Learning what’s relevant in a largely irrelevant world: The role of selective attention in learning

Abstract

This thesis presents a framework to study the interaction between attention and learning. The framework proposes that learning processes act on an attentionally-filtered representation of the environment and that the attention filter is dynamically modulated by the outcomes of ongoing learning. These assumptions were tested in a series of experiments in which participants performed a multi-dimensional decision-making task with probabilistic rewards. Choice behavior was analyzed using computational models. Some of these models incorporated information about participants’ focus of attention, which was decoded on each trial by combining eye-tracking with pattern classification of functional magnetic resonance imaging (fMRI) data. Model-based analysis of behavior provided preliminary evidence that attention helps determine what we learn about, but we also learn what to attend to.