Reviewed by David Celiberti, PhD, BCBA-D and Prathik Sasikumar, Extern
Association for Science in Autism Treatment

practice parameters for Artificial Intelligence use in applied behavior analysisThe last few years have witnessed a dramatic surge in the use of Artificial Intelligence (AI) across disciplines beyond behavior analysis—including medicine, law, and education. For example, an American Medical Association (2025) study found that by the fall of 2024, roughly 66% of surveyed physicians were using healthcare AI, representing a 78% increase from the year prior. Yet this rapid expansion has come with significant drawbacks, owing largely to the early stage of AI’s development. Among the first and most consequential problems posed by this rapid expansion is the authenticity of academic writing. For instance, by May 2025, Neurosurgical Review had retracted 129 articles for the unethical use of large language models without proper citation; in comparison, only 191 neurosurgical articles in total were retracted across PubMed between 2012 and 2017. This represents a striking increase in retractions attributable to AI relative to recent years (Madhugiri et al., 2020).

The accelerating use of generative AI in healthcare is reshaping clinical service delivery while concurrently raising complex ethical and practical concerns. Tung et al. (2025) reported that generative AI in healthcare poses significant practical and ethical challenges, including bias within the algorithms, data security risks, unclear accountability, and widespread utilization in the absence of clear regulations. Providers are also sounding the alarms. For example, Hou et al. (2025) found that healthcare professionals are primarily concerned about AI’s threats to patient autonomy, privacy, transparency, and ethical decision-making. Taken together, there is urgency for clear, comprehensive, and specific guidance on ethical implementation. As behavior analysts, we are clearly not alone in facing both the abundant opportunities and challenges of AI and have benefited from the early work of many of our colleagues (e.g., Behavior Analysis Certification Board [BACB], 2024; Jennings & Cox, 2023; Peck et al., 2025; Walz, 2024).

It is in this context that we are delighted to share a review of Practice Parameters for Artificial Intelligence Use in Applied Behavior Analysis, recently published by the Council of Autism Service Providers (CASP). Such guidance is very timely, as many agencies have been utilizing AI in a myriad of ways, including, but not limited to, diverse functions such as public relations, human resources, staff training, and clinical programming. Unfortunately, its use has outpaced the development of protocols, laws, regulations, and ethical guidelines. As a result, behavior analytic and non-behavior analytic organizations alike have been tasked to develop guardrails and guideposts quickly and with few adequate models.

As per CASP, the intent of the guidelines is to address this significant need and provide guidance on payer, regulatory, and ethical matters as well as organizational oversight of AI. As will be highlighted below, the parameters offer strategies on AI selection and deployment, change management, monitoring, and auditing practices. We applaud the decision to place an annotated glossary of over two dozen terms at the very beginning of the resource to prepare the reader, particularly those who are new to AI and its rapid evolution of technology and associated terminology. The inclusion of examples to accompany the definitions is very helpful and appreciated.

There a plethora of external factors that must be considered by organization leaders, such as changing laws, funding policies, and ethics. As of now, federal and state laws on AI in healthcare are constantly being updated, so it is essential to remain apprised, look to available models, and consult with legal experts for the most accurate guidance on governing AI requirements. Although mandates from funding sources have not yet been fully articulated, the CASP Guidelines discourage providers from assuming that all AI use will be permitted or reimbursed. Ethical obligations under BACB codes require clinicians to take full responsibility for all clinical decision-making, so the use of AI is only supplementary to such decisions. Organizations are required to follow new AI governance frameworks that prioritize patient care by emphasizing transparency, data privacy, and clients’ rights.

With respect to organizational oversight encompassing selection, implementation, and monitoring, the CASP Guidelines address seven areas that are briefly highlighted below:

      1. AI System Selection: Akin to our work as behavior analysts, the guidelines urge users to conduct an appropriate assessment prior to implementation and incorporation with an eye toward impact of AI use on roles and processes, including additional responsibilities that would arise. This also includes evaluation of both organizational and client-side needs.
      2. Transparency: Three dimensions of transparency are highlighted in this resource. Accountability pertains to understanding how the system works and how output is derived from a decision-making standpoint. Safety and quality are inextricably linked to transparency and highlight the need for continuous and well conceptualized improvement. Lastly, informed consent pertains not just to the agency using (and understanding the processes involved) but that sufficient information is shared with consumers as well.
      3. Social Significance: Evaluating the social significance of AI use not only benefits the agency but also impacts services and outcomes for consumers, both desired and unintended ones. For instance, impact on privacy, individualization, and fidelity will require careful monitoring.
      4. Deployment: Adequate communication with employees and supervisees is foundational to effective and ethical incorporation of AI, as well as careful timing, and adequate training both with respect to onboarding and ongoing implementation. Social significance also spills over into deployment, as it is important to assess the comfort level of all employees employing AI.
      5. Monitoring and Auditing: Careful oversight enables the agency to assess if the AI system is operating as intended. This oversight includes identifying the individual or individuals responsible, how and when monitoring occurs, and the development of a decision tree to assess how to manage unexpected or intended outcomes.
      6. Error Reporting: These unexpected or intended outcomes may take many forms, and users need a structured and transparent process to report them. The systemic, administrative, and training implications of errors would need to be considered carefully to maintain commitment to high quality and ethical practices.
      7. AI System Deprecation: The agency should be cognizant when chosen AI systems have exceeded their shelf life and make the full array of needed adjustments as warranted, again while maintaining service quality.

Again, we encourage readers to peruse the full resource to learn more about the above dimensions of organizational oversight, as the authors have done a wonderful job in presenting the information in a cohesive and helpful manner. Undoubtedly, the scope and depth of AI will continue to advance in the years ahead, and it will remain important for both agencies and individuals to ensure alignment and compliance with federal and state laws, emerging ethical standards, and obligations to clients, as well as funding sources. We are confident that CASP will continue to support the community to best leverage this rapidly evolving technology and are grateful to CASP for publishing this resource.

References

American Medical Association. (2025, February 26). 2 in 3 physicians are using health AI—up 78% from 2023. https://www.ama-assn.org/practice-management/digital-health/2-3-physicians-are-using-health-ai-78-2023

Behavior Analysis Certification Board (2024, July). BACB Newsletter https://www.bacb.com/wp-content/uploads/2024/07/BACB_July2024_Newsletter-240911-a.pdf

Hou, J., Cheng, X., Liao, J., Zhang, Z., & Wang, W. (2025). Ethical concerns of AI in healthcare: A systematic review of qualitative studies. Nursing Ethics. Advance online publication. https://doi.org/10.1177/09697330251385024

Jennings, A. M., & Cox, D. J. (2023). Starting the conversation around the ethical use of artificial intelligence in applied behavior analysis. Behavior Analysis in Practice, 17(1). https://doi.org/10.1007/s40617-023-00868-z

Madhugiri, V. S., Nagella, A. B., & Uppar, A. M. (2020). An analysis of retractions in neurosurgery and allied clinical and basic science specialties. Acta Neurochirurgica, 163(1), 19–30. https://doi.org/10.1007/s00701-020-04615-z

Peck, S., O’Brien, C., Bourret, J., & Agostinelli, D. (2025). ChatGPT versus clinician responses to questions in ABA: Preference, identification, and level of agreement. Journal of Applied Behavior Analysis, 1–13. https://doi.org/10.1002/jaba.70029

Ravindranath, P. (2025, May 27). 129 papers retracted, and recurrent use of a single sentence in innumerable papers by Indian researchers. Science Chronicle. https://sciencechronicle.in/2025/05/27/129-papers-retracted-a-single-sentence-recurs-in-innumerable-papers-by-indian-researchers/

Tung, T., Hasnaeen, S. M. N., & Zhao, X. (2025). Ethical and practical challenges of generative AI in healthcare and proposed solutions: a survey. Frontiers in digital health, 7 . https://doi.org/10.3389/fdgth.2025.1692517

Walz, R. (2024, January 12). Navigating the ethical terrain of generative AI in behavior analysis: A three-part series. https://science.abainternational.org/2024/01/12/6836/

Reference for this article:

Celiberti, D., & Sasikumar, P. (2025). Review of Practice parameters for Artificial Intelligence use in applied behavior analysis. Science in Autism Treatment, 23(1).

Other Related ASAT Reviews:

Related ASAT Articles:

 

#BehaviorAnalysts #EarlyCareerAndStudents #Faculty #TrainingandSupervision

 

Print Friendly, PDF & Email