A Rapid Review of Training Future Behavior Analysts Through Computer-Generated Simulations: Challenges and Opportunities in Using Artificial Intelligence and Virtual Reality

Abstract

This rapid review explores the potential of computer-generated simulations in training future behavior analysts. While competencies in behavior analysis are critical for improving client outcomes, gaps remain in the availability and fidelity of training methods within higher education. Emerging technologies, such as artificial intelligence (AI) and virtual simulations, offer innovative solutions to address these challenges. This review synthesizes literature from 2019 to 2024, examining the role of computer-generated simulations in teaching essential skills at the undergraduate and graduate levels. The findings summarize existing research, identify key variables influencing training effectiveness, and highlight implications for future educational practices. By leveraging AI and virtual reality (VR), educators may enhance experiential learning, promote culturally responsive care, and improve skill acquisition for aspiring behavior analysts. This review provides recommendations to guide future research and the ethical integration of AI technologies in behavior analytic education.

There has been a significant growth in demand for behavior analysts in the United States, with a 5852% increase from 2010 to 2021 and 38% from 2020 to 2021 (Du & Lee, 2024). With the enhanced need comes increased pressure to prepare well-rounded future behavior analysts with a repertoire of behavior-analytic skills as well as skills to think on their feet, solve problems, adapt to change, and be compassionate professionals who can make a positive change in people’s lives. It is imperative that behavioral interventions are implemented with a high degree of quality by well-trained professionals (Leaf et al., 2023). Du and Lee (2024) discussed the importance of training programs focusing on developing a behavior analyst’s expertise in verbal behavior about the science, contingency-shaped repertoires, and verbally mediated repertoires. There has also been criticism that current educational programs lack research-based evidence of their effectiveness on actual behavior change (Kazemi et al., 2022). Shortfalls have been noted within interprofessional collaboration, conflict resolution, in-the-moment treatment decisions, and compassionate and culturally responsive care (Busch et al., 2020; Du & Lee, 2024; Goldman et al., 2018; Henderson et al., 2023; Jimenez‐Gomez & Beaulieu, 2022; Kazemi et al., 2022; LaFrance et al., 2019; Leaf et al., 2023; Shea et al., 2024). Researchers have expressed the need for future behavior analysts to include training in self-evaluation, cultural responsiveness, and relevant cultural variables and how they impact values and goal selection, as well as collaborative, compassionate, and client-centered care (Baires et al., 2023; Beaulieu & Jimenez-Gomez, 2022; Fong et al., 2017; Jimenez‐Gomez & Beaulieu, 2022; LeBlanc et al., 2020; Martinez & Mahoney, 2022; Marchese & Weiss, 2023; Nohelty et al., 2024; O’Neill et al., 2023; Penney et al., 2023; Plattner & Anderson, 2023; Rodriguez et al., 2023; Slim & Reuter-Yuill, 2021; Taylor et al., 2019; West et al., 2024). Overall, effective training programs for behavior analysts are essential to ensure that individuals receive high-quality, ethical, and compassionate care that leads to meaningful improvements in their lives.

Educational institutions play a critical role in providing academic instructions, shaping attitudes, and developing skills in future behavior analysts. Some researchers have explored the field of behavior analysis to develop effective teaching strategies (Du & Lee, 2024; Leaf et al., 2023). One such strategy, behavioral skills training (BST), has been extensively researched (Miltenberger et al., 1999; Slane & Lieberman-Betz, 2021; Smith et al., 2024). Research has indicated that BST has a statistically significant positive effect (Fingerhut & Moeyaert, 2022), yet a scoping review found that BST was often not used or used with poor fidelity (Briggs et al., 2024). Briggs et al. (2024) found that time constraints and lack of resources were significant barriers that prevented the efficacious adoption of this evidence-based training. Within undergraduate and graduate-level education, these barriers are present due to the size of the classes and the diversity of necessary scenarios. This challenge may be bridged by providing additional computer-generated simulations within the BST model when training future behavior analysts.

Computer-Generated Simulations

Some computer-generated simulations utilize advanced technologies such as large language models (LLMs) and machine learning algorithms to create interactive scenarios that closely mimic real-world interactions. Machine learning is a subset of artificial intelligence (AI) that enables systems to learn patterns from data and make predictions or decisions without being explicitly programmed (Devlin et al., 2019). LLMs are trained in vast amounts of text data, enabling the models to understand and generate text in a human-like fashion (Devlin et al., 2019). Conversational AI is made possible by the generative pretrained transformer (GPT) architecture that powers LLMs like the popular ChatGPT by OpenAI (Vaswani et al., 2017). With conversational AI, participants in simulations can engage in discussions with virtual patients and doctors whose responses are generated dynamically in real time. The characters can be customized to reflect any characteristics desired for the simulation, making it possible to have a diversity of personalities that would be financially and logistically difficult, if not impossible, to recreate through in-person actors. The cost savings, convenience, safety, and efficiency of AI-powered simulations are evidenced by their adoption across a wide spectrum of applications, including education, healthcare, and industry training.

