341,0.7773,0.7231,0.3612,0.7427,0.6271,0.7974}]$$

Second iteration

Using the update equation:$$C_1^{(2)}=f({\sum }_{j=1}^{10}{w}_{1j}{C}_j^{(1)})$$

Replacing the weights $w_{1j}$ and the concept values $C_1^{(2)}$$$\begin{array}{cc}\hfill C_1^{\left(2\right)}& =f\left(0·C_1^{\left(1\right)}+0.3·C_2^{\left(1\right)}+0.4·C_3^{\left(1\right)}-0.2·C_4^{\left(1\right)}+0.5·C_5^{\left(1\right)}+0.3·C_6^{\left(1\right)}-0.3·C_7^{\left(1\right)}+0.2·C_8^{\left(1\right)}-0.1·C_9^{\left(1\right)}+0.4·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(0+0.3·0.6726+0.4·0.7841-0.2·0.6341+0.5·0.7773+0.3·0.7231-0.3·0.3612+0.2·0.7427-0.1·0.6271+0.4·0.7974\right)\hfill \\ \hfill & =f\left(0+0.2018+0.3136-0.1268+0.3886+0.2169-0.1084+0.1485-0.0627+0.3189\right)\hfill \\ \hfill & =f\left(1.2905\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-1.2905}}\approx 0.7842\hfill \\ \hfill \end{array}$$

Calculating $C_2^{\left(2\right)}$$$\begin{array}{cc}\hfill C_2^{\left(2\right)}& =f\left(0.2·C_1^{\left(1\right)}+0·C_2^{\left(1\right)}+0.4·C_3^{\left(1\right)}-0·C_4^{\left(1\right)}+0.2·C_5^{\left(1\right)}+0.1·C_6^{\left(1\right)}-0.1·C_7^{\left(1\right)}+0.5·C_8^{\left(1\right)}-0·C_9^{\left(1\right)}+0.3·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(0.2·0.6502+0+0.4·0.7841+0+0.2·0.7773+0.1·0.7231-0.1·0.3612+0.5·0.7427+0+0.3·0.7974\right)\hfill \\ \hfill & =f\left(0.1300+0+0.3136+0+0.1555+0.0723-0.0361+0.3713+0+0.2392\right)\hfill \\ \hfill & =f\left(1.2458\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-1.2458}}\approx 0.7766\hfill \\ \hfill \end{array}$$

Calculating $C_3^{\left(2\right)}$$$\begin{array}{cc}\hfill C_3^{\left(2\right)}& =f\left(0.3·C_1^{\left(1\right)}+0.5·C_2^{\left(1\right)}+0·C_3^{\left(1\right)}-0.4·C_4^{\left(1\right)}+0.5·C_5^{\left(1\right)}+0.4·C_6^{\left(1\right)}-0.2·C_7^{\left(1\right)}+0.1·C_8^{\left(1\right)}-0.3·C_9^{\left(1\right)}+0.4·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(0.3·0.6502+0.5·0.6726+0+0.4·0.6341+0.5·0.7773+0.4·0.7231-0.2·0.3612+0.1·0.7427+0.3·0.6271+0.4·0.7974\right)\hfill \\ \hfill & =f\left(0.1951+0.3363+0+0.2536+0.3886+0.2892-0.0722+0.0743+0.1881+0.3189\right)\hfill \\ \hfill & =f\left(1.9720\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-1.9720}}\approx 0.8778\hfill \\ \hfill \end{array}$$

Calculating $C_4^{\left(2\right)}$$$\begin{array}{cc}\hfill C_4^{\left(2\right)}& =f\left(-0.3·C_1^{\left(1\right)}+0.2·C_2^{\left(1\right)}+0.4·C_3^{\left(1\right)}+0+0.1·C_5^{\left(1\right)}+0.2·C_6^{\left(1\right)}-0.4·C_7^{\left(1\right)}+0.1·C_8^{\left(1\right)}+0.2·C_9^{\left(1\right)}+0.5·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(-0.3·0.6502+0.2·0.6726+0.4·0.7841+0+0.1·0.7773+0.2·0.7231-0.4·0.3612+0.1·0.7427+0.2·0.6271+0.5·0.7974\right)\hfill \\ \hfill & =f\left(-0.1951+0.1345+0.3136+0+0.0777+0.1446-0.1445+0.0743+0.1254+0.3987\right)\hfill \\ \hfill & =f\left(0.9291\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-0.9291}}\approx 0.7169\hfill \\ \hfill \end{array}$$

Calculating $C_5^{\left(2\right)}$$$\begin{array}{cc}\hfill C_5^{\left(2\right)}& =f\left(0.5·C_1^{\left(1\right)}+0.2·C_2^{\left(1\right)}+0.4·C_3^{\left(1\right)}+0.3·C_4^{\left(1\right)}+0.4·C_6^{\left(1\right)}-0.2·C_7^{\left(1\right)}+0.1·C_8^{\left(1\right)}+0.2·C_9^{\left(1\right)}+0.4·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(0.5·0.6502+0.2·0.6726+0.4·0.7841+0.3·0.6341+0+0.4·0.7231-0.2·0.3612+0.1·0.7427+0.2·0.6271+0.4·0.7974\right)\hfill \\ \hfill & =f\left(0.3251+0.1345+0.3136+0.1902+0+0.2892-0.0722+0.0743+0.1254+0.3189\right)\hfill \\ \hfill & =f\left(1.6990\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-1.6990}}\approx 0.8454\hfill \\ \hfill \end{array}$$

Calculating $C_6^{\left(2\right)}$$$\begin{array}{cc}\hfill C_6^{\left(2\right)}& =f\left(0.3·C_1^{\left(1\right)}+0.2·C_2^{\left(1\right)}+0.5·C_3^{\left(1\right)}+0.3·C_4^{\left(1\right)}+0.4·C_5^{\left(1\right)}+0·C_6^{\left(1\right)}-0.3·C_7^{\left(1\right)}+0.1·C_8^{\left(1\right)}+0.2·C_9^{\left(1\right)}+0.4·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(0.3·0.6502+0.2·0.6726+0.5·0.7841+0.3·0.6341+0.4·0.7773+0-0.3·0.3612+0.1·0.7427+0.2·0.6271+0.4·0.7974\right)\hfill \\ \hfill & =f\left(0.1951+0.1345+0.3920+0.1902+0.3109+0-0.1084+0.0743+0.1254+0.3189\right)\hfill \\ \hfill & =f\left(1.6330\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-1.6330}}\approx 0.8365\hfill \\ \hfill \end{array}$$

Calculating $C_7^{\left(2\right)}$$$\begin{array}{cc}\hfill C_7^{\left(2\right)}& =f\left(-0.4·C_1^{\left(1\right)}-0.1·C_2^{\left(1\right)}-0.2·C_3^{\left(1\right)}+0.3·C_4^{\left(1\right)}-0.3·C_5^{\left(1\right)}-0.3·C_6^{\left(1\right)}+0·C_7^{\left(1\right)}-0.2·C_8^{\left(1\right)}-0.3·C_9^{\left(1\right)}+0.1·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f\left(-0.4·0.6502-0.1·0.6726-0.2·0.7841+0.3·0.6341-0.3·0.7773-0.3·0.7231+0-0.2·0.7427-0.3·0.6271+0.1·0.7974\right)\hfill \\ \hfill & =f\left(-0.2601-0.0673-0.1568+0.1902-0.2332-0.2169+0-0.1485-0.1881+0.0797\right)\hfill \\ \hfill & =f\left(-0.8010\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-0.8010}}\approx 0.3098\hfill \\ \hfill \end{array}$$

Calculating $C_8^{\left(2\right)}$$$\begin{array}{cc}\hfill C_8^{\left(2\right)}& =f\left(0.3·C_1^{\left(1\right)}+0.5·C_2^{\left(1\right)}+0.4·C_3^{\left(1\right)}+0.2·C_4^{\left(1\right)}+0.2·C_5^{\left(1\right)}+0.3·C_6^{\left(1\right)}-0.2·C_7^{\left(1\right)}+0·C_8^{\left(1\right)}+0.1·C_9^{\left(1\right)}+0.4·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f(0.3·0.6502+0.5·0.6726+0.4·0.7841+0.2·0.6341+0.2·0.7773+0.3·0.7231\hfill \\ \hfill & -0.2·0.3612+0+0.1·0.6271+0.4·0.7974)\hfill \\ \hfill & =f\left(0.1951+0.3363+0.3136+0.1268+0.1555+0.2169-0.0722+0+0.0627+0.3189\right)\hfill \\ \hfill & =f\left(1.6537\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-1.6537}}\approx 0.8394\hfill \\ \hfill \end{array}$$

Calculating $C_9^{\left(2\right)}$$$\begin{array}{cc}\hfill C_9^{\left(2\right)}& =f\left(-0.2·C_1^{\left(1\right)}+0.1·C_2^{\left(1\right)}+0.2·C_3^{\left(1\right)}+0.3·C_4^{\left(1\right)}+0.1·C_5^{\left(1\right)}+0.2·C_6^{\left(1\right)}-0.3·C_7^{\left(1\right)}+0.2·C_8^{\left(1\right)}+0·C_9^{\left(1\right)}+0.3·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f(-0.2·0.6502+0.1·0.6726+0.2·0.7841+0.3·0.6341+0.1·0.7773+0.2·0.7231\hfill \\ \hfill & -0.3·0.3612+0.2·0.7427+0+0.3·0.7974)\hfill \\ \hfill & =f\left(-0.1300+0.0673+0.1568+0.1902+0.0777+0.1446-0.1084+0.1485+0+0.2392\right)\hfill \\ \hfill & =f\left(0.7860\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-0.7860}}\approx 0.6869\hfill \\ \hfill \end{array}$$

