Information Theoretic Characterization of Uncertainty Distinguishes Surprise From Accuracy Signals in the Brain

Abstract

Uncertainty presents a problem for both human and machine decision-making. While utility maximization has traditionally been viewed as the motive force behind choice behavior, it has been theorized that uncertainty minimization may supersede reward motivation. Beyond reward, decisions are guided by belief, i.e., confidence-weighted expectations. Evidence challenging a belief evokes surprise, which signals a deviation from expectation (stimulus-bound surprise) but also provides an information gain. To support the theory that uncertainty minimization is an essential drive for the brain, we probe the neural trace of uncertainty-related decision variables, namely confidence, surprise, and information gain, in a discrete decision with a deterministic outcome. Confidence and surprise were elicited with a gambling task administered in a functional magnetic resonance imaging experiment, where agents start with a uniform probability distribution, transition to a non-uniform probabilistic state, and end in a fully certain state. After controlling for reward expectation, we find confidence, taken as the negative entropy of a trial, correlates with a response in the hippocampus and temporal lobe. Stimulus-bound surprise, taken as Shannon information, correlates with responses in the insula and striatum. In addition, we also find a neural response to a measure of information gain captured by a confidence error, a quantity we dub accuracy. BOLD responses to accuracy were found in the cerebellum and precuneus, after controlling for reward prediction errors and stimulus-bound surprise at the same time point. Our results suggest that, even absent an overt need for learning, the human brain expends energy on information gain and uncertainty minimization.

Surprise: a unified theory and experimental predictions

“…The neural signatures of the Shannon surprise (and the state prediction error, c.f. Proposition 3) have been found in numerous EEG (Gijsen et al, 2021;Kolossa et al, 2015;Kopp & Lange, 2013;Mars et al, 2008;Meyniel et al, 2016;Modirshanechi et al, 2019;Mousavi et al, 2020), MEG (Maheu et al, 2019;Meyniel, 2020;Mousavi et al, 2020), and fMRI (Gläscher et al, 2010;Konovalov & Krajbich, 2018;Loued-Khenissi & Preuschoff, 2020) studies. In particular, it is well-known that the amplitude of the EEG P300 component correlates with the Shannon surprise (Kolossa et al, 2015;Kopp & Lange, 2013;Mars et al, 2008;Meyniel et al, 2016;Modirshanechi et al, 2019;Mousavi et al, 2020).…”

“…In particular, it is well-known that the amplitude of the EEG P300 component correlates with the Shannon surprise (Kolossa et al, 2015;Kopp & Lange, 2013;Mars et al, 2008;Meyniel et al, 2016;Modirshanechi et al, 2019;Mousavi et al, 2020). (Kopp & Lange, 2013), MEG (Meyniel, 2020), and fMRI (Konovalov & Krajbich, 2018;Loued-Khenissi & Preuschoff, 2020).…”

Gambling on Others’ Health: Risky Pro-social Decision-Making in the Era of Covid19

“…Uncertainty is an intrinsic feature of decision-making as we base choices on imperfect knowledge and incomplete information ( [59]). We focus here on risk, an expected form of uncertainty ( [67])which is known to steer decision-making away from maximizing predicted gains (expected value theory ( [11]; [2]; [25]; and Prospect Theory [85]): agents are risk-averse in the gain domain, assigning more value to less risky options but less value to sure options when facing loss.…”

The Oculomotor Signature of Expected Surprise

“…Prediction is based on prior information learned in a previously experienced similar context. However, it is inevitably tainted with some spatial and/or temporal uncertainty that has to be explicitly estimated in the decision process 4 . The decision to initiate a particular behavioral response could be influenced by this estimation of uncertainty.…”

“…Surprise is a polysemic term that can be interpreted in several different ways and is hypothesized to underlie essential functions like learning 27,28 and emotions 29,30 . In the present study, we will use 'surprise' to quantify the unexpectedness of a particular event occurring from a random variable ('stimulus-bound' surprise 4 ). Surprise will be computed using the definition from Shannon's information theory and expressed in bits 31 .…”