Within higher education, there is a huge potential for varied applications. AI in healthcare can simulate surgical procedures or clinical decision-making. Within emergency services, simulations can provide opportunities to address emergency situations, such as fire or police foundations or hazard identification in a workplace environment. Within education, simulations can be created to teach future educators to respond to an aggressive child or a child in emotional distress, and within the field of applied behavior analysis (ABA), simulations can be used to teach discrete-trial training, functional behavior assessments, reinforcement strategies, prompting and fading techniques, and data collection and analysis. These simulations provide students in higher education with opportunities to develop practical skills in a risk-free setting and the potential to provide immediate personalized feedback and adaptive responses based on student input, thereby enhancing experiential learning and critical thinking skills (Fingerhut & Moeyaert, 2022; Ng et al., 2023). By allowing students to engage in complex scenarios, such as patient-care tasks in nursing or culturally responsive consultations in behavior analysis, AI simulations could support skill-building in ways that traditional classroom learning cannot (Briggs et al., 2024; Fingerhut & Moeyaert, 2022; Ng et al., 2023). This approach also addresses several logistical challenges in higher education, such as limited access to real-world clinical placements and high instructor-student ratios, by offering scalable, self-guided practice (Babaita et al., 2024). For instance, using AI-driven avatars in counseling simulations enables students to practice culturally sensitive communication, a skill known to improve client outcomes in applied fields (Jimenez‐Gomez & Beaulieu, 2022). AI tools like Body Works, Teachable Machine, and Google’s Teachable Machine platform also make simulation-based learning accessible even to nontechnical learners, aligning well with global educational initiatives to enhance digital literacy and AI competencies across disciplines (Hallmark et al., 2021).

Given the need for innovative and effective methods to train future behavior analysts alongside advancements in computer-generated simulations, it is essential to explore this potential in greater depth. An exploration of the past five years was conducted due to the rapidly changing technology and the importance of providing a summary of the research using the most current AI tools. While the first working model of a neural network was created in the 1950s (Rosenblatt, 1957), and head-mounted VR displays in the 1960s (Sutherland, 1968), the transformer architecture that powers modern AI tools such as Chat GPT was proposed less than ten years ago (Vaswani et al., 2017). In 2019, OpenAI released its second GPT model, which offered expanded abilities to generate text in a conversational style using natural language processing. This advancement ushered in a new era in human-computer interaction and opened many new possibilities for researchers. Focusing on articles from the past five years ensures relevance and alignment with current research and practices. Research evolves rapidly, especially in fields like behavior analysis, where newer studies often build on or refute earlier findings. By concentrating on recent work, we ensured our review reflected the latest theories, methodologies, and conclusions, avoiding reliance on outdated information. Additionally, many academic and professional guidelines recommend prioritizing recent literature to maintain relevance and credibility.

A rapid review was used to objectively analyze the research. A rapid review is an evolving, streamlined form of systematic review aimed at producing evidence quickly and efficiently without compromising the rigor of the process. According to Smela et al. (2023), rapid reviews follow many systematic review principles but simplify or omit certain steps to accelerate evidence synthesis. Rapid reviews employ strategies like narrowing the research scope and limiting database searches (Smela et al., 2023).

The purpose of this paper is to review the current literature by examining research studies from 2019 to 2024 and exploring the use of computer-generated simulations in teaching and training undergraduate and graduate students in the college education sector. This research will summarize the literature on computer-generated simulations, synthesize the findings, identify variables relevant to effective training, and offer suggestions to guide future research in this area.

Method

Protocol and Search Strategy

The PICOS model (population, intervention, comparator, outcome, study design; Cochrane Library (2020) was used as a structured framework for formulating research questions in systematic and rapid reviews. In the context of rapid reviews, the framework is often adapted to balance comprehensiveness and feasibility. A PRISMA-ScR checklist was used to guide this review. The databases PsycINFO and ERIC were used to identify peer-reviewed studies published in journals written in English between January 2019 and September 2024. The search terms and Boolean phrases were as follows: (randomized controlled trials or rct or randomised control trials or qualitative or quantitative) AND (AI generated simulation* OR virtual simulation* or avatar* or avatar based or avatar-based OR virtual reality) AND (undergraduate students or graduate students or university students or college students). In addition, a manual search of the Journal of Applied Behavior Analysis and Behavior Analysis in Practice was conducted for the period January 2019 to July 2024 in an attempt to ensure research in behavior analysis was included.

Inclusion Criteria

Studies were included if they (a) were published in a peer-reviewed journal within the past five years, (b) used a computer-generated simulation, (c) taught a clinical or industry skill of any kind to undergraduate or graduate level students in a college environment, and (d) evaluated the effectiveness of training using some form of empirical data, including either quantitative or self-report of skill change.

Exclusion Criteria

This study excluded meta-analyses, systematic reviews, dissertations, general opinion articles simulations, and research involving therapy for students using computer-generated simulations.

Data Extraction

For each article selected in the review, the following information was extracted: (a) participant characteristics (level of education and discipline); (b) study characteristics (study design); (c) skill taught; (d) training format (how computer-generated simulations were used within the skill acquisition); (e) training outcomes (skill acquisition effectiveness or self-reported effectiveness); and (f) social validity (any measures of student perceptions of training effectiveness).

Results

Across the two databases, 54 studies were identified. Covidence identified four duplications, and one duplication was identified manually. Two researchers conducted an initial abstract screening, resulting in 21 studies meeting inclusion criteria, which moved to a full-text screening. Two researchers reviewed these 21 articles, and 16 were included. There were three disagreements, creating an interrater agreement of 85%. Disagreements were resolved by the primary researcher after rereading the article and reviewing the inclusion and exclusion criteria a second time. The full review is captured in the PRISMA-ScR flow chart (Fig. 1). The PRISMA-ScR checklist is included in Fig. 2. Data extraction was completed by the primary author and reviewed by the secondary author. The secondary author agreed with the data extraction and suggested the inclusion of one other item. The additional item identified by the secondary author was included in the paper.