Calculating $C_{10}^{\left(2\right)}$$$\begin{array}{cc}\hfill C_{10}^{\left(2\right)}& =f\left(0.5·C_1^{\left(1\right)}+0.3·C_2^{\left(1\right)}+0.4·C_3^{\left(1\right)}+0.4·C_4^{\left(1\right)}+0.5·C_5^{\left(1\right)}+0.4·C_6^{\left(1\right)}+0.1·C_7^{\left(1\right)}+0.3·C_8^{\left(1\right)}+0.2·C_9^{\left(1\right)}+0·C_{10}^{\left(1\right)}\right)\hfill \\ \hfill & =f(0.5·0.6502+0.3·0.6726+0.4·0.7841+0.4·0.6341+0.5·0.7773+0.4·0.7231\hfill \\ \hfill & +0.1·0.3612+0.3·0.7427+0.2·0.6271+0)\hfill \\ \hfill & =f\left(0.3251+0.2018+0.3136+0.2536+0.3886+0.2892+0.0361+0.2228+0.1254+0\right)\hfill \\ \hfill & =f\left(2.1562\right)\hfill \\ \hfill f\left(x\right)& =\frac{1}{1+e^{-2.1562}}\approx 0.8962\hfill \\ \hfill \end{array}$$

Summary of second iteration

After calculating all concepts, the new values for the second iteration are:$$C^{\left(2\right)}=[\text{0.7842,0.7766,0.8778,0.7169,0.8454,0.8365,0.3098,0.8394,0.6869,0.8962}]$$

Convergence check

If the FCM has reached equilibrium, the difference between the current iteration's concept values and the previous iteration's values must be assessed to identify if it is less than δ = 0.001.

Calculations of the absolute difference for each concept:$$\begin{array}{c}\mathrm{\Delta }C=\left|C^{\left(2\right)}-C^{\left(1\right)}\right|\\ \mathrm{\Delta }{C}_1=\left|0.7842-0.6502\right|=0.1340\\ \mathrm{\Delta }{C}_2=\left|0.7766-0.6726\right|=0.1040\\ \mathrm{\Delta }{C}_3=\left|0.8778-0.7841\right|=0.0937\\ \mathrm{\Delta }{C}_4=\left|0.7169-0.6341\right|=0.0828\\ \mathrm{\Delta }{C}_5=\left|0.8454-0.7773\right|=0.0681\\ \mathrm{\Delta }{C}_6=\left|0.8365-0.7231\right|=0.1134\\ \mathrm{\Delta }{C}_7=\left|0.3098-0.3612\right|=0.0514\\ \mathrm{\Delta }{C}_8=\left|0.8394-0.7427\right|=0.0967\\ \mathrm{\Delta }{C}_9=\left|0.6869-0.6271\right|=0.0598\\ \mathrm{\Delta }{C}_{10}=\left|0.8962-0.7974\right|=0.0988\\ \end{array}$$

Based on the above, 12 iterations were performed until none of the differences between iterations are greater than the predefined threshold points δ = 0.001. The results of the iterations are shown in the tables below.

Table 4 shows the concept values at different iteration stages in the FCM system. The values demonstrate how the learning environment dynamically adapts based on learner interactions to optimize engagement and knowledge retention.

Table 4. Concept values for each iteration

	Iteration i
	3	4	5	6	7	8	9	10	11	12
${\mathit{C}}_1^{\left(\mathit{\iota }\right)}$	0.8481	0.8849	0.9059	0.9188	0.9251	0.9285	0.9310	0.9325	0.9329	0.9331
${\mathit{C}}_2^{\left(\mathit{\iota }\right)}$	0.8226	0.8502	0.8670	0.8772	0.8801	0.8829	0.8852	0.8865	0.8871	0.8875
${\mathit{C}}_3^{\left(\mathit{\iota }\right)}$	0.9208	0.9446	0.9576	0.9649	0.9681	0.9698	0.9708	0.9715	0.9717	0.9718
${\mathit{C}}_4^{\left(\mathit{\iota }\right)}$	0.7667	0.7978	0.8178	0.8318	0.8372	0.8402	0.8425	0.8435	0.8439	0.8440
${\mathit{C}}_5^{\left(\mathit{\iota }\right)}$	0.8878	0.9138	0.9298	0.9406	0.9450	0.9475	0.9490	0.9498	0.9500	0.9501
${\mathit{C}}_6^{\left(\mathit{\iota }\right)}$	0.8779	0.9036	0.9189	0.9299	0.9351	0.9373	0.9387	0.9394	0.9396	0.9392
${\mathit{C}}_7^{\left(\mathit{\iota }\right)}$	0.2986	0.2939	0.2914	0.2901	0.2896	0.2894	0.2893	0.2893	0.2892	0.2892
${\mathit{C}}_8^{\left(\mathit{\iota }\right)}$	0.8776	0.9023	0.9172	0.9270	0.9315	0.9337	0.9352	0.9356	0.9358	0.9358
${\mathit{C}}_9^{\left(\mathit{\iota }\right)}$	0.7239	0.7458	0.7589	0.7677	0.7710	0.7732	0.7744	0.7748	0.7750	0.7751
${\mathit{C}}_{10}^{\left(\mathit{\iota }\right)}$	0.9354	0.9565	0.9681	0.9752	0.9773	0.9792	0.9803	0.9808	0.9810	0.9811

Table 5 shows the threshold values of each iteration of the FCM calibration process. The table’s data show how the system reaches equilibrium state after the iterations, delivering stable adaptive learning responses.

Table 5. Threshold Points for FCM Convergence

	Iteration i
	3	4	5	6	7	8	9	10	11	12
$\mathbf{\Delta }{\mathit{C}}_1^{\left(\mathit{\iota }\right)}$	0.0639	0.0368	0.0210	0.0129	0.0063	0.0034	0.0025	0.0015	0.0004	0.0002
$\mathbf{\Delta }{\mathit{C}}_2^{\left(\mathit{\iota }\right)}$	0.0460	0.0276	0.0168	0.0102	0.0029	0.0028	0.0023	0.0013	0.0006	0.0004
$\mathbf{\Delta }{\mathit{C}}_3^{\left(\mathit{\iota }\right)}$	0.0430	0.0238	0.0130	0.0073	0.0032	0.0017	0.0010	0.0007	0.0002	0.0001
$\mathbf{\Delta }{\mathit{C}}_4^{\left(\mathit{\iota }\right)}$	0.0498	0.0311	0.0200	0.0140	0.0054	0.0030	0.0023	0.0010	0.0004	0.0001
$\mathbf{\Delta }{\mathit{C}}_5^{\left(\mathit{\iota }\right)}$	0.0424	0.0260	0.0160	0.0108	0.0044	0.0025	0.0015	0.0008	0.0002	0.0001
$\mathbf{\Delta }{\mathit{C}}_6^{\left(\mathit{\iota }\right)}$	0.0414	0.0257	0.0153	0.0110	0.0052	0.0022	0.0014	0.0007	0.0002	0.0004
$\mathbf{\Delta }{\mathit{C}}_7^{\left(\mathit{\iota }\right)}$	0.0112	0.0047	0.0025	0.0013	0.0005	0.0002	0.0001	0.0000	0.0001	0.0000
$\mathbf{\Delta }{\mathit{C}}_8^{\left(\mathit{\iota }\right)}$	0.0382	0.0247	0.0149	0.0098	0.0045	0.0022	0.0015	0.0004	0.0002	0.0000
$\mathbf{\Delta }{\mathit{C}}_9^{\left(\mathit{\iota }\right)}$	0.0370	0.0219	0.0131	0.0088	0.0033	0.0022	0.0012	0.0004	0.0002	0.0001
$\mathbf{\Delta }{\mathit{C}}_{10}^{\left(\mathit{\iota }\right)}$	0.0392	0.0211	0.0116	0.0071	0.0021	0.0019	0.0011	0.0005	0.0002	0.0001

As of right now, none of the differences between iterations are greater than the predefined threshold points δ = 0.001. This indicates that after the 12 th iteration, equilibrium has been reached.

By comparing a group using an FCM-based adaptive hint system with another using a static, linear hint system, the study thoroughly evaluates the VR education game's effectiveness in user performance, engagement, and learning. It also examines the advantages that adaptive learning mechanisms bring, such as deep comprehension and motivation. The following section describes the methodology to be used in assessing the impact the game has on improving both the learning experience and performance of participants.

Participants and experimental setup

The participants in this study were students, both at undergraduate and postgraduate levels, attending lessons at the Department of Informatics and Computer Engineering of the University of West Attica and who attended lessons during the second semester of the academic year 2023–2024. 140 students were split into two groups of 70 students each at random:

VR-FCM group (n = 70):

The VR-FCM Group participated in a VR educational game that combines FCMs and adaptive mechanisms to teach Java programming concepts. As players advance through the game, FCMs are in charge of providing a personalized in-game hint system and adapting the difficulty levels for each player according to their skills and knowledge in order to improve learning outcomes and player engagement.

VR-Hint group (n = 70):

The VR-Hint Group played the VR game, which was the same as the VR-FCM Group and used the same user interface, apparatus and learning objectives. The main and important difference between these two groups is that the VR-Hint Group delivered a learning experience to the users with a hint system that used a linear approach to adjust the difficulty. The hint system delivered hints by taking into consideration the number of errors and completion times without making use of the dynamic personalization that FCMs deliver.

