Predicting Heuristic Decisions in Child Welfare: A Neural Network Exploration

Behavior analysts have long recognized the benefits of closely following their data; however, the data we are following may be moving faster than the tools we have to accurately analyze and predict future behaviors. This problem even saturates behavior-analytic investigations that focus on the evaluation of complex data related to public policy issues in areas such as poverty, geriatrics, and child welfare practice. In the face of this research enigma, there exists a more powerful and precise set of classification and prediction platforms for researchers in the behavioral sciences. In this article, we describe a combination of neural network strategies that predict child welfare professionals’ decision making. Extending the data analysis from the Yelick and Thyer (2019) study, we employed our current version of the Kohonen self-organizing map in conjunction with our deep neural network as a strategy for identifying participants who were at high probability for making heuristic decisions.