Estimating slow drifts in decision criterion using a hierarchical linear dynamical system (hLDS)

Our daily life is permeated with decisions. Prominent psychological models formalize decision-making as a process of evidence accumulation which is terminated once a boundary is reached. These models assume fixed parameters over the course of a block or a condition, and trial-by-trial changes are considered noise. However, decisions are not isolated events with unchanged dynamics when stimuli are repeatedly presented; decision policies are affected by previous experience in specific ways. Most prominently, decision-making is characterized by choice history biases, a behaviorally observed tendency to repeat (or alternate) previous responses in two-alternative forced-choice tasks. Currently, it remains unclear how such history biases can be understood within evidence accumulation model. In the current project, I will investigate whether a single mechanism underlies choice history bias, and its modulation by contextual factors such as decision confidence. Moreover, I will aim to demonstrate that all these findings are naturally accounted for within Bayesian decision theory. Finally, novel predictions from this Bayesian account concerning the neural mechanisms underlying choice history biases will be tested using EEG recordings. Collectively, results from this project will form a solid basis for a more dynamic view on decision-making, whereby decisions reach further than the current sensory input.