Table 6 presents study participants'demographic distribution by gender, level of degree, and experience with VR. In addition to the information presented in Table 6, participants'ages ranged from 19 to 27 years, with a mean age of 22.3 years (SD = 2.1). Among undergraduates, most were in their third or fourth year of study, while postgraduate participants were enrolled in the first year of their program. Regarding programming experience, participants reported between 1 and 4 years of prior experience, with an average of 2.3 years (SD = 0.9). The distribution of gender, degree level, and prior VR experience between the VR-FCM and VR-Hint groups was comparable, ensuring the validity of between-group comparisons and supporting future replication under similar conditions.

Table 6. Participant demographics

Group	Gender	Degree level	Prior VR experience
VR-FCM group	45% male (n = 32)	60% undergraduates (n = 42)	30% prior experience (n = 21)
	50% female (n = 35)	40% postgraduates (n = 28)	70% no prior experience (n = 49)
	5% preferred not to define (n = 3)
VR-hint group	48% male (n = 34)	58% undergraduates (n = 41)	38% prior experience (n = 21)
	50% female (n = 35)	42% postgraduates (n = 29)	62% no prior experience (n = 49)
	2% preferred not to define (n = 1)

Ensuring that both groups balance the demographic distribution will ultimately produce more solid results with the comparability in gender, degree level and previous experience with VR. All students who had participated in the experimental stages took part in a 45-min session playing the game, but also in filling out questionnaires relevant to the intervention. The questionnaires were to measure the learning outcome, motivation, user experience and effectiveness of the hint system mechanics in the game.

Game structure overview

Meta Quest 2 VR was the HMD that players used to experience the VR educational game, which was developed by employing Unity 3D. The goal of the game's development was to teach students about Java programming loops. To accomplish this, the VR-FCM Group utilized Gamification, Flow Theory and Fuzzy Cognitive Maps (FCMs) to create an engaging and adaptable educational VR game for players. However, the VR-Hint Group students played the same game without the adaptive features that the FCMs feedback system offers.

• Level 1: Introduction to Syntax

  In this level, players have to answer quiz questions about Java Programming that are provided by an NPC (Non-Player Character) who is located in a virtual village in order to move on to the next stage.

• Level 2: Logical Sequencing

  The players advance to the second level after completing the first one successfully. In this level they must activate four torches in the right order and form a line of Java code related to loops.

• Level 3: Debugging and Optimization

  After completing the second level successfully, players must match the right stones that project a line of code with programming statements pertaining to Java loops (represented by rupees).

For all three levels:

VR-FCM Group: The player can view the hints as a result of the real-time FCM mechanism, which dynamically modifies their support and difficulty levels based on the player's engagement and knowledge level.

VR-Hint Group: This group uses a simple hint system with predetermined conditions, such as a prolonged period of time or player errors that are used to deliver this system. The main difference between the VR-FCM Group and the VR-Hint Group is that the latter does not make use of the personalized adaptation provided by FCM.

The fact that the game is designed in such a way that both the control and the experimental groups are exposed to the same content and user interface enables players to assess the FCM adaptive feedback as the primary differential between the groups.

Study procedure

To analyze and evaluate the effectiveness of the VR game and focus on the Fuzzy Cognitive Maps (FCMs) in adaptive learning, questionnaires were employed. Both the VR-FCM Group and VR-Hint Group have completed a series of the same type of questionnaires thus maintaining the congruency of the data collected across the groups:

• Presence Questionnaire: The Presence Questionnaire is a 19-item questionnaire, which is based on the framework created by Witmer & Singer (Witmer & Singer, 1998). This questionnaire examines and quantifies the players'sense of presence and control within the VR environment. For this reason, the questions were required to be rated on a 7-point Likert scale ranging from 1 ("Not at all") to 7 ("Completely"). Aspects such as the naturalness of interaction, control over events and involvement in the VR environment were required to be answered.

• Motivated Strategies for Learning Questionnaire (MSLQ): This questionnaire included 44 items that had to be filled by the players (Pintrich et al., 1993). This questionnaire was meant to measure and report participants'motivation and learning strategies. These included dimensions of VR interventions related to intrinsic motivation, self-regulation and test-taking strategies on a scale of 7 points ranging from 1 ("Not true of me at all") to 7 ("Very true of me"). In addition to motivation, the MSLQ was also used to assess learner engagement. Engagement was operationalized using the Cognitive and Metacognitive Strategies and Time and Study Environment Management subscales of the MSLQ (Pintrich et al., 1993). These subscales capture students’ active involvement in their learning process, including planning, monitoring, and time management—core indicators of academic engagement. Participants rated each item on a 7-point Likert scale ranging from 1 (“Not true of me at all”) to 7 (“Very true of me”). These subscales have demonstrated acceptable internal consistency in similar educational research contexts (Pintrich et al., 1993).

• User Experience Questionnaire (UEQ): This questionnaire was delivered to the players. 26 items were rated on a 7-point scale with contrasting pairs, such as annoying, enjoyable, boring and exciting, addressing aspects such as user experience with the VR game, as well as usability, enjoyment and perceived effectiveness (Schrepp et al., 2017).

• Knowledge Questionnaire: To assess conceptual understanding of the programming concepts addressed in the VR game (particularly Java loops), a 20-item Knowledge Questionnaire was administered both before and after the intervention. The questionnaire was designed by the research team and reviewed by two university-level programming instructors to ensure content validity and alignment with the learning objectives of the VR experience.

• The questionnaire included multiple-choice questions targeting core Java concepts taught during the VR experience, such as loop syntax, control flow and debugging logic. Items were designed to assess both surface-level recall and deeper conceptual understanding.

• Example items include:

• “What is the correct syntax for a ‘for’ loop that runs 5 times?”

• “Which type of loop is best when the number of iterations is unknown?”

• “What is the output of: int i = 0; while (i < 2) { System.out.print(i); i + +;}”

• A pilot test with 10 students was conducted to assess clarity and timing. The internal consistency of the post-test questionnaire yielded a Cronbach’s alpha of 0.81, indicating good reliability.

• Perception of AI-Driven Feedback Questionnaire: In addition to the above measures, participants in the VR-FCM group completed a Perception of AI-Driven Feedback Questionnaire. This custom-developed instrument consisted of 10 items rated on a 7-point Likert scale ranging from 1 ("Strongly disagree") to 7 ("Strongly agree"). The questionnaire was designed to assess students’ perceptions of the usefulness, clarity, and effectiveness of the real-time AI-driven personalized feedback provided during gameplay. Items targeted aspects such as the helpfulness of hints, the appropriateness of difficulty adjustments, and the perceived impact of feedback on learning progress.A pilot test was conducted with 10 students to ensure the clarity and relevance of the items. The internal consistency of the questionnaire was high, with a Cronbach’s alpha of 0.86. Representative items include:

“The AI mechanisms helped me better understand the content.”

“The game’s assistance helped me understand complex concepts effectively."The full Perception of AI-Driven Feedback Questionnaire is provided in Table 14 of the Appendix.

Minor contextual adaptations were made to the original instruments to ensure their relevance to the VR/AI-based educational environment while maintaining their established structure and validity. Specifically, wording adjustments were applied to the Presence Questionnaire items to better reflect interaction with an immersive VR learning system rather than traditional VR simulations (e.g., emphasizing educational task performance and system responsiveness). Similarly, the MSLQ was used without altering the core constructs but framed in instructions to refer explicitly to the VR-based programming activity rather than general academic tasks. For the User Experience Questionnaire (UEQ), the original bipolar adjectives were preserved, but the introductory instructions were adjusted to focus on evaluating the VR learning environment. No changes were made to the content or rating scales of the instruments to ensure comparability with existing validated versions.

The VR-FCM and VR-Hint groups'immersion, motivation, user experience and learning efficacy can be directly compared by gathering data from both groups, excluding the impact of FCM's in-game mechanics on the VR-Hint Group.

Results

This section presents the findings of the performance metrics, user engagement and learning outcomes. Key comparisons involve the VR-FCM and VR-Hint groups, with a focus on the impact of adaptive feedback on learning effectiveness. The detailed analysis of results is presented in the sections that follow.

Task performance and error analysis

The game held some important data metrics regarding task completion times, error rates and hint usage of both groups. These data are useful for assessing the effectiveness of the adaptive hint system that FCMs deliver for the students for the VR-FCM Group.

Table 7 presents key performance metrics comparing the two study groups. It includes task completion time, error rates and frequencies in utilizing hints and provides an overview of how adaptive FCM-driven feedback influences learning efficiency and engagement.

Table 7. Performance metrics for VR-FCM and VR-hint groups

Metric	VR-FCM group (Mean ± SD)	VR-hint group (Mean ± SD)
Average task completion time	10.5 ± 2.1 min	12.3 ± 2.8 min
Average error rate per task	2.5 ± 0.9 errors	3.6 ± 1.2 errors
Hint usage frequency per task	1.8 ± 0.7 hints	2.6 ± 1.0 hints
Percentage of tasks requiring hints	55%	68%
Tasks completed without errors	30%	20%
Average time on most complex task	14.2 ± 3.2 min	17.0 ± 3.6 min
Hint request rate on complex task	65%	75%

Statistical analysis: ANOVA

The VR-FCM and VR-Hint groups'performance was compared across the measured metrics using a one-way ANOVA for Completion Time, Error Rate and Hint Usage:

Table 8 provides a statistical comparison and further supports the effectiveness of the adaptive FCM-based system. A one-way ANOVA test was conducted and revealed a difference in task completion time (F(1,138) = 21.78, p < 0.001), error rate (F(1,138) = 58.40, p < 0.001), and use of hints (F(1,138) = 33.77, p < 0.001) between groups. The VR-FCM group performed better than the VR-Hint group with faster completion times, fewer errors and use of fewer hints. These findings supported the adaptive system's impact on learning efficiency and interest.

