Combining psychological models with machine learning to better predict people’s decisions

Abstract

Creating agents that proficiently interact with people is critical for many applications. Towards creating these agents, models are needed that effectively predict people's decisions in a variety of problems. To date, two approaches have been suggested to generally describe people's decision behavior. One approach creates a-priori predictions about people's behavior, either based on theoretical rational behavior or based on psychological models, including bounded rationality. A second type of approach focuses on creating models based exclusively on observations of people's behavior. At the forefront of these types of methods are various machine learning algorithms. This paper explores how these two approaches can be compared and combined in different types of domains. In relatively simple domains, both psychological models and machine learning yield clear prediction models with nearly identical results. In more complex domains, the exact action predicted by psychological models is not even clear, and machine learning models are even less accurate. Nonetheless, we present a novel approach of creating hybrid methods that incorporate features from psychological models in conjunction with machine learning in order to create significantly improved models for predicting people's decisions. To demonstrate these claims, we present an overview of previous and new results, taken from representative domains ranging from a relatively simple optimization problem and complex domains such as negotiation and coordination without communication.

Behavioral Analysis of Insider Threat: A Survey and Bootstrapped Prediction in Imbalanced Data

“…While the accurate prediction of an insider's moves was the main motivation, only Nash equilibrium analysis has been used to capture the insider's future actions, in order to respond properly. It is well known that people do not follow Nash equilibrium strategies [77], [78] including intelligent security people [79]. In addition, the game is a one-shot game, and a discussion on repeated games was not included.…”

Using a Stochastic Agent Model to Optimize Performance in Divergent Interest Tacit Coordination Games

“…It has already been suggested that achieving successful human–machine collaboration requires the modeling of human behavior for predicting people’s decisions [ 7 , 8 ]. It has been shown, for example, that information about drivers (i.e., driving style) improves the prediction models regarding the use of an automated assistive system [ 9 ].…”

Making Friends on the Fly: Advances in Ad Hoc Teamwork

“…We expect that this general approach will apply well to dealing with humans despite the wide range of behaviors that humans can exhibit. This expectation is supported by research showing that a small number of behaviors captures the majority of human behaviors in certain tasks [108]. Similarly, research on bandit problems suggests that only a limited number of strategies are viable in social settings [97].…”

“…This case is of interest because a finite set of behaviors can often cover the space of likely behaviors. For example, analysis of bandit problems [97], ad hoc teamwork [19], and using machine learning with psychological models [108] suggests that a small number of behaviors can represent the spread of possible behaviors.…”