Fig. 1.

PRISMA-ScR flow chart describing the article selection process

Fig. 2.

PRISMA-ScR checklist

Table 1 describes the participants, discipline, design, and the skills targeted. These studies included 977 college or university students within both graduate and undergraduate programs. Within the 977 subjects, 668 individuals used some form of computer-generated avatars to learn some industry or clinical skill, with the remaining as control subjects. Within these articles, 29 students were identified as college students, 711 students were identified as undergraduate students, 89 students were identified as graduate students, and 148 were identified as university students. A wide range of disciplines were included in these studies, half of which were within the medical field. The research designs included randomized controlled trials, qualitative, quantitative, nonequivalent, pre–post control group, and within-subject research design. The 16 studies taught a variety of skills during simulations. A large percentage addressed medical skills (41%). Soft skills were targeted within 18% of the studies (interprofessional communication, breaking bad news, ethical decision-making). One study targeted interprofessional communication. No studies were directed at compassionate or culturally responsive behaviors.

Table 1.

Overview of research in training using computer-generated simulations

Study	Number of participants	College level	Discipline	Design	Skill taught
Aksoy et al. 2023	14 Exp. 15 Control	College	Anesthesiology	Randomized control	Crisis Management, technical skills
Alahmadi and Muslim Alraddadi, 2020	90	Undergraduate	Preparation year to university	Qualitative	English
Andalib and Monsur, 2024	16	Undergraduate	Landscape architecture	Qualitative	Understanding of landscape architecture construction details
Babaita et al. 2024	27 Exp. 30 Control	Undergraduate	Nursing	Randomized control	Closed tracheal suction (including oral suction) for mechanically ventilated patients.
Carrard et al. 2020	23	Undergraduate	Medical students	Qualitative	Breaking bad news and responding to patient reactions
Chan et al. 2021	38 Exp. 39 Control	Undergraduate	Nursing	Randomized control	Chemotherapy administration, including drug preparation, personal protection, confirmation of the tubing set, patient education, and the safe handling of hazardous drugs.
Chang et al. 2023	23 in person 30 hybrid 23 VR	University students	Not reported	Randomized control	CPR
Hu et al. 2023	37 Exp. 35 Control	University Students	Business	Pre–post-test nonequivalent control group	Ethical decision-making skills
Huang et al. 2024	26 Exp. 30 Control	Undergraduate	Medical students	Randomized control	Spinal injury management, composite injury treatment, CPR, tracheal intubation, bronchoscopic lung segment recognition
Liaw et al. 2020	60 Virtual 60 Simulations	Undergraduate	Medical and nursing students	Randomized control	Interprofessional communication skills
Moro et al. 2021	19 Mobile based 19 HoloLens	Undergraduate	Health and biomedical science	Randomized control	Anatomy of the brain
Nuutinen et al. 2024	23 VR 17 video lectures	Undergraduate	Medical students	Randomized control	Incising an abscess
Patton et al., 2024	69	Graduate Level	Health Professions	Within subject	Substance use disorder screening, brief intervention, and referral.
Randolph et al. 2024	Exp.10 Control 10	Graduate students	Special Education	Randomized control	Differential reinforcement, prompting, restating expectations
Remacle et al. 2023	21 Exp. 20 Control	Undergraduate	Teacher College	Randomized control	Vocal behaviors in noisy classroom, improving communication protecting voice.
Sunday et al. 2022	153	Undergraduate	Computer Science	Qualitative	Object oriented programming

Table 2 summarizes the types of training procedures used within the simulation research, the results of the study, and any information shared on social validity. Among the identified 16 research articles, five research studies (31%) used computer-generated simulations in isolation (Aksoy et al., 2023; Chan et al., 2021; Huang et al., 2024; Nuutinen et al., 2024; Remacle et al., 2023). The remaining articles were blended treatments using AI-generated simulations as one component. Other components that were frequently combined with the simulations were online modules, group debriefing (facilitated discussions following simulations were used to analyze performance and reinforce learning), in-class lectures (traditional lectures were used to introduce concepts, models, and theoretical frameworks that were then applied in simulations), self-observation and reflection (students engaged in self-review, often using video recordings of their simulation sessions, to enhance self-awareness and refine their communication skills), live modeling (instructors or experienced practitioners demonstrated the desired skills and behaviors), live simulations (role-playing exercises that were not AI-driven), hand sketching, class discussions (facilitated group discussions). The computer-generated simulations included practice scenarios where students interacted with avatars to develop targeted skills, computer-generated environments, VR games that focused on specific skills, collaborative missions using avatars, and a virtual platform to allow for collaboration and team-based simulations (Aksoy et al., 2023; Andalib & Monsur, 2024; Hu et al., 2023; Huang et al., 2024). AI was also used as a tool for feedback and evaluation (Aksoy et al., 2023).

Table 2.