Table 8. Statistical analysis of learning outcomes

Metric	F-statistic	p-value	Interpretation
Completion time	21.78	7.15e-06	Statistically significant difference: VR-FCM group completed tasks faster
Error rate	58.40	3.26e-12	Statistically significant difference. VR-FCM group made fewer errors
Hint usage	33.77	4.09e-08	Statistically significant difference. VR-FCM group required fewer hints

Figure 6 shows a bar chart comparing task completion time, error rate, and use of hints in the VR-FCM and VR-Hint groups. The results reveal that adaptive FCM-based feedback improves performance by reducing errors and use of hints without diminishing engagement.

[See PDF for image]

Fig. 6

Comparison of task performance, error rate, and hint usage in VR learning systems

Interpretation of results

Following ANOVA analysis, the three metrics (Completion Time, Error Rates and Hint Usage) revealed significant differences (p < 0.001) between the two groups:

1.Completion Time When compared to students in the VR-Hint group, the VR-FCM group had faster task completion times. This indicates how well the feedback mechanisms improve the system's efficiency.

Figure 7 depicts the average task completion time for each task in both study groups. In this figure, it can be observed that the students who participated in the VR-FCM group were faster in task performance when compared to those in the VR-Hint group. These results make evident the effectiveness of adaptive real-time feedback in achieving optimal learning speed.

[See PDF for image]

Fig. 7

Average task completion time per task

2.Error Rate Fewer errors were made by students in the VR-FCM group intervention. This shows how the Fuzzy Cognitive Maps feedback system's personalized hints efficiently supported learning and minimized students'errors.

Figure 8 shows error rate analysis by task for both study groups. Error rate refers to the percentage of faulty attempts by task and shows how adaptive support impacts learning performance.

[See PDF for image]

Fig. 8

Average error rate per task

Figure 9 shows the percentage of error-free task completion in both study groups. The VR-FCM group had a much higher percentage of error-free task completion, which demonstrates that adaptive learning can be effective in skill acquisition and reducing user frustration.

[See PDF for image]

Fig. 9

Percentage of tasks completed without errors

3.Hint Usage Compared to students in the VR-Hint group, participants in the VR-FCM group received considerably fewer hints. This indicates that students in the VR-FCM group comprehended the subject matter better and depended less on external assistance than those in the VR-Hint group.

Figure 10 shows the number of hints used per task in both groups. The VR-FCM group relied less on hints, while the VR-Hint group required more frequent external assistance. This suggests that adaptive personalization fosters independent learning by gradually reducing reliance on system-provided hints.

[See PDF for image]

Fig. 10

Hint usage frequency per task

Figure 11 depicts the percentage of tasks on which hints were required from players in order to progress. The VR-FCM group received fewer hints than did the VR-Hint group in support of the observation that adaptive intervention supported learners'ability in task completion.

[See PDF for image]

Fig. 11

Percentage of tasks requiring hints

Figure 12 depicts rates of hint requests for hard tasks for both groups. The group that employed VR-FCM requested hints at a rate that was lower than that of the group that employed VR-Hint, indicating greater problem-solving ability through adaptive learning mechanisms.

[See PDF for image]

Fig. 12

Hint request rate on complex task

Compared with non-adaptive systems, these results clearly show that adaptive VR can improve instructional processes and produce better learning outcomes.

Participants in the VR-FCM Group completed tasks faster than those in the VR-Hint Group, as shown by shorter completion times. The VR-FCM Group received fewer hints and committed a much lower error rate. This reflects that the learning process was effectively supported by adaptive feedback mechanisms of the VR-FCM Group.

The findings of research identify that adaptive VR-based learning can be extended to other programming languages by modifying the Fuzzy Cognitive Map model to accommodate diverse syntax structures, debugging and problem-solving styles. Similar research has identified diverse learning approaches and their evolving nature in learning scenarios, particularly in higher education, where hybrid learning patterns have recorded significant growth and development over a period of time (Qi et al., 2024).

The flexibility enables extension to other languages such as Python, C + +, and JavaScript so that personalized adaptation is facilitated in a range of programming environments. The adaptive features of the system can be extended to other educational areas such as STEM subjects, medical education, and language education where real-time personalization is critical in enhancing learning outcomes. However, in spite of its advantages, adaptive personalization is prone to certain potential limitations like greater computation needs in real-time adaptation, need for large data for fine-tuning weights in the FCM model and inability to match adaptive difficulty levels with diverse learner needs. Future research needs to explore means by which the system can be made scalable so that these limitations are overcome and make it usable in a large variety of learning situations, as well as sustain student motivation over a long duration.

Results from questionnaires

Presence results

The Presence Questionnaire is to evaluate the level of the players'sense of presence and control in the virtual world.

The results from the Presence Questionnaire Results showed that the students of the VR-FCM Group scored higher on the items related to immersion and control. They scored 5.3 on average in interaction control while the students of R-Hint Group scored 4.8. Additionally, students of the VR-FCM Group scored 5.4 on sense of movement compared to 5.0 on what the students of the VR-Hint Group scored.

Figure 13 shows the Presence Questionnaire results for Sense of Movement and Interaction Control. The results indicate that participants in the VR-FCM group reported a greater sense of smooth movement and precise interaction control in the VR environment, contributing to a more immersive and responsive learning experience.

[See PDF for image]

Fig. 13

Presence questionnaire results—sense of movement and interaction control

According to these comparisons, FCM-based feedback enhances the VR experience, providing the students with a feeling of being more engaged and intuitive within the environment. According to the results, adaptive mechanisms can enhance user immersion and control in VR-based education.

Motivation and learning strategies (MSLQ results)

The MSLQ assesses motivation, learning strategies and self-regulation dimensions. The VR-FCM and VR-Hint groups rated all 44 items.

The MSLQ results revealed major differences in the two groups'methods to learning and motivation. The average score for the VR-FCM Group's students on intrinsic motivation items was 5.4. On the contrary, the VR-Hint Group received a score of 4.4. On items related to self-regulation, students in the VR-FCM Group also scored an average of 5.2 on average, while students in the VR-Hint Group scored an average of 4.3.

Figure 14 displays results from the Motivated Strategies for Learning Questionnaire (MSLQ) measuring Intrinsic Motivation and Self-Regulation. The results indicate that students in the VR-FCM group reported higher intrinsic motivation and improved self-regulation skills than students in the VR-Hint group, reinforcing the role of adaptive learning mechanisms in fostering self-directed engagement.

[See PDF for image]

Fig. 14

MSLQ results—intrinsic motivation and self-regulation

These differences show that adaptive feedback assisted by FCMs delivered a more stimulating VR-based education. It also motivated students to be more engaged with the learning material. These results make it evident that adaptive mechanisms enhance motivation and learning outcomes within the VR educational context.

User experience and learning effectiveness (UEQ, VR group only)

Students provided valuable responses to the User Experience Questionnaire (UEQ) about their experiences playing the VR game. UEQ's primary goal is to evaluate aspects like the game's learning effectiveness, enjoyment and usability. Answers for all 26 items were provided by the VR-FCM and VR-Hint groups.

The UEQ produced results based on the students’ perceptions and showed that the VR-FCM group evaluated the VR experience higher in general than the VR-Hint group. To be more specific, the VR-FCM group’s score on enjoyment was 5.6 and 5.0 for the VR-Hint group. A notable difference was also observed in the results related to ease of learning, noting that the VR-FCM group scored 5.5 versus 5.1 for the VR-Hint group.

Figure 15 presents the outcomes of the User Experience Questionnaire (UEQ) measuring Perceived Usability and Learning Experience. The participants in the group that used VR-FCM found the system more usable and enjoyable and more effective in overall learning, suggesting that adaptive personalization makes learning more effective in VR learning environments.

[See PDF for image]

Fig. 15

User experience questionnaire (UEQ) results—perceived usability and learning experience

These results show that the VR-FCM group delivers a more enjoyable, intuitive and engaging experience than the VR-Hint group, meaning that adaptive feedback had a positive impact on the users’ experience in the VR educational context.

Perception of AI-driven feedback

To evaluate how the students perceived AI Mechanisms and overall experience, related questionnaires were provided to the ones who participated in the VR-FCM Group. Students were asked to complete the questionnaires and rate how effective the FCM-driven system was in delivering personalized support. They were also asked how it improved their learning process and how it enhanced their engagement. For this reason, they had to rate statements on a 5-point Likert scale, ranging from 1 ("Strongly Disagree") to 5 ("Strongly Agree").

Figure 16 illustrates the evaluation ratings for adaptive AI mechanisms in the FCM-based VR system. The results reveal high effectiveness in real-time adaptation in learning, engagement support and personalized guidance, confirming the advantage of intervention through AI in education in a VR-based system.

[See PDF for image]

Fig. 16

Evaluation of FCM-driven AI mechanisms

The results showed that the FCM mechanisms supporting the game averaged 4.5 on items measuring engagement support. The same group also scored 4.4 on how the students perceived the personalized feedback, showing how the system was effective in meeting the students’ learning requirements. These results show that FCM-based mechanisms can enhance student engagement and personalized assistance to improve learning outcomes.

