|
| Selectionism and Machine Learning |
| Sunday, May 24, 2020 |
| 4:00 PM–4:50 PM |
| Marriott Marquis, Level M1, Georgetown |
| Area: PCH |
| Instruction Level: Intermediate |
| Chair: Temple S Lovelace (Duquesne University) |
| |
| Selection Sciences: A Five-Level Undergraduate Course in Selection by Consequences |
| Domain: Theory |
| CRISS WILHITE (California State University Fresno ) |
| |
| Abstract: Skinner formally introduced selection by consequences to behavior analysis in 1981. Since then, behavior analysts and researchers from other disciplines have expanded our understanding of selection to include units of selection, variation of those units, unit-environment interaction, selection, and maintenance of the units. Many researchers now recognize five levels at which selection occurs: evolution, epigenetics, immunology, behavior, and cultural practices. The Department of Psychology at Fresno State has offered a topics course called “Selection Sciences” for four years. The main content is how these systems work in general, with detailed analyses at each level. Additional topics include the history of selectionist approaches, dynamical systems, complexity, emergence, and chaos theory. The course is open to all undergraduates, fulfills a Psychology major requirement, and has been well received. Classes such as this may help bridge gaps between behavior analysis and other sub-disciplines within psychology along with related sciences, as selection processes are inherently multidisciplinary. |
| |
| Machine Learning and Behavior Analysis: Can Artificial Intelligence Reduce Bias in the Functional Behavior Assessment Process? |
| Domain: Theory |
| TEMPLE S LOVELACE (Duquesne University) |
| |
| Abstract: Artificial Intelligence has become a central part of understanding the teaching and learning processes in education (Nafea, 2017). In education, machine learning has been touted as a way for educators to improve efficiency and to personalize the learning that students and teachers undertake (Petrilli, 2018). As more efficient methods of teaching and practice are sought to improve the way we approach behavioral analytic training, such as virtual and mixed reality, a single question remains, “When will behavior analysis get bit by the artificial intelligence bug?” Using data from the author’s systematic review on functional behavior assessment, the authors will walk participants through “FBAware” a simulated application that can detect if a behavior analyst has bias in determining the function of behavior. Next, the authors will present sample linear models, R^2, and adjusted R^2 values of the predictive models as a use case for what a predictive analytics could look like for organizations and universities. Lastly, the authors will talk about the promise and pitfalls that predictive analytics could bring in indicating where bias exists and how the field can become forward thinking in order to shape a future where artificial intelligence and behavior analysis coincide. |
| |
|
| |