Methods and outcomes of research in training using computer-generated simulations

Study	Training procedure	Training outcomes	Social validity
Aksoy et al. 2023	Lecture was followed by skills training with live facilitators, then the control group completed simulation training while the VR group, trained with the VR-based game module	Conventional training vs. virtual reality based Technical scores higher in conventional group. (p = .03)	The majority of the participants prefer VR-based serious game-based training
Alahmadi and Muslim Alraddadi, 2020	Students practiced speaking English within a virtual classroom format	Virtual classes vs. face-to-face classes No difference	Positive attitudes toward virtual classes, format helped overcome learning obstacles, such as anxiety when interacting in English
Andalib and Monsur, 2024	Using computer-generated VR simulations as a co-created, visualized 3D models and explored views of the construction details. This experience was part of a larger curriculum that included traditional lectures, hand sketches, 2D CAD drawings, and 3D modeling	Students affirmed the utility of VR in aiding their understanding​	Some students reported issues such as dizziness and visual discomfort. Increased enjoyment.
Babaita et al. 2024	3-hour theoretical class, Meta Quest VR training vs face to face instructor demonstration	Face to face vs. 360-degree VR No difference (p = 0.874)	Face to face group had higher satisfaction. Reports of “feeling drunk.” VR reported as more interesting
Carrard et al. 2020	Computer-generated virtual patient simulation where students interacted with an avatar in a virtual medical consultation. Experiences videotaped for self-observation	Effective in helping students develop technical communication skills, such as organizing information and becoming more aware of their verbal and non-verbal behaviors	Students appreciated the stress-free environment to practice. Students reported lack of interactivity and realism. Self-observation was valued, allowing for critical self-reflection
Chan et al. 2021	Experimental group used VR-based documents, which included a 3D simulation of chemotherapy administration. Used smartphones to engage in virtual simulations. Control group receiving traditional documents	VR training vs traditional methods VR higher knowledge score (p = .003)	Students commented on the immersive and realistic nature of the simulations lead to effective learning. Some noted technical issues
Chang et al. 2023	All groups received a 40-minute lecture followed by 70 minutes of practice. The VR delivered lecture and practice on VR platform	VR vs hybrid vs in person Intention: hybrid and VR higher (p = .05) Knowledge: hybrid and in person higher (p = .05) Accuracy: in person and hybrid higher (p = .05)	No reported social validity
Hu et al. 2023	Lecture was presented to both groups followed by Avatar based digital collaboration missions on the Gather Town Platform vs the asynchronous learning platform for the control group	Avatar based vs synchronous learning platform Motivation: avatar based higher (p < .05) Ethical decision: avatar higher (p < .05) Academic achievement: avatar higher (p < .05)	Did not directly measure social validity
Huang et al. 2024	Theoretical and practical lessons followed by: Metaverse vs in person simulations	VR Metaverse vs traditional Higher VR (p < .05)	High satisfaction 88; lower level in comfort of use 69
Liaw et al. 2020	Students interacted within a 3D platform CREATIVE in a virtual hospital using avatars to simulate nurse-physician team-based scenarios	VR vs live simulation Communication: no difference (p = .29) Teamwork attitudes: no difference (p = .14)	Students reported virtual reality provided opportunities for anonymous social interactions, which may reduce social anxiety and stress for students
Moro et al. 2021	A short lecture-style audio-stream was used with a 3D model of a brain. The user was able to dissect and remove parts and of the brain	Holo lens vs AR group Knowledge: no difference (p = .69)	The participant in the HoloLens group reported greater amounts of general discomfort, fatigue, headache, eye strain and blurred vision and significantly more dizziness
Nuutinen et al. 2024	In the VR-simulation of the abscess incision, the students wore the VR glasses and received feedback when they touched anything. Within the video group, the video lecture of the abscess lasted two minutes	Virtual reality vs video lecture Self-assessment knowledge: no difference Self-confidence: no difference Learning value: no difference Test scores: no difference	The VR group evaluated the exercise added value to the teaching
Patton et al. 2024	Self-paced online modules. Two virtual simulations (screening brief intervention and referral to treatment of adolescent avatars), Synchronous debriefing	Pre–post-test, no control Drug perception: increase (p < .001) Interprofessional competency (p < .001)	Enjoyed interprofessional teamwork
Randolph et al. 2024	Online module, videos, and readings. Then participants avatar platform Teach Live™ verses participants in zoom with actor groups	Virtual platform TeachLivE™ vs Zoom with actors Accuracy: TeachLivE™ higher mean 76 vs 67	Participants felt the instructional format was an effective way to learn content
Remacle et al. 2023	Three individual voice training sessions including virtual reality (VR) classroom simulations. Simulated noisy environments and pupil agitation to mimic real-life teaching conditions. Participants practiced vocal skills in these immersive settings with increasing levels of difficulty	VR vs control Phonetographic surface: no difference (p = .666) Vocal effort: no difference (p = .702) Self-efficacy: VR higher (p = .009) Immersion tendency: VR higher (p = .01)	Participant satisfaction
Sunday et al. 2022	A brief lecture, questions and discussions, students played Imikode VR game for 30 minutes	Increased understanding of the subject and programming skills. Increased motivation of students to learn material	Students report dizziness. Students reported that the VR helped them in understanding and found it interesting and motivating

Further categorizing findings into key themes allowed for the comparison of computer-generated simulation methods and highlighted areas for future research. Within this synthesis, themes such as overall training outcomes, types of soft skills addressed, social validity results, and a review of skills targeted were explored. This synthesis attempted to relate the types of skills taught to educational requirements for teaching future behavior analysts.