Conceptual understanding of programming constructs

To assess students'conceptual understanding of programming constructs, a post-intervention Knowledge Questionnaire was administered. This test focused solely on specific questions related to Java and, more specifically, in programming loops so their comprehension of the subject matter during the VR interventions could be measured. The two groups completed the questionnaire containing the same items designed to assess their conceptual understanding of programming constructs.

Table 9 presents the results of the post-intervention Knowledge Questionnaire for both groups. The findings suggest that the adaptive FCM-based feedback enhanced students'conceptual understanding of Java programming constructs.

Table 9. Knowledge questionnaire results

Knowledge questionnaire results	VR-FCM group (Mean ± SD)	VR-hint group (Mean ± SD)
Average score (%)	85 ± 6	76 ± 8
Percentage of correct responses on complex questions (%)	83 ± 5	71 ± 7
Percentage of correct Responses on basic questions (%)	87 ± 6	79 ± 7

The VR-FCM group achieved an average score of 85%, compared to 76% in the VR-Hint group, indicating a higher level of conceptual understanding. It is essential to mention that the VR-FCM Group also scored 83% on average on more complex questions, while the VR-Hint group scored 71%. These results highlight the role of adaptive feedback mechanisms in supporting deeper conceptual understanding, particularly when the learning material increases in complexity.

Participant feedback and future improvements

Students of both groups were also asked for their feedback and suggestions about the VR experience. The feedback was collected via a questionnaire offering multiple-choice and open-ended questions. The primary objective of this questionnaire was to assess the potential dimensions of the system that make it effective or flawed and perhaps potential improvements.

Table 10 displays that each group's percentage of participants who reported that the VR system helped to understand programming concepts is provided. The responses demonstrate positive feedback in general towards the effectiveness of adaptive learning mechanisms.

Table 10. User feedback on VR system usefulness

Question	VR-FCM group—Yes	VR-FCM group—No	VR-hint group—Yes	VR-hint group—No
Did the VR system help you understand the content better?	92%	8%	78%	22%
Would you consider using VR for other educational purposes in the future?	88%	12%	74%	26%

88% of the participants of the VR-FCM Group and 74% in the VR-Hint Group stated that they would consider a VR educational application in the future. This highlights the potential and the value of VR as an educational approach. This is also promising because the students of the VR-FCM Group showed greater enthusiasm than the ones of the VR-Hint Group, showing that the adaptive mechanisms offered by FCMs can make VR a beneficial learning approach.

Open-ended feedback provided valuable insights into the system’s strengths and areas for improvement

The students of the VR-FCM Group stated that the interactions within the virtual environment that the VR experience offers and, most importantly, the adaptive feedback made the players feel more engaged and present within the VR environment. Adaptive feedback also helps students better comprehend programming concepts. Students who participated in VR-Hint Group also gave positive feedback on the immersive content, and some of them noted that it would be beneficial to have adaptive support through the learning experience as it would improve their experience.

An area in which this VR experience can be improved is the one related to technical issues. This includes smoother navigation and improved responsiveness and could improve usability.

The feedback provided shows the potential of the system as an educational approach providing adaptive feedback by utilizing FCMs to enhance user engagement and comprehension.

Discussion

This study highlights how Fuzzy Cognitive Maps'(FCMs') adaptive feedback mechanisms contribute to the effectiveness of VR learning environments. Two groups, one that played the VR game with the FCMs integrated into it, and the other that used standard, non-adaptive hints, were used for evaluating the game. These interventions produced valuable information about personalized assistance in VR learning environments.

This study is supported by previous works (Segura et al. 2020; Srimadhaven et al., 2020), which identified that gamified VR has a promising effect in enhancing engagement and logical thinking for programming education. Although these studies stand to validate the role of immersive learning environments, this study presents a significant advancement. Since FCMs are dynamic, nonlinear systems that allow personalized real-time tuning of feedback and level of challenges, they enable statistically significant improvement in task completion, reduction in error rates and enhanced conceptual understanding of programming constructs among subjects within the VR-FCM group. In this work, the combination of FCMs with Flow Theory and gamification, allows the challenge-skill balance to be handled dynamically, enabling a more comprehensive and adaptive learning experience than that compared to previously reported approaches.

The findings of this study reinforce prior research on the pedagogical benefits of immersive VR environments in programming education. For example, Segura et al. (2020) and Srimadhaven et al. (2020) demonstrated how gamified virtual environments can enhance learner motivation, engagement, and logical reasoning. Our results extend these insights by incorporating an adaptive mechanism based on Fuzzy Cognitive Maps (FCMs), which enables dynamic, real-time personalization of challenge levels and feedback. Unlike previous systems that relied on static or generalized interactions, the FCM-based system in our study significantly improved task completion times, reduced error rates, and optimized hint usage—demonstrating the advantages of adaptivity in immersive educational contexts.

In addition, the observed improvements in knowledge retention align with the findings of Abuhassna et al., (2022a, b), who emphasized the importance of adaptive learning environments in fostering deeper comprehension through tailored feedback. While those studies focused primarily on online and blended learning modalities, our work applies similar adaptive principles within an immersive VR framework. This integration of FCMs with pedagogical theories such as Flow Theory and Gamification presents a novel approach to supporting both the cognitive and motivational aspects of learning, contributing meaningfully to the design of intelligent, learner-centered educational technologies.

Learning outcomes and comprehension

The VR-FCM Group's students outperformed the VR-Hint Group on the Knowledge Questionnaire (85% vs. 76% for the VR-Hint Group). Additionally, students performed significantly better on challenging questions. These findings show that the students'understanding of the material was improved by adaptive, personalized feedback in the form of hints provided by the FCMs (Fig. 17).

[See PDF for image]

Fig. 17

Knowledge questionnaire performance comparison

The clear performance difference between the two groups shows that FCMs'adaptive hinting mechanisms work well when they aim to provide players with challenging content that they can comprehend more easily and effectively with the help of adaptive guiding mechanisms.

Engagement and motivation

According to the results of the Motivated Strategies for Learning Questionnaire (MSLQ), students'motivation and engagement are significantly impacted by the adaptive mechanisms. The students who took part in the VR-FCM Group scored 5.2 on self-regulation and 5.4 on intrinsic motivation. However, the students in the VR-Hint Group performed considerably less than those in the VR-FCM Group, scoring 4.4 on intrinsic motivation and 4.3 on self-regulation, respectively (Fig. 18).

[See PDF for image]

Fig. 18

MSLQ results–motivation and engagement

This difference may be explained by the effectiveness of the FCM-driven adaptive hint system that the players get, which, taken as a whole, makes the virtual environment more interactive and engaging and encourages them to interact with the virtual learning content and be more inclusive throughout the learning process.

This is in line with the concept of Flow Theory, which is that motivation and immersion are enhanced when challenge and skill are balanced properly.

Flow Theory and Gamification support one another in ensuring motivation in VR-based learning platforms through ensuring an appropriate balance between ability and challenge and introducing motivational elements that enhance engagement. Flow Theory is concerned with being totally immersed in an activity and achieving a psychological state of complete absorption when skills are in congruence with task difficulty (Csikszentmihalyi, 1997a, b). Within this research's educational game in VR, this equilibrium was dynamically maintained through Fuzzy Cognitive Maps (FCMs) so as not to cause disengagement through boredom (as a result of a lack of challenge) or through frustration (as a result of undue difficulty). Gamification features like reward schemes, progression systems, and adaptive difficulty are complemented by Flow Theory through enhancing motivation and engagement. Implementation of gamification in Flow-based learning has been reported to enhance persistence and cognitive engagement (Hamari et al., 2014). Interactive challenges, adaptive feedback, and progress based on milestones in our VR system are a few of the mechanisms that prompted learners to remain engaged by providing instant feedback and effort reward and hence sustained long-term motivation. The combination of these two frameworks in an adaptive VR system is a creative solution for fostering engagement in that it dynamically adjusts according to users'performance and provides a consistent learning experience based on individual needs.

The FCMs maintained participants'interest by adapting the levels of difficulty and offering hints based on real-time performance, which reduced frustration while keeping a level of difficulty appropriate for each learner.

The adaptability of the system is critical in providing a personalized learning experience to diverse learners, particularly beginners and advanced learners. With the help of Fuzzy Cognitive Maps (FCMs), the system dynamically adjusts task difficulty, scaffolding and feedback intensity real time. Greater guided support, more hints at more regular intervals and less demanding tasks are provided by the system to novices in order to prevent cognitive overload and gradually increase complexity as their proficiency increases. In contrast, more advanced learners encounter more complex issues, fewer hints and problem-solving challenges that promote independent learning and more active engagement. This is in accordance with research on adaptive learning environments that underscore the importance of adapting difficulty in real-time to facilitate optimal learning flow and reduce disengagement (Ma et al., 2014). Studies have determined that adaptive VR-based systems improve learning effectiveness and engagement by ensuring that each learner is challenged at a suitable level without being bored or overwhelmed (Khosravi et al., 2022). Our findings support these principles, demonstrating that the system successfully personalizes the learning experience, fostering engagement and enhancing skill development across diverse learner profiles.

User experience and presence

Results from the UEQ and Presence Questionnaire reveal that the adaptive feedback mechanism in VR-based educational environments has an enormous impact on positive User Experience and Presence. It was observed that students of the VR-FCM Group seemed to feel more in control (5.3 vs. 4.8), natural interaction (5.2 vs. 4.9) and overall enjoyment (5.6 vs. 5.0) as compared to the results of VR-Hint Group students. FCMs added into a VR learning environment not only enhance satisfaction but create more immersive and natural-like experiences for the target audience. According to these scores, adaptive feedback improves user engagement and gives students a chance to engage with the virtual environment in a meaningful manner, which helps them learn more effectively and strengthen any prior knowledge they may have. It also improves the user experience (Fig. 19).

