Prediction in human decision making: a modeling approach

ABSTRACT

Human beings can determine optimal behaviors, which depends on the ability to make planned and adaptive decisions. Decision making is defined as the ability to choose between different alternatives. Purpose: this study, we have addressed the prediction aspect of human decision making from neurological, experimental and modeling points of view. Methods: We used a predictive reinforcement learning framework to simulate the human decision making behavior, concentrating on the role of frontal brain regions which are responsible for predictive control of human behavior. The model was tested in a maze task and the human subjects were asked to do the same task. A group of six volunteers including three men and three women at the age of 23-26 participated in this experiment. Results: The similarity between responses of the model and the human behavior was observed after varying the prediction horizons. We found that subjects with less risky choices usually decide based on considering long term advantages of their action selections, which is equal to the longer prediction horizon. However, they are more susceptible to reach suboptimal solutions if their predictions become wrong due to some reasons like changing environment or inaccurate models. Conclusion: The concept of prediction result in faster learning and minimizing future losses in decision making problems. Since the problem solving in human beings is very faster than a trial and error system, considering this ability will help to describe the human behavior more desirably. This observation is compatible to the recent findings about the role of Dorsolateral Prefrontal Cortex in prediction and its relations to Anterior Cingulate Cortex with the ability of conflict monitoring and action selection.