Abstract:
For decision making in environments with hidden changes in context and consequent uncertainty, it is suboptimal to perfectly accumulate evidence. Instead, an ideal strategy involves a non-linear belief updating process that is sensitive to these unexpected changes and adapts according to the changing statistics of the environment. Brain states described by fluctuating levels of arousal dictate variability in information processing. Neuromodulators such as phasic norepinephrine can reconfigure large-scale neural circuits of the decision machinery, but are they involved in mediating flexible decision-making? Building on what we know about neural mechanisms underlying behaviour that also cause changes in pupil size, what can we infer about these flexible mechanisms that are typically inaccessible, from measurements of the pupil? We use pupillometry to obtain pupil dilation as a proxy for arousal and employ a non-linear normative model to infer latent parameters of the belief updating process in healthy human participants performing a hierarchical decision task involving uncertainty. We find that behavioural measures of change point detection and uncertainty are encoded in the dynamics of pupil diameter on fast and slow time scales, respectively. Our findings implicate the involvement of arousal systems of the brainstem in the belief-updating process, whereby arousal serves to facilitate flexible decision-making in uncertain environments.