[See PDF for image]

Fig. 19

User experience and presence

Potential for VR in broader educational contexts

The students provided feedback and suggestions regarding their VR experience. 88% of the participants in the VR-FCM Group showed a great willingness to use VR educational applications in the future, while the corresponding percentage in the VR-Hint Group was 74% (Fig. 20).

[See PDF for image]

Fig. 20

Willingness to use VR in future

This also reveals the true potential of VR as a learning method especially when it is blended with AI mechanisms. The participants in the VR-Hint Group suggested indicating that personalized assistance as a pedagogic approach will improve their perception of the learning experience by assessing how adaptive feedback support plays its role.

The results of this study indicate that adaptive VR can assist students to learn at their own pace, considering their individual performance and skills to deliver personalized, effective and engaging learning.

Limitations and future research

FCMs used in this study's VR-based educational environment offer adaptive support while considering the skills and performance of the students. However, there are certain limitations to this adaptive approach when attempting to manage complex in-game interactions.

Although learning is facilitated by the proposed VR system, its implementation in real life is not without its challenges. Costly hardware would limit implementation in low-funded organizations (Radianti et al., 2020), and teacher training remains a requirement for successful implementation (Marougkas et al., 2023a, b). Student adjustment problems like motion sickness and cognitive overload would also potentially impact participation (Makransky & Petersen, 2021).

Also, the scalability problem arises as real-time adaptive feedback is computationally expensive. These problems will have to be tackled by cost-effective hardware, professional development among teachers, and optimized VR experiences so that it is applied more extensively in the classroom.

The relationships between nodes become more complex and demanding of the system's real-time flexibility as new ideas are added to old ones. As independent hardware like Meta Quest 2 consumes more processing power, which constrains their limited capability as well as battery life, the capability of independent VR HMD hardware is also an issue.

“Error Rate” as a construct, for instance, would have an impact on “Hint Effectiveness” and “Player Knowledge”. As more constructs are added to the scenarios like the ones with indirect relationships among constructs, however, they probably have difficulty making proper adaptations.

The system's scalability to different study disciplines like STEM education, vocational education, medical education and corporate learning should be the subject of future research. Scaling its application will open the way to test its adaptability and effectiveness across different learning environments. Scaling its application will give more validation of effectiveness and adaptability across different learning environments. Longitudinal studies could also examine the long-term retention of learning and acquisition of skills due to adaptive VR intervention. Another research area with potential is the application of more advanced AI-based approaches, such as reinforcement learning and deep learning models, to facilitate real-time personalization and learning performance. Dynamic learning environments with collaborative VR, through which learners engage with other learners, could also facilitate social learning and problem-solving abilities, opening the way for socially adaptive learning models.

Conclusion

Virtual Reality (VR) is already being used as an effective way of delivering educational material, providing an immersive and interactive way of learning. The majority of educational VR systems do not offer students’ experiences that are tailored to their educational requirements and do not integrate learning theories, which are critical for effective learning outcomes. By using technologies powered by AI mechanisms such as Fuzzy Cognitive Maps (FCMs), these gaps are filled with allowing real-time adaptation and personalized feedback based on the needs of each learner.

This study presents a system which integrates FCMs with Flow Theory and gamification to develop a virtual reality learning environment that dynamically adapts to the challenges by providing learners with adaptive assistance in the form of hints that support them throughout the educational process and provide an efficient learning approach.

The VR-FCM group outperformed the VR-Hint group in terms of performance, engagement, and comprehension thanks to these adaptive mechanisms. In addition to enhancing students'comprehension of Java programming concepts, personalized feedback provided by an AI-supported system additionally enhanced their motivation and sense of control in the virtual environment.

This study brings new insights into the manner in which adaptivity driven by AI enhances virtual learning, paving the way to more effective learning. Through the integration of Flow Theory, Fuzzy Cognitive Maps (FCMs), and Gamification, the study presents an adaptive VR system that adjusts automatically to learners'different competencies to achieve an optimal level of challenge and engagement. Unlike traditional e-learning environments, which rely on static instructional design, our approach leverages real-time AI adaptation to personalize learning pathways. This educational approach advances VR-based educational research and lays the groundwork for future innovations in scalable, immersive, and intelligent adaptive learning systems.

The findings of this study yield several practical implications for educators, instructional designers, and researchers. First, educators aiming to integrate VR in programming education are encouraged to adopt adaptive mechanisms such as Fuzzy Cognitive Maps (FCMs) to support differentiated learning pathways. Personalized feedback, as implemented in this study, was shown to improve learners’ understanding, engagement, and motivation—key factors for successful learning in immersive environments.

For developers of VR/AI-based learning environments, our results emphasize the importance of real-time adaptivity and responsiveness. Systems that respond to learners’ behaviors—such as time on task, error frequency, or prior performance—can offer more meaningful support and promote deeper learning experiences. The successful use of FCMs in our design demonstrates the potential of combining pedagogical modeling with AI to enhance user interaction and feedback quality.

Acknowledgements

Not applicable.

Author contributions

Andreas Marougkas conceived the presented idea and was responsible for the original draft, validation, methodology, formal analysis, and data curation. Christos Troussas contributed to writing—both review and original draft—and was responsible for visualization, validation, methodology, formal analysis, and conceptualization. Akrivi Krouska contributed to writing—review and editing—and provided validation, supervision, methodology, and formal analysis. Cleo Sgouropoulou also contributed to writing—review and editing—and was responsible for validation, supervision, methodology, and formal analysis. All authors discussed the results and contributed to the final manuscript.

Funding

None.

Availability of data and materials

Data will be made available on request.

Declarations

Competing interests

The authors declare that they have no competing of interest.

Abbreviations

VR

Virtual reality

FCMs

Fuzzy cognitive maps

HMD

Head-mounted device

VLEs

Virtual learning environments

AI

Artificial intelligence

NPC

Non-player character

UEQ

User experience questionnaire

MSLQ

Motivated strategies for learning questionnaire

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

Abuhammad, A; Falah, J; Alfalah, SFM; Abu-Tarboush, M; Tarawneh, RT; Drikakis, D; Charissis, V. MedChemVR: A virtual reality game to enhance medicinal chemistry education. Multimodal Technologies and Interaction; 2021; 5, 3 10. [DOI: https://dx.doi.org/10.3390/mti5030010]

Abuhassna, H. From theory to practice: The impact of e-learning on student performance, trends, and educational horizons. Qeios; 2024; [DOI: https://dx.doi.org/10.32388/XKZC3Z.2]

Abuhassna, H; Awae, F; Bayoumi, K; Al Zitawi, DUD; Alsharif, AH; Yahaya, N. Understanding online learning readiness among university students: A bibliometric analysis. International Journal of Interactive Mobile Technologies (iJIM); 2022; 16, 13 pp. 81-94. [DOI: https://dx.doi.org/10.3991/ijim.v16i13.30605]

Abuhassna, H; Van, NT; Yahaya, N; Zakaria, MAZM; Awae, F; Al Zitawi, DUD; Bayoumi, K. Strategies for successful blended learning—A bibliometric analysis and reviews. International Journal of Interactive Mobile Technologies (iJIM); 2022; 16, 13 pp. 66-80. [DOI: https://dx.doi.org/10.3991/ijim.v16i13.30739]

Abuhassna, H; Yahaya, N; Zakaria, MAZM; Zaid, NM; Samah, NA; Awae, F; Nee, CK; Alsharif, AH. Trends on using the Technology Acceptance Model (TAM) for online learning: A bibliometric and content analysis. International Journal of Information and Education Technology; 2023; 13, 1 pp. 131-142. [DOI: https://dx.doi.org/10.18178/ijiet.2023.13.1.142]

Adnan, A; Ya Shak, M; Abd Karim, R; Mohd Tahir, MH; Mohamad Shah, D. 360-degree videos, VR experiences and the application of Education 4.0 technologies in Malaysia for exposure and immersion. Advances in Science, Technology and Engineering Systems Journal; 2020; 5, 1 pp. 373-381. [DOI: https://dx.doi.org/10.25046/aj050148]

Akman, E; Çakır, R. Pupils’ opinions on an educational virtual reality game regarding flow experience. International Journal of Emerging Technologies in Learning (iJET); 2019; 14, pp. 121-137. [DOI: https://dx.doi.org/10.3991/ijet.v14i15.10576]

Bendeck Soto, J; Toro Ocampo, D; Colon, L; Oropesa, A. Perceptions of ImmerseMe virtual reality platform to improve english communicative skills in higher education. International Journal of Interactive Mobile Technologies (iJIM); 2020; 14, 7 pp. 4-19. [DOI: https://dx.doi.org/10.3991/ijim.v14i07.12181]

Bennie, S; Ranaghan, K; Deeks, H; Goldsmith, H; O'Connor, M; Mulholland, A; Glowacki, D. Teaching enzyme catalysis using interactive molecular dynamics in virtual reality. Journal of Chemical Education; 2019; 96, 11 pp. 2488-2496. [DOI: https://dx.doi.org/10.1021/acs.jchemed.9b00181]

Bowman, DA; McMahan, RP. Virtual reality: How much immersion is enough?. Computer; 2007; 40, 7 pp. 36-43. [DOI: https://dx.doi.org/10.1109/MC.2007.257]

Buckingham, G. Hand tracking for immersive virtual reality: opportunities and challenges. Front Virtual Real; 2021; 2, 728461. [DOI: https://dx.doi.org/10.3389/frvir.2021.728461]