The studies examined different constructs within their results. Thirteen of the 16 studies (81%) measured quantitative competencies or skill changes. Of these, one study (8%) found hands-on training more effective, four studies (31%) showed the VR simulation group outperformed the control group, five studies (38%) found no difference between in-person and VR simulation groups, two studies (15%) showed mixed results, and one study (8%) reported skill increases following baseline with VR simulation without a control group comparison. The remaining three studies (19% of the original 16 articles) did not have quantitative skill measurement competencies. However, students in these studies all self-reported increases in both competencies and confidence. These studies demonstrated positive changes following the VR simulations. Ten articles involved a direct comparison between different educational methods, technologies, or learning environments, as seen in Table 2. Nine studies evaluated soft skills within this synthesis.

Soft skills (interpersonal skills that facilitate effective social and professional interaction) were evaluated through a variety of methods, including self-assessment, questionnaires, and observational methods. Carrard et al. (2020) explored the use of virtual patient (VP) simulations to train medical students in breaking bad news, such as a surgeon reporting that a tumor could not be completely removed or providing a cancer diagnosis. They found that students valued self-observation through video recording as a way to reflect on their communication behaviors. The students identified that the simulation helped them develop skills such as observing their verbal and nonverbal communication, becoming more aware of their own reactions, and managing their emotional responses, pauses, and silences. Liaw et al. (2020) investigated the use of VR for team training in healthcare. They used the Interprofessional Socialization and Valuing Scale (ISVS) to assess changes in attitudes. Hu et al. (2023) explored the use of avatar-based digital platforms for improving ethical decision-making skills and evaluated the decision-making skills by assessing participants’ answers to scenarios. Remacle et al. (2023) evaluated a voice-related prevention program for teachers using VR. They measured oral communication self-efficacy using a 15-item self-rated questionnaire, with participants rating their capability in various situations on a 7-point Likert scale. Patton et al. (2024) used the Interprofessional Education Collaborative Competency (IPEC) Self-Assessment to evaluate interprofessional readiness and competency.

Social validity measures were reported in 14 of the 16 studies (88%), with several common themes emerging across these evaluations. Related to the VR headsets, a primary concern expressed by participants was the experience of nausea, dizziness, or a “feeling of being drunk,” which was reported in four studies (27%). Despite these occasional adverse effects, student responses were overwhelmingly positive, highlighting increased motivation, enjoyment, and reduced anxiety during practice sessions. Notably, participants appreciated the safe environment provided by virtual reality (VR) and simulation-based training, allowing them to refine their skills without posing any risk to actual clients.

Several studies explored the application of VR and simulation technologies for teaching specific clinical and technical skills across various disciplines. While some of these studies targeted fields outside of ABA, they nonetheless provide insight into how VR can be leveraged to teach technical ABA competencies. For example, Chang et al. (2023) examined the effectiveness of hybrid and VR methods for CPR training, while Aksoy et al. (2023) used VR-based serious games to reinforce knowledge of advanced life support (ALS). Additional applications of VR training included its use in a preparatory language program to enhance student engagement in English (Alahmadi & Muslim Alraddadi, 2020), the instruction of landscape architecture design techniques (Andalib & Monsur, 2024), and the teaching of closed tracheal suction procedures through 360° VR video simulations (Babaita et al., 2024).

Several studies were particularly relevant to behavior analysis, demonstrating the potential of VR to teach skills related to addictions, counseling, caregiver communication, and crisis management. For instance, Patton et al. (2024) utilized virtual simulations to train an interprofessional group of 69 graduate students in substance use screening and brief intervention techniques. Additionally, a VR-based serious game was developed to automatically assess crisis response management skills alongside other technical competencies (Aksoy et al., 2023). Another notable study by Remacle et al. (2023) employed VR teaching simulations to support teacher trainees in developing communication skills and self-efficacy, with a focus on improving vocal health in noisy classroom environments. The emphasis on effective communication aligns with core competencies in behavior analysis, where practitioners must engage in clear, precise, and contextually appropriate communication across consultation, training, and intervention settings. The adaptability of VR in simulating real-world conditions, such as challenging classroom environments, suggests its potential for preparing future behavior analysts to navigate diverse professional settings effectively. Another area of relevance is the use of VR-based training to teach ethical decision-making skills.

Hu et al. (2023) examined the use of an avatar-based digital collaboration platform to enhance students’ ethical decision-making, learning motivation, and academic achievement in a business ethics course. The researchers employed a pre–post-test design with nonequivalent control groups, comparing an experimental group using an avatar-based digital platform with a control group utilizing a conventional asynchronous learning platform. The results indicated that the experimental group demonstrated statistically significant improvements in ethical decision-making skills and academic performance compared to the control group. Another related area, interprofessional skills, was examined among medical and nursing students.

Liaw et al. (2020) investigated the effectiveness of computer-based virtual reality (VR) team training in comparison to live simulations for enhancing teamwork attitudes and communication skills among healthcare students. The study involved 120 medical and nursing students who were randomly assigned to either the live simulation or VR-based training groups. Both training modalities simulated real-life scenarios, incorporating multiuser participation, including students, a simulated patient, and a facilitator, to support collaborative learning. The findings indicated that VR-based training was not inferior to live simulations in improving teamwork attitudes and communication skills. Furthermore, the computer-based VR platform, which utilized avatars within a 3D virtual hospital environment, successfully replaced human-controlled facilitators and simulated patients, enhancing the scalability and sustainability of the training.

