Association for Behavior Analysis International

More than 40 years ago, Hayes et al. (1980) shared concerns about the science of applied behavior analysis, namely what they viewed as the proliferation of behavioral technology unmoored by coherent and evolving theories of learning (see also Pierce & Epling, 1980). This trend has maintained and, perhaps, worsened in the intervening years (e.g., Mahoney et al. 2019; Sosine & Cox, 2023). Behavior Analysis as a science and profession, specifically as practiced by BCBAs, is currently at a watershed moment with the proliferation of big data and AI-powered tools. Behavior analysis (as a profession) currently lacks standard clinical care metrics or systematic and replicable methods to recommend treatment hours outside very narrow contexts and populations (i.e., early intervention for individuals with autism/autistic individuals). Big data and AI-powered tools offer potential methods to fill these gaps by bridging basic and applied science in new ways. In this symposium, presenters will share their current and planned work in this area, such as precisely targeting interventions and treatment recommendations based on individual client needs. Presenters will also lead a discussion around the evolving impact of big data and AI-powered tools on clinical practice and what the skills of tomorrow’s most effective behavior analysts might look like.

Entry level practitioners and researchers in behavior analysis or related fields.

Artificial intelligence (AI) is evolving at a breathtaking pace. Building impactful AI products requires large amounts of data relevant to the tasks the AI tools augment. Fortunately, behavior analytic services involve tremendous amounts of data already being collected that can, in turn, be used to build useful AI systems. AI applied to behavior analytic service provision can: increase the efficiency and effectiveness of behavior analysts’ impact for the patients they serve; provide greater consistency in treatment recommendations and approaches; and improve the speed and degree to which recommendations can be tailored based on each learner’s unique profile. In this presentation, we discuss how RethinkFirst is leveraging AI to improve, inform, and augment the clinical decisions and subsequent impact of certified behavior analysts’ skill sets. Specifically, we discuss how AI allows us to identify unique patient and provider clusters, match patients to providers who are best equipped to maximize progress and create meaningful change for individual learners, and recommend treatment goals and targets to optimize learning pathways tailored to individual learners. Along the way, we also highlight the necessary rigor, safeguards, and ongoing ethical considerations that technology developers and consumers should proactively discuss when developing and using AI systems at scale.

A great deal of attention is being allocated to artificial intelligence (AI) among the general and professional public. This attention spans everything from curiosity and concern to excitement and naiveté. Within behavior analysis, AI understandably is receiving greater attention with concerns over its potential role and how integrating AI into areas of our work will bring us closer or further away from our subject matter. Central Reach as a provider of end-to-end software solutions for behavior analysts working in the Autism and IDD care space has begun a concerted effort to explore what role, if any, AI can play in supporting aspects of applied work. Utilizing Central Reach’s internal technical and behavior analytic expertise several initiatives have been undertaken to determine the feasibility and efficacy that AI informed tools can contribute to aspects of the applied behavior analytic workflow. This presentation reviews these efforts with particular attention to the prudent iterative development process, quality assurance, and the validity parameters that have been created. Furthermore, implications of employing these tools and others like them will be discussed.

In the ever-evolving landscape of Applied Behavior Analysis (ABA), the integration of Artificial Intelligence (AI) presents an exciting opportunity to revolutionize clinical decision-making. Keohane and Greer (2005) demonstrated the benefits of data-informed decision-making, and how an algorithm that increased the quantity of decisions made ultimately led to students learning more quickly. With the emergence of Large Language Models (LLMs) like ChatGPT, the opportunities to integrate AI into the clinical decision-making process have expanded even further. This presentation explores how incorporating LLMs into the clinical decision-making process can amplify the effects demonstrated by Keohane and Greer on a larger scale. In partnership with computer scientists, we trained an LLM on select clinical data of thousands of clients, and created individualized reports for clients, providing key information to clinicians that they need to make critical decisions. By harnessing AI, we showcase the ability to scale a decision-making model across 400 BCBAs and analyze 20,000 graphs, ultimately fostering more informed daily clinical choices. This presentation is tailored for behavior analysts, emphasizing that AI is not a replacement but a powerful support tool. It shifts the paradigm of decision-making from individual analysis to a data-driven synthesis, aiding BCBAs in prioritizing clients, program progress evaluation, and drawing on past effective decisions for similar cases. The future of ABA lies in the synergy of human expertise and AI's analytical capabilities. Join us as we unravel the transformative potential of AI, bridging the gap between data and effective clinical practices in Applied Behavior Analysis.

Standardized models of analysis in behavior analysis are restricted to either descriptive assessments or experimental conditions framed within a linear direct contingency framework. These models have come under scrutiny in recent years from all directions highlighting their limited scope and scale. Additionally, advancements in machine learning and data science have compounded such critiques producing the capacity for greater degrees of precision across modalities. The purpose of this paper was to take the first step in establishing a comprehensive interpretative framework consistent with a multi-scalar analysis of behavior (Baum, 2002). To accomplish this goal the group of authors conducted a systematic review of the Journal of Applied Behavior Analysis (JABA) to isolate critical case variables and evaluate their predictive validity of treatment selection across thirteen years. Preliminary results indicate the potential for a robust area of research in quantifying clinical decision-making and effectively tying practitioner choices to the literature. Implications will be proposed.