Cai, Y; Miao, C; Tan, A-H; Shen, Z; Li, B. Creating an immersive game world with evolutionary fuzzy cognitive maps. IEEE Comput Graph Appl; 2010; 30, 2 pp. 58-70. [DOI: https://dx.doi.org/10.1109/MCG.2009.80]

Chang, SC; Hsu, T; Chen, YN; Morris, J. The effects of spherical video-based virtual reality implementation on students’ natural science learning effectiveness. Interactive Learning Environments; 2018; 52, 3 226. [DOI: https://dx.doi.org/10.1080/10494820.2018.1548490]

Chen, B; Wang, Y; Wang, L. The effects of virtual reality-assisted language learning: A meta-analysis. Sustainability; 2022; 14, 6 3147. [DOI: https://dx.doi.org/10.3390/su14063147]

Chen, CM; Lee, HM; Chen, YH. Personalized e-learning system using item response theory. Computers & Education; 2005; 44, 3 pp. 237-255. [DOI: https://dx.doi.org/10.1016/j.compedu.2004.01.006]

Chen, PH. The design of applying gamification in an immersive virtual reality virtual laboratory for powder-bed binder jetting 3DP training. Education Sciences; 2020; 10, 7 172. [DOI: https://dx.doi.org/10.3390/educsci10070172]

Chiossi, F; Ou, C; Gerhardt, C; Putze, F; Mayer, S. Designing and evaluating an adaptive virtual reality system using EEG frequencies to balance internal and external attention states. International Journal of Human-Computer Studies; 2025; 196, 103433. [DOI: https://dx.doi.org/10.1016/j.ijhcs.2024.103433]

Csikszentmihalyi, M. (1997). Finding flow: The psychology of engagement with everyday life. Basic Books.

Csikszentmihalyi, M. (1997). Flow and the psychology of discovery and invention. HarperPerennial.

Esposito, D., Schaumann, D., Camarda, D., spsampsps Kalay, Y. E. (2020). A multi-agent simulator for infection spread in a healthcare environment. In Advances in Practical Applications of Agents, Multi-Agent Systems, and Trustworthiness. The PAAMS Collection (pp. 408–411). Springer. https://doi.org/10.1007/978-3-030-49778-1_36

Felix, G; Nápoles, G; Falcon, R et al. A review on methods and software for fuzzy cognitive maps. Artificial Intelligence Review; 2019; 52, pp. 1707-1737. [DOI: https://dx.doi.org/10.1007/s10462-017-9575-1]

Godovykh, M; Baker, C; Fyall, A. VR in tourism: A new call for virtual tourism experience amid and after the COVID-19 pandemic. Tourism and Hospitality; 2022; 3, 1 pp. 265-275. [DOI: https://dx.doi.org/10.3390/tourhosp3010018]

Hamari, J., Koivisto, J., & Sarsa, H. (2014). Does gamification work? A literature review of empirical studies on gamification. 2014 47th Hawaii International Conference on System Sciences, Waikoloa, HI, USA, 3025–3034. https://doi.org/10.1109/HICSS.2014.377

Hein, G. E. (1991). Constructivist learning theory. Institute for Inquiry. Available at: http://www.exploratorium.edu/ifi/resources/constructivistlearning.html

Hossain, S; Brooks, L. Fuzzy cognitive map modelling educational software adoption. Comput Educ; 2008; 51, 4 pp. 1569-1588. [DOI: https://dx.doi.org/10.1016/j.compedu.2008.03.002]

Hutson, J; Olsen, T. Digital humanities and virtual reality: A review of theories and best practices for art history. International Journal of Technology in Education (IJTE); 2021; 4, 3 pp. 491-500. [DOI: https://dx.doi.org/10.46328/ijte.150]

Innocenti, E; Geronazzo, M; Vescovi, D; Nordahl, R; Serafin, S; Ludovico, L; Avanzini, F. Mobile virtual reality for musical genre learning in primary education. Computers & Education; 2019; 139, pp. 102-117. [DOI: https://dx.doi.org/10.1016/j.compedu.2019.04.010]

Isik, B., Emir Isik, G., spsampsps Zilka, M. (2024). A Unity3D-based interactive educational game of compressed air system maintenance. In Concli, F., Maccioni, L., Vidoni, R., spsampsps Matt, D. T. (Eds.), Latest Advancements in Mechanical Engineering. ISIEA 2024. Lecture Notes in Networks and Systems, vol 1124. Springer, Cham. https://doi.org/10.1007/978-3-031-70462-8_27

Johnson-Glenberg, M., Ly, V., Su, M., Zavala, R., Bartolomeo, H., & Kalina, E. (2020). Embodied agentic STEM education: Effects of 3D VR compared to 2D PC. In Proceedings of the 6th International Conference of the Immersive Learning Research Network (iLRN 2020), 24–30. https://doi.org/10.23919/iLRN47897.2020.9155155

Kersting, M; Bondell, J; Steier, R; Myers, M. Virtual reality in astronomy education: Reflecting on design principles through a dialogue between researchers and practitioners. International Journal of Science Education, Part B; 2023; 14, 2 pp. 157-176. [DOI: https://dx.doi.org/10.1080/21548455.2023.2238871]

Khosravi, H; Buckingham Shum, S; Chen, G; Conati, C; Tsai, Y-S; Kay, J; Knight, S; Martinez-Maldonado, R; Sadiq, S; Gašević, D. Explainable artificial intelligence in education. Computers and Education: Artificial Intelligence; 2022; 3, 100074. [DOI: https://dx.doi.org/10.1016/j.caeai.2022.100074]

Klochko, O. V., Tkachenko, S. V., Babiichuk, I. M., Fedorets, V. M., spsampsps Galych, T. V. (2022). Using virtual reality technologies for teaching computer science at secondary school. In Ignatenko, O., et al. (Eds.), ICTERI 2021 Workshops. ICTERI 2021. Communications in Computer and Information Science, vol 1635. Springer, Cham. https://doi.org/10.1007/978-3-031-14841-5_15

Korniejenko, K; Kontny, B. The usage of virtual and augmented reality in underwater archeology. Applied Sciences; 2024; 14, 18 8188. [DOI: https://dx.doi.org/10.3390/app14188188]

Kosko, B. Fuzzy cognitive maps. International Journal of Man-Machine Studies; 1986; 24, 1 pp. 65-75. [DOI: https://dx.doi.org/10.1016/S0020-7373(86)80040-2]

Kot, T; Novák, P. Application of virtual reality in teleoperation of the military mobile robotic system TAROS. Int J Adv Robot Syst; 2018; [DOI: https://dx.doi.org/10.1177/1729881417751545]

Krouska, A; Troussas, C; Sgouropoulou, C. Mobile game-based learning as a solution in COVID-19 era: Modeling the pedagogical affordance and student interactions. Education and Information Technologies; 2022; 27, pp. 229-241. [DOI: https://dx.doi.org/10.1007/s10639-021-10672-3]

Lee, JJ; Hammer, J. Gamification in education: What, how, why bother?. Academic Exchange Quarterly; 2011; 15, 2 pp. 146-151.

Liao, T; She, J. How does virtual reality (VR) facilitate design? A review of VR usage in early-stage engineering design. Proceedings of the Design Society; 2023; 3, pp. 2115-2124. [DOI: https://dx.doi.org/10.1017/pds.2023.212]

Lin, Y; Wang, S; Lan, Y. The study of virtual reality adaptive learning method based on learning style model. Computer Applications in Engineering Education; 2022; 30, pp. 396-414. [DOI: https://dx.doi.org/10.1002/cae.22462]

Liu, Y., Lin, Y., Shi, R., Luo, Y., & Liang, H. N. (2021). RelicVR: A virtual reality game for active exploration of archaeological relics. Extended Abstracts of the 2021 Annual Symposium on Computer Human Interaction in Play, 326–332. https://doi.org/10.48550/arXiv.2109.14185

Ma, W; Adesope, OO; Nesbit, JC; Liu, Q. Intelligent tutoring systems and learning outcomes: A meta-analysis. Journal of Educational Psychology; 2014; 106, 4 pp. 901-918. [DOI: https://dx.doi.org/10.1037/a0037123]

Maddalon, L; Minissi, ME; Parsons, T; Hervas, A; Alcaniz, M. Exploring adaptive virtual reality systems used in interventions for children with autism spectrum disorder: Systematic review. Journal of Medical Internet Research; 2024; 26, [DOI: https://dx.doi.org/10.2196/57093] e57093.

