Updating Beliefs under Perceived Threat

Abstract

Humans are better at integrating desirable information into their beliefs than undesirable information. This asymmetry poses an evolutionary puzzle, as it can lead to an underestimation of risk and thus failure to take precautionary action. Here, we suggest a mechanism that can speak to this conundrum. In particular, we show that the bias vanishes in response to perceived threat in the environment. We report that an improvement in participants' tendency to incorporate bad news into their beliefs is associated with physiological arousal in response to threat indexed by galvanic skin response and self-reported anxiety. This pattern of results was observed in a controlled laboratory setting (Experiment I), where perceived threat was manipulated, and in firefighters on duty (Experiment II), where it naturally varied. Such flexibility in how individuals integrate information may enhance the likelihood of responding to warnings with caution in environments rife with threat, while maintaining a positivity bias otherwise, a strategy that can increase well-being.SIGNIFICANCE STATEMENT The human tendency to be overly optimistic has mystified scholars and lay people for decades: How could biased beliefs have been selected over unbiased beliefs? Scholars have suggested that although the optimism bias can lead to negative outcomes, including financial collapse and war, it can also facilitate health and productivity. Here, we demonstrate that a mechanism generating the optimism bias, namely asymmetric information integration, evaporates under threat. Such flexibility could result in enhanced caution in dangerous environments while supporting an optimism bias otherwise, potentially increasing well-being.

Sympathetic and parasympathetic involvement in time constrained sequential foraging

“…ACC has also been implicated amongst other prefrontal sites in tracking reward history (Seo et al, 2007;Bernacchia et al, 2011). This would align with previous neurophysiological work in human learning, which implicates heightened stress states -assayed via electrodermal activity -with closer trialwise tracking of adjustments in information integration (Li et al, 2011), with selective stress-driven sensitivity increases to negative prediction errors, eradicating positive learning biases that emerge under calm conditions when stress levels are normal (Garrett et al, 2018).…”

“…Appraising sequential offers of reward relative to an unknown future opportunity and a time cost requires an optimization policy that draws on a belief about the richness characteristics of the current environment. Across a range of experiments, including reinforcement-learning tasks, belief updating paradigms and prey selection, information integration shows a positive bias Sharot, 2014, 2017;Garrett et al, , 2018Kuzmanovic et al, 2015Kuzmanovic et al, , 2016Eil and Rao, 2011;Korn et al, 2012;Kuzmanovic and Rigoux, 2017;Lefebvre et al, 2017;Garrett and Daw, 2019).…”

“…Appraising sequential offers of reward relative to an unknown future opportunity and a time cost requires an optimization policy that draws on a belief about the richness characteristics of the current environment. Across a range of experiments, including reinforcement-learning tasks, belief updating paradigms and prey selection, information integration shows a positive bias (Eil & Rao, 2011;Garrett, González-Garzón, Foulkes, Levita & Sharot, 2018;Korn, Prehn, Park, Walter, & Heekeren, 2012;Kuzmanovic and Rigoux, 2017;Kuzmanovic, Jefferson & Vogeley, 2015Lefebvre et al, 2017;Garrett & Daw, 2019). That is, the rate at which humans update their belief about a probability, association or reward rate is more sluggish if the new information carries a negative or aversive valence or telegraphs a deterioration of the current belief.…”

Biased belief updating and suboptimal choice in foraging decisions

“…The second (Asymmetric Model) learned to a different degree from positive vs. negative prediction errors. This feature, which has been observed in a number of other learning scenarios (Eil and Rao, 2011;Sharot, 2014, 2017;Garrett et al, , 2018Korn et al, 2012;Kuzmanovic and Rigoux, 2017;Kuzmanovic et al, 2015Kuzmanovic et al, , 2016Lefebvre et al, 2017;Mobius et al, 2011;Moutsiana et al, 2013Moutsiana et al, , 2015Sharot and Garrett, 2016;Sharot et al, 2011) but not explored in the context of foraging, gave the model the capacity to express learning asymmetries and to predict and capture specific patterns of experience-dependent biases in choices.…”

“…For instance, a model that defined an MVT-like threshold from a simple incremental running average of recent events (Constantino and Daw, 2015) would treat either direction of change symmetrically. We hypothesized, however, that the difference could arise from asymmetric learning, as has been studied in other decision domains (Eil and Rao, 2011;Sharot, 2014, 2017;Garrett et al, , 2018Korn et al, 2012;Kuzmanovic and Rigoux, 2017;Kuzmanovic et al, 2015Kuzmanovic et al, , 2016Lefebvre et al, 2017): individuals changing their subjective estimates of the environment's reward rate differentially depending whether the update was in a positive or negative direction. To formally test this, we adapted a computational model (Constantino and Daw, 2015) to capture how participants maintained ongoing estimates of the environments reward rate during the task and tested whether endowing this model with the capacity to shift estimates up and down at different rates depending on whether the information received suggested that the environment was improving (as would be the case when transitioning from poor to rich) versus deteriorating (as would be the case when transitioning from rich to poor) could account for the relative hesitance of participants to change choices when the environment became worse.…”