CCN meeting | Jaqueline Gottlieb (Columbia University, USA)

04-03-2021 from 15:00 to 16:00

Attention as rational choice

Despite important advances in our understanding of decision mechanisms, we are severely limited in understanding decisions made under conditions of realistic complexity. A key limitation is the assumption that decision makers attend to information that is given to them by an exogenous process. This contrasts with realistic behavior, in which agents endogenously decide which information to attend to - or be curious about – given their goals and behavioral situation. I will describe new findings from my laboratory about the neural mechanisms of visual information sampling decisions in the fronto-parietal network. I will present the Bayesian prescription attentional sampling should be optimally guided by prior uncertainty and the validity (likelihood) of the available information and describe evidence that these quantities are encoded in the monkey fronto-parietal network. I will show that sampling deviates from the Bayes-optimal prescription in two ways - by virtue of imperfect filtering of low likelihood stimuli and by virtue of biases toward stimuli associated with desirable outcomes – and these deviations may be traced to the neural mechanisms that integrate prior and likelihood and coordinate information gathering with expected rewards.  By revealing the biological mechanisms of realistic information gathering policies, the results advance our understanding of decision making under realistic complexity.