Using Machine Learning to Create an Adaptable, Scalable, and Interpretable Behavioral Model

In this study, we introduce Adaptable Scalable Best Estimate and Sampling Tools (AS-BEAST), an interpretable model of human decision-making under uncertainty, that fuses the foundational principles of BEAST, a behavioral model grounded in psychological theory, with the capabilities of machine learning techniques. Our strategy involves mathematically formalizing BEAST as a differentiable function and representing it in a computational graph. This approach facilitates the learning of model parameters using automatic differentiation and gradient descent. AS-BEAST scales to larger data sets and adapts to new data more efficiently, while preserving the psychological interpretability of the original model. Evaluation of AS-BEAST on the largest publicly accessible data set of human choice under uncertainty shows that it predicts choice at state-of-the-art levels, similar to those of less interpretable deep neural networks and better than several benchmarks, including the original BEAST model. Importantly, AS-BEAST provides interpretable explanations for choice behavior, leading to the extraction of novel psychological insights from the data. This research demonstrates the potential of machine learning techniques to enhance the scalability and adaptability of models rooted in psychological theory, without compromising their interpretability or insight generation capabilities.