A particularly relevant study for the training of future behavior analysts is Randolph et al. (2024), which explored the effectiveness of two virtual training platforms, TeachLivE™ (avatar-based) and Zoom (actor-based), in training preservice teachers in behavior management skills. The researchers implemented BST to teach differential reinforcement, prompting, and restating expectations. Graduate students enrolled in special education programs across three universities were divided into two groups, receiving identical training modules and practice scenarios via different platforms. The TeachLivE™ group engaged with lifelike avatars in a virtual classroom, whereas the Zoom group participated in traditional role-play with peer actors. Training effectiveness was assessed using a percent-correct measure of participants’ application of behavior-reduction techniques. The results demonstrated that the TeachLivE™ group exhibited a higher level of success in applying behavior management strategies compared to the Zoom group, suggesting that the immersive nature of TeachLivE™ may offer a more effective training environment. However, both groups demonstrated improvements, reinforcing the potential benefits of virtual practice in skill acquisition. This study underscores the value of virtual simulations in teaching essential behavior management techniques to future behavior analysts.

Discussion

A rapid review of the use of computer-generated simulations to train clinical/industry skills in undergraduate and graduate education programs was conducted. Sixteen articles were analyzed, and information related to:

1. Participant characteristics.

2. Study characteristics.

3. Skills taught.

4. Training outcomes, and

5. Social validity was extracted.

These data yielded valuable information regarding current practices, evidence of effectiveness, and practical implications related to teaching compassionate, culturally responsive consultation and transdisciplinary skills to graduate and undergraduate students.

This rapid review search using ERIC and PsycINFO revealed that research on computer-generated simulations is more prevalent in the healthcare sector than in other fields. However, Lin et al. (2024) reported that the use of AI in education has been steadily increasing over the past decades. According to their findings, the adoption of VR in higher education institutions has risen by 46% in the United States and 96% in the United Kingdom. Additionally, Lin et al. (2024) noted the establishment of dedicated VR laboratories at leading universities, including Harvard. Despite the growing interest and implementation of computer-generated simulations, the rapid pace of technological advancements poses challenges in developing research-based tools grounded in emerging technologies. This underscores the necessity of conducting rapid reviews to keep pace with these developments. Nevertheless, the practical implications of computer-generated simulations in education remain substantial, offering opportunities for enhanced learning experiences, improved skill acquisition, and scalable training solutions.

Ethics of Using AI Simulation Within Pedagogy

The integration of AI simulations into educational settings introduces several ethical considerations that must be carefully addressed to ensure fairness, transparency, and effectiveness in learning environments. One primary concern is data privacy, as AI-driven simulations may collect user interaction data, including personal and performance-related information. Clear policies on data protection and management are essential to safeguard student privacy and prevent unauthorized use of sensitive information (Ng et al., 2023). Additionally, AI models can inherit biases from their training data, potentially leading to inaccurate or inequitable portrayals within simulations, particularly in culturally diverse contexts (Long & Magerko, 2020). To mitigate these risks, AI tools must be designed with transparency, allowing both educators and students to understand their limitations and potential biases.

The ethical considerations for AI-driven training are critical when preparing future behavior analysts, as these professionals will work with vulnerable populations. AI simulations used in this context must prioritize cultural sensitivity, client dignity, and ethical service delivery. Simulations should incorporate diverse avatars and realistic client interactions that reflect the varied backgrounds and needs of service recipients (Jimenez‐Gomez & Beaulieu, 2022). Moreover, AI models should align with core behavior-analytic principles, particularly in fostering compassionate and culturally responsive care. The design of these simulations must avoid reinforcing stereotypes or oversimplified behaviors that could mislead trainees and impact their clinical judgment.

As computer-generated simulations continue to be integrated into behavior analysis education, it is essential to provide students with the knowledge to critically assess their functionality, scope, and limitations. AI cannot fully replicate the nuances of human behavior, and overreliance on these tools may hinder the development of essential interpersonal and clinical skills. Furthermore, there is a risk that cost-saving initiatives may prioritize AI-based training over its integration with comprehensive, instructor-led experiences. To maximize the benefits of AI-driven pedagogy while addressing these challenges, educators may consider implementing AI-enhanced training as part of a balanced curriculum that combines technological advancements with evidence-based instructional methods.

Implications for Teaching Compassionate, Culturally Responsive Consultation and Transdisciplinary Skills

The integration of VR into ABA training offers a promising solution to longstanding gaps in education related to cultural responsiveness and interdisciplinary collaboration. Unlike traditional didactic methods, VR-based simulations provide a dynamic, interactive platform where students can repeatedly engage in real-world scenarios, receive immediate feedback, and refine their skills in a controlled environment. By incorporating avatars representing diverse backgrounds, VR can expose students to culturally nuanced interactions, fostering empathy and potentially removing implicit biases through repeated exposure to counter-stereotypical portrayals (Belisle & Dixon, 2020; Dixon et al., 2009; Kirschner et al., 2022; Shea et al., 2023; Sivaraman et al., 2023). Additionally, VR-based training can simulate interdisciplinary team meetings, allowing students to navigate professional communication and collaboration with practitioners from various fields. Given the documented struggles behavior analysts face in interprofessional settings (Kazemi et al., 2022), these simulations may provide much-needed preparation for managing workplace conflicts and fostering effective teamwork.