Makransky, G; Petersen, GB. The cognitive affective model of immersive learning (CAMIL): A theoretical research-based model of learning in immersive virtual reality. Educational Psychology Review; 2021; 33, pp. 937-958. [DOI: https://dx.doi.org/10.1007/s10648-020-09586-2]

Mariscal, G; Jiménez-García, E; Vivas, MD; Redondo, S; Moreno-Pérez, S. Virtual reality simulation-based learning. Education in the Knowledge Society (EKS); 2020; 21, 15. [DOI: https://dx.doi.org/10.14201/eks.20809]

Marougkas, A; Troussas, C; Krouska, A; Sgouropoulou, C. Virtual reality in education: A review of learning theories, approaches and methodologies for the last decade. Electronics; 2023; 12, 13 2832. [DOI: https://dx.doi.org/10.3390/electronics12132832]

Marougkas, A., Troussas, C., Krouska, A., spsampsps Sgouropoulou, C. (2023). Virtual reality in education: Reviewing different technological approaches and their implementations. In Krouska, A., Troussas, C., spsampsps Caro, J. (Eds.), Novel spsampsps Intelligent Digital Systems: Proceedings of the 2nd International Conference (NiDS 2022). Lecture Notes in Networks and Systems, vol 556. Springer, Cham. https://doi.org/10.1007/978-3-031-17601-2_8

Marougkas, A; Troussas, C; Krouska, A; Sgouropoulou, C. How personalized and effective is immersive virtual reality in education? A systematic literature review for the last decade. Multimedia Tools and Applications; 2024; 83, pp. 18185-18233. [DOI: https://dx.doi.org/10.1007/s11042-023-15986-7]

Memarsadeghi, N; Varshney, A. Virtual and augmented reality applications in science and engineering. Computing in Science & Engineering; 2020; 22, 3 pp. 4-6. [DOI: https://dx.doi.org/10.1109/MCSE.2020.2987151]

Monita, F; Ikhsan, J. Development Virtual Reality IPA (VR-IPA) learning media for science learning. Journal of Physics: Conference Series; 2020; 1440, 012103. [DOI: https://dx.doi.org/10.1088/1742-6596/1440/1/012103]

Onderdijk, KE; Bouckaert, L; Van Dyck, E. Concert experiences in virtual reality environments. Virtual Reality; 2023; 27, pp. 2383-2396. [DOI: https://dx.doi.org/10.1007/s10055-023-00814-y]

Oyekunle, D; Claude, BEA; Waliu, AO; Adekunle, TS; Matthew, UO. Cloud-based adaptive learning system: Virtual reality and augmented reality-assisted educational pedagogy development on clinical simulation. Journal of Digital Health; 2024; 3, 1 pp. 49-62. [DOI: https://dx.doi.org/10.55976/jdh.32024126849-62]

Oyelere, SS; Bouali, N; Kaliisa, R. Exploring the trends of educational virtual reality games: A systematic review of empirical studies. Smart Learning Environments; 2020; 7, 31. [DOI: https://dx.doi.org/10.1186/s40561-020-00142-7]

Papageorgiou, EI; Groumpos, PP. A weight adaptation method for fuzzy cognitive map learning. Soft Computing; 2005; 9, pp. 846-857. [DOI: https://dx.doi.org/10.1007/s00500-004-0426-z]

Papakostas, C; Troussas, C; Krouska, A; Sgouropoulou, C. Measuring user experience, usability and interactivity of a personalized mobile augmented reality training system. Sensors; 2021; 21, 11 3888. [DOI: https://dx.doi.org/10.3390/s21113888]

Papakostas, C; Troussas, C; Krouska, A; Sgouropoulou, C. Personalization of the learning path within an augmented reality spatial ability training application based on fuzzy weights. Sensors; 2022; 22, 18 7059. [DOI: https://dx.doi.org/10.3390/s22187059]

Parmar, D; Bertrand, J; Babu, S; Chalil Madathil, K; Zelaya, M; Wang, T; Wagner, J; Gramopadhye, A; Frady, K. A comparative evaluation of viewing metaphors on psychophysical skills education in an interactive virtual environment. Virtual Reality; 2016; 20, pp. 141-157. [DOI: https://dx.doi.org/10.1007/s10055-016-0287-7]

Pérez-Muñoz, S; Castaño Calle, R; Morales Campo, PT; Rodríguez-Cayetano, A. A systematic review of the use and effect of virtual reality, augmented reality and mixed reality in physical education. Information; 2024; 15, 9 582. [DOI: https://dx.doi.org/10.3390/info15090582]

Pintrich, PR; Smith, DAF; Garcia, T; Mckeachie, WJ. Reliability and predictive validity of the motivated strategies for learning questionnaire (MSLQ). Educational and Psychological Measurement; 1993; 53, 3 pp. 801-813. [DOI: https://dx.doi.org/10.1177/0013164493053003024]

Qi, P., Jumaat, N. F. B., Abuhassna, H., spsampsps Ting, L. (2024). A comprehensive bibliometric analysis of blended learning in language education. In Edwards, B. I., Tanko, B. L., Klufallah, M., Abuhassna, H., spsampsps Chinedu, C. C. (Eds.), Reimagining Transformative Educational Spaces. Lecture Notes in Educational Technology. Springer, Singapore. https://doi.org/10.1007/978-981-97-8752-4_11

Radianti, J; Majchrzak, TA; Fromm, J; Wohlgenannt, I. A systematic review of immersive virtual reality applications for higher education: Design elements, lessons learned, and research agenda. Computers & Education; 2020; 147, 103778. [DOI: https://dx.doi.org/10.1016/j.compedu.2019.103778]

Remolar, I; Rebollo, C; Fernández-Moyano, JA. Learning history using virtual and augmented reality. Computers; 2021; 10, 11 146. [DOI: https://dx.doi.org/10.3390/computers10110146]

Rokooei, S; Shojaei, A; Alvanchi, A; Azad, R; Didehvar, N. Virtual reality application for construction safety training. Safety Science; 2022; 157, 105925. [DOI: https://dx.doi.org/10.1016/j.ssci.2022.105925]

Rychkova, A; Korotkikh, A; Mironov, A; Smolin, A; Maksimenko, N; Kurushkin, M. Orbital Battleship: A multiplayer guessing game in immersive virtual reality. Journal of Chemical Education; 2020; 97, 11 pp. 4184-4188. [DOI: https://dx.doi.org/10.1021/acs.jchemed.0c00866]

Sánchez Laws, AL; Utne, T. Ethics guidelines for immersive journalism. Front Robot AI; 2019; 6, 28. [DOI: https://dx.doi.org/10.3389/frobt.2019.00028]

Scavarelli, A; Arya, A; Teather, RJ. Virtual reality and augmented reality in social learning spaces: A literature review. Virtual Reality; 2021; 25, pp. 257-277. [DOI: https://dx.doi.org/10.1007/s10055-020-00444-8]

Schrepp, M; Hinderks, A; Thomaschewski, J. Construction of a benchmark for the user experience questionnaire (UEQ). International Journal of Interactive Multimedia and Artificial Intelligence; 2017; 4, 4 pp. 40-44. [DOI: https://dx.doi.org/10.9781/ijimai.2017.445]

Segura, RJ; del Pino, FJ; Ogáyar, CJ; Rueda, AJ. VR-OCKS: a virtual reality game for learning the basic concepts of programming. Comput Appl Eng Educ; 2020; 28, pp. 31-41. [DOI: https://dx.doi.org/10.1002/cae.22172]

Srimadhaven, T; Junni, CAV; Harshith, N; Ebenezer, JS; Girish, SS; Priyaadharshini, M. Learning analytics: Virtual reality for programming course in higher education. Procedia Computer Science; 2020; 172, pp. 433-437. [DOI: https://dx.doi.org/10.1016/j.procs.2020.05.095]

Stauffert, J; Niebling, F; Latoschik, ME. Latency and cybersickness: Impact, causes, and measures. A Review. Frontiers in Virtual Reality; 2020; 1, 582204. [DOI: https://dx.doi.org/10.3389/frvir.2020.582204]

Stefan, H; Mortimer, M; Horan, B. Evaluating the effectiveness of virtual reality for safety-relevant training: A systematic review. Virtual Reality; 2023; 27, pp. 2839-2869. [DOI: https://dx.doi.org/10.1007/s10055-023-00843-7]

Strousopoulos, P., Troussas, C., Krouska, A., & Sgouropoulou, C. (2024). The gamification advantage: Innovating education through engagement. In 2024 9th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM) (pp. 204–208). Athens, Greece. https://doi.org/10.1109/SEEDA-CECNSM63478.2024.00045

Strousopoulos, C., Troussas, C., Krouska, A., Mylonas, P., & Sgouropoulou, C. (2023). Integrating stereotype user models for adaptive scenarios in game playing within immersive virtual environments. In 2023 International Conference on Semantic and Social Media Adaptation and Personalization (SMAP)* (pp. 1–6). IEEE. https://doi.org/10.1109/SMAP59435.2023.10255160

Surer, E; Erkayaoğlu, M; Öztürk, ZN et al. Developing a scenario-based video game generation framework for computer and virtual reality environments: A comparative usability study. Journal of Multimodal User Interaction; 2020; 15, pp. 393-411. [DOI: https://dx.doi.org/10.1007/s12193-020-00348-6]

Troussas, C; Krouska, A; Sgouropoulou, C. Collaboration and fuzzy-modeled personalization for mobile game-based learning in higher education. Computers & Education; 2020; 144, 103698. [DOI: https://dx.doi.org/10.1016/j.compedu.2019.103698]

Wang, J., Shi, R., Xiao, Z., Qin, X., & Liang, H. (2022). Effect of render resolution on gameplay experience, performance, and simulator sickness in virtual reality games. ArXiv.https://arxiv.org/abs/2203.12294

Witmer, BG; Singer, MJ. Measuring presence in virtual environments: A presence questionnaire. Presence: Teleoperators and Virtual Environments; 1998; 7, 3 pp. 225-240. [DOI: https://dx.doi.org/10.1162/105474698565686]

Wu, T; Ho, CB. A scoping review of metaverse in emergency medicine. Australasian Emergency Care; 2023; 26, 1 pp. 75-83. [DOI: https://dx.doi.org/10.1016/j.auec.2022.08.002]

Zawacki-Richter, O; Marín, VI; Bond, M et al. Systematic review of research on artificial intelligence applications in higher education—where are the educators?. International Journal of Educational Technology in Higher Education; 2019; 16, 39. [DOI: https://dx.doi.org/10.1186/s41239-019-0171-0]