Cognitive Models as a Tool to Link Decision Behavior with EEG Signals

EEG is a direct measure of cortical neuronal activities, making it particularly well-suited to identify latent computations that underlie learning and decision making. This chapter takes a targeted look at how EEG signals and cognitive models of decision making can be linked in a mutually beneficial way. The first section begins with a tutorial on commonly used “linking” approaches, with a focus on sequential sampling models of decision making. This includes a depiction of increasingly sophisticated studies: from condition-level summaries of EEG signals with independently estimated model parameters to regressing trial-level EEG signals with trial-level estimates of model parameters. We then draw attention to joint modeling approaches that assess the bidirectional relationship between EEG signals and parameters of cognitive models. These approaches provide integrated, confirmatory frameworks that propose a common latent source that generates predictions for multiple outputs, such as behavior and neural data. The second section is a review of linking approaches in reinforcement learning models of decision making. This includes a brief history of formal linking approaches between EEG and reinforcement learning, from origins in qualitative comparison of prediction errors and aggregate EEG summaries through to regression of individual-trial data. The chapter concludes with some caveats to current linking approaches and a discussion of potential future directions for advancing the methods of linking EEG signals with cognitive models.