Studies suggest that VR enhances skills such as interprofessional communication and perceptions of the importance of interprofessional communication (Carrard et al., 2020; Hu et al., 2023; Liaw et al., 2020; Patton et al., 2024), supporting its potential utility in ABA education. Patton et al. (2024) found that scores in the Problems Perceptions Questionnaire (PPQ) were higher following virtual simulations, which indicated improved trainee’s perception of individuals who use drugs, suggesting a shift toward more positive views of this marginalized group. While the integration of VR into behavior analytic training is still in its early stages, it has the potential to enhance education in this area. By leveraging VR technology, ABA education programs may enhance student competencies in culturally responsive compassionate care and interdisciplinary collaboration, ultimately improving service delivery and client outcomes.

Best Practice Standards

Literature has suggested that research has struggled to keep pace with rapid technological advancements (Lin et al., 2024; Ng et al., 2023; Randolph et al., 2024). Ongoing research and evaluation are crucial to optimizing VR integration and identifying strategies to maximize student engagement and knowledge retention (Briggs et al., 2024). Future standards must evolve to incorporate these ongoing changes to remain relevant. This review provides some reflections on developing best practice standards regarding simulation design, equipment, and simulation integration into the curriculum, evaluation of learning outcomes, and sustainable VR integration in higher education.

Simulation Design

One approach to making LLMs more suitable for a specific application is through fine-tuning. This involves adjusting pretrained LLMs on selected new datasets that are relevant to their use case, allowing the model to generate more appropriate and context-specific outputs (Howard & Ruder, 2018). Fine-tuning can, for instance, help an LLM understand the specific language used in behavior analysis, recognize particular client scenarios, and provide feedback aligned with best practices.

Creating effective simulations that meet the cultural and ethical requirements of the Behavior Analyst Certification Board (BACB) codes may be most effectively completed by using a team of subject specialists. In an era of rapidly advancing technology, it may be tempting to collaborate with simulation specialists to develop modules for courses quickly. However, it would be more effective for diverse teams, including subject matter experts from various cultural backgrounds, to work together in designing and creating materials for global use. These teams should also continuously adapt and monitor the materials to ensure their relevance and effectiveness. Within nursing, the International Nursing Association for Clinical Simulation and Learning (INACSL) Standards Committee has been developed to guide the integration, use, and advancement of simulation-based experiences within research, academia, and clinical practice (Hallmark et al., 2021). A similar special interest group within BACB may act to enlist subject experts to work alongside simulation specialists on topics such as developing technical skills, ethical decision-making skills, supervision skills, culturally responsive and compassionate ABA skills, and effective transdisciplinary skills. This team should include specialists from varied cultures, backgrounds, and areas of expertise to foster diverse perspectives and innovative solutions.

When designing these modules, consideration should be given to creating avatars representing a range of cultures and placing avatars representing marginalized groups in leadership roles. This will enhance the material, resources, global reach, and applicability, ensuring that resources are accessible to a broader audience while promoting diversity, equity, and inclusion.

Equipment

Student feedback within 24% of the articles indicated difficulties with vertigo when using VR devices (Andalib & Monsur, 2024; Babaita et al., 2024; Moro et al., 2021; Sunday et al., 2022). This, along with the higher level of technical skills required to calibrate VR devices, training requirements to use VR devices, potential technical complications associated with VR devices, and the additional costs associated with VR devices, all create considerable deliberations of the equipment requirements of the immersive experience for students. The computer-generated scenarios can be accessed on computers and phones without VR devices. This equipment would not offer the feedback that the VR experiences provide (measuring hand and head movements and eye contact) and would limit the immersive experience. Additional advantages to accessing through individuals’ phones and computers are the lower cost of equipment and the added advantage that students have the flexibility to complete the scenarios at any time from any location. This may increase student participation and potentially increase the number of scenarios practiced. Moro et al. (2021) compared VR to mobile augmented reality in a randomized controlled trial to teach the physiology and anatomy of parts of the brain. The authors found no difference between the test scores of the VR group and those of the mobile-based augmented reality group. However, there was a significant increase in reported dizziness in the VR users (Moro et al., 2021). Given Moro et al.’s research, along with the high incidence of students expressing concerns about dizziness and “feeling drunk” across several research articles, it may be prudent to examine the cost and benefits in terms of DICE.

DICE (dangerous/difficult, impossible/impractical, counterproductive, or expensive) refers to a decision-making process to help guide individuals in considering the use of VR (Learning Guild, 2019). Dangerous/difficult refers to the opportunity VR provides to practice and improve emotional and behavioral responses in the safety of a simulation. Impossible refers to the fact that through VR, learners can experience a situation from multiple perspectives that would be impossible to replicate. Counterproductive refers to the elimination of minor injuries which may be avoided when using simulations. Finally, expensive refers to the potential to create experiences that are too expensive or rare to practice in the real world (Learning Guild, 2019). While one research article indicated equivalent results of using AI without VR, it may be argued that the immersive experience provided by VR may be an important feature when considering the full benefits outlined in DICE. The benefits of immersion may be difficult to evaluate. By mitigating risks, fostering perspective-taking, and reducing costs associated with real-world practice, VR offers a compelling case for its integration into training environments, even when drawbacks are considered. There is a need for future research targeting mitigating strategies targeted at VR-induced vertigo, such as leaning forward, ginger candy, and limiting exposure to VR. Regardless of the form of the AI integration, incorporating computer-generated simulation should also be considered.

