How robust is the optimistic update bias for estimating self-risk and population base rates?

Abstract

Humans hold unrealistically optimistic predictions of what their future holds. These predictions are generated and maintained as people update their beliefs more readily when receiving information that calls for adjustment in an optimistic direction relative to information that calls for adjustment in a pessimistic direction. Thus far this update bias has been shown when people make estimations regarding the self. Here, we examine whether asymmetric belief updating also exists when making estimations regarding population base rates. We reveal that while participants update beliefs regarding risk in the population in an asymmetric manner, such valence-dependent updating of base rates can be accounted for by priors. In contrast, we show that optimistic updating regarding the self is a robust phenomenon, which holds even under different empirical definitions of desirable information.

Figure 1. Paradigm.

(a) In each trial, participants were presented with a short description of 1 of 80 adverse events and asked to estimate how likely this event was to occur to themselves in the future and how likely the event was to happen on average in the population. They were then presented with the base rate in a demographically similar population. Finally, participants were asked to rate how negative/positive they found this information. The second session was the same as the first except that the base rate was not presented and participants did not submit any ratings. Examples of trials in which the participant's estimate of the event occurring to themselves and the base rate was (b) higher or (c) lower than the provided base rate. In the specific examples shown here, under either classification scheme therefore these trials would be categorized as desirable and undesirable trials respectively.

Figure 2. Update for self risk and base rates under different classifications of desirability.

Participants update estimates of their self risk more when the information they received was desirable compared to undesirable. Participants update estimates of their base rates more when the information they received was desirable compared to undesirable under classification one but