Investigating Surprise Signals in the Anterior Insula

Overview

The investigators propose a behavioral experiment with SEEG recording and stimulation, to both confirm the role of a brain region known as the anterior insula in identifying surprise, and disambiguate between competing principles behind adaptation: optimizing and satisficing. Optimizers continue to learn and adapt if performance can be improved, while satisficers are satisfied with a good enough performance and will cease adapting once that is reached. To study surprise signals in the anterior insula, a brain structure where these signals have been very prominent, the investigators will employ an experiment with subjects who are under SEEG (stereoelectroencephalogram) recording, that is, recording from electrodes which have been surgically implanted in the brain. These recordings will be done as patients perform a task where they try to anticipate the movements of a target on a line in two different learning environments (conditions). The experimenters will then determine whether these signals reflect surprise relative to past engagement with the environment, or surprise that reveals that the agent no longer feels in control because uncertainty is not in line with the reference model. If evidence is consistent with the former, adaptation reflects traditional reinforcement and aims at optimizing behavior. If evidence instead is consistent with the latter, behavior is guided by a prior model (a reference model) and behavior is satisficing. An fMRI study by d'Acremont and Bossaerts provides initial evidence that activation in the anterior insula supports the satisficing hypothesis, however it lacks the temporal granularity to completely rule out optimizing. In the current project, the investigators propose to use the higher time resolution of SEEG recordings to confirm these findings and reject the optimizing hypothesis. Additionally, stimulations of the anterior insula during a subset of trials will be used to determine whether insular activation following surprise signals and preceding changes in behavior (learning) is merely correlational or in fact causal. Stimulation will allow us to determine to what extent the subjects' sense of control and subsequent behavior can be influenced in accordance with surprise-based modeling of behavior. The cohort for this study will be patients with drug-resistant, focal epilepsy and who are hospitalized at the Hôpitaux Universitaires de Genève (HUG) for pre-surgical evaluation of their epilepsy using SEEG. The protocol will run in parallel with the patients' clinical procedures.