A novel approach to studying strategic decisions with eye-tracking and machine learning

We propose a novel method of using eye-tracking to study strategic decisions. The conventional approach is to hypothesize what eye-patterns should be observed if a given model of decision-making was accurate, and then proceed to verify if this occurs. When such hypothesis specification is difficult a priori, we propose instead to expose subjects to a variant of the original strategic task that should induce processing it in a way consistent with the postulated model. It is then possible to use machine learning pattern recognition techniques to check if the associated eye-patterns are similar to those recorded during the original task. We illustrate the method using simple examples of 2x2 matching-pennies and coordination games with or without feedback about the counterparts' past moves. We discuss the strengths and limitations of the method in this context.