Learning to sample: Using contextual cues to flexibly adjust information sampling in decision making

Standard sequential sampling models have been highly successful in explaining how we make everyday, yet important, decisions, such as when it is safe to cross the road. One important recent proposal to improve upon these models has been to adapt the decision process to the estimated instability of the environment. In everyday life, however, we often have to switch back and forth between multiple environments, differing in instability, and learn to adapt to them simultaneously. Therefore, we propose a novel model that is able to simultaneously adapt the decision process to different environments by associating contextual cues with environment instabilities. We will develop this model and run behavioural experiments in which participants perform a random dot motion task, while switching back and forth between different environments, recognisable by contextual cues. We will evaluate whether our model explains the behavioural data better than other models. Moreover, we will investigate whether people can appropriately generalise these learned context-instability associations, when having to make optimal decisions in novel, yet similar, environments. Finally, these paradigms will also give us a newfound window onto the neural signatures of these task execution parameters. Therefore, using representational similarity analysis, we will study which brain regions show a voxel pattern response that is consistent with the context-specific regulation of this decision making parameter.