Simulation Integration Into the Curriculum

While there has been no component analysis of the efficacy of specific VR techniques, researchers have provided a variety of effective strategies for integrating VR into curricula. Most researchers in this study (69%) used a blended learning model, combining VR technologies with traditional instructional methods. For example, Andialib and Monsur (2024) combined lectures, readings, and hand sketching prior to moving to the VR experience, allowing for a solid theoretical foundation before engaging with more advanced VR technologies. Several researchers provided opportunities for students to practice skills, make errors, and receive constructive feedback prior to the simulations (Randolph et al., 2024; Remacle et al., 2023). Randolph et al. (2024) used a BST model incorporating VR as a strategy to further practice behavioral strategies. Yet another useful strategy was creating a collaborative learning environment to create opportunities for knowledge sharing and sharing of diverse perspectives (Hu et al., 2023; Patton et al., 2024). An alternative use of VR was to create opportunities or real-world simulations. Chan et al. (2021) created a hospital setting to help nursing students handle chemotherapeutic drugs, and Remacle et al. (2023) created a classroom setting to teach education students better oral communication skills in noisy environments. Another key component to consider in pedagogy is the strategy to evaluate outcomes.

Evaluation of Learning Outcomes

Research was varied on the evaluation of the skills. In some cases, computer-generated simulations provided evaluations and in other cases, self-reflections of the skills were used. Finally, professors evaluated the skills separately following the practice with the avatars. It is suggested that further research is necessary to determine the agreement between computer-generated feedback and feedback from professors. Understanding whether these align before relying on AI to evaluate students’ performance is important.

Sustainable VR Integration in Higher Education

A system-wide approach, incorporating a support system including collaboration among faculty, IT professionals, and institutional leadership, is key to sustaining VR adoption in higher education. Universities and colleges should provide ongoing technical support, access to resources, and continuous professional development opportunities to help educators refine their VR teaching strategies over time. Professional development for educators to effectively integrate VR into curricula should include training in both the pedagogical and technical aspects of these tools. Faculty must develop the knowledge and skills required to navigate VR hardware, operate simulation software, and design engaging, interactive learning activities that align with course learning outcomes (Long & Magerko, 2020). Training should also emphasize inclusivity, ensuring that VR applications support students with different abilities, allowing all learners to benefit from immersive, experiential education. By investing in teaching methodologies, support systems, and interdisciplinary collaboration, institutions can create an environment where VR technology meaningfully contributes to student learning, making education more immersive, interactive, and accessible for all.

Future Research

Computer-generated simulations hold promise for teaching clinical skills to undergraduate and graduate-level students, offering an immersive, interactive method for practicing critical competencies in a controlled environment. However, the existing body of research on this pedagogy remains limited, with few studies directly comparing its efficacy to traditional methods of instruction. Preliminary evidence indicates that simulations can improve learning by providing opportunities for repetition, immediate feedback, and a risk-free environment for making mistakes. However, further rigorous research is required to validate these findings (Cook et al., 2013). Additionally, there is a lack of research specifically within ABA. There is also a lack of standardized best practice guidelines for integrating AI-driven simulations into educational curricula. Establishing such standards would ensure consistency in training outcomes and guide educators on the appropriate balance between AI tools and hands-on, real-world experiences. Further research is required to determine the effectiveness of computer-generated assessments and determine if these can effectively be used to evaluate students.

Limitations

The scope of the review was limited to the ERIC and PsycINFO databases and two journal searches. These searches may not adequately cover or include publications and resources, particularly in STEM (Science, Technology, Engineering, and Mathematics). A more comprehensive search may generate broader disciplines and subjects. This review was presented from a behavior-analytic perspective, and other consumers may have different perspectives related to computer-generated simulations.

Although the guidelines of PRISMA-ScR were used, steps were omitted. Namely, the search was not registered. There was not a critical appraisal of the sources of evidence. Also, there were only two evaluators of the sources of evidence.

Conclusions

This rapid review examined the application of computer-generated simulations in training undergraduate and graduate students in industry and clinical skills, with a particular focus on behavior analysis education. One of the primary purposes of this review was to identify key variables influencing training effectiveness in computer-generated simulations. The findings suggest that computer-generated simulations, including avatar-based and virtual reality platforms, offer promising opportunities for skill acquisition by providing scalable, immersive, and interactive learning environments. These technologies address common training challenges such as limited access to real-world placements, resource constraints, and the need for repeated practice in controlled settings. The review also highlights key considerations for future implementation. AI-driven training tools must be designed to ensure cultural responsiveness, ethical integrity, and alignment with evidence-based practices. While computer-generated feedback and virtual practice opportunities enhance training efficiency, research is needed to validate their effectiveness compared to traditional instructional methods. Additionally, concerns related to accessibility, technological limitations, and student discomfort with VR applications warrant further exploration.

Given the increasing reliance on AI technologies in education, it is imperative that best practice standards be developed to guide the integration of computer-generated simulations into behavior analytic training. Future research should focus on refining these tools to enhance their educational impact, evaluating their efficacy in promoting compassionate and culturally responsive consultation skills and establishing guidelines to ensure their ethical use. By leveraging the advantages of AI-driven training while addressing its limitations, educators can enhance the preparation of future behavior analysts to meet the complex and evolving needs of diverse client populations.

Author Contributions

Pamela Shea contributed to the study conception and design. Material preparation, and data collection were completed by Pamela Shea and the analysis was completed by Pamela Shea and Erin Cote. The first draft of the manuscript was written by Pamela Shea with sections of AI by Michael Carter. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

No funding was received to complete this study.

Availability of data and material

Not applicable.

Declarations

Conflicts of Interest

There are no conflicts of interest to declare.

Ethics approval

Not applicable.