People adaptively use information to improve their internal states and external outcomes

Abstract

Information can strongly impact peoples’ affect, their level of uncertainty and their decisions. It is assumed that people seek information with the goal of improving all three. But are they successful at achieving this goal? Answering this question is important for assessing the impact of self-driven information consumption on people’s well-being. Here, over four experiments (total N = 518) we show that participants accurately predict the impact of information on their internal states (e.g., affect and cognition) and external outcomes (e.g., material rewards), and use these predictions to guide information-seeking choices. A model incorporating participants’ subjective expectations regarding the impact of information on their affective, cognitive, and material outcomes accounted for information-seeking choices better than standard models currently used in the literature, which include objective proxies of those subjective measures. This model also accounted for individual differences in information-seeking choices. By balancing considerations of the impact of information on affective, cognitive and material outcomes when seeking knowledge, participants became happier, more certain and earned more points when they purchased information relative to when they did not, suggesting they adopted an adaptive strategy.

A neural mechanism for conserved value computations integrating information and rewards

“…Here, we take advantage of the fact that our task can present full probability distributions of rewards to test between three hypothesized families of uncertainty measures (Fig 2A,B). One family of uncertainty measures depends only on the probabilities of the potential outcomes, most prominently Shannon entropy ( 51 ) and related quantities such as Kullback-Leibler divergence ( 52 ), which have been used to model information preferences ( 44, 53, 54 ) and many other forms of motivation, cognition, and neural computation ( 55–59 ). A second family depends only on the magnitudes of potential outcomes, including the range, which has been proposed to regulate the dynamic range of neural activity ( 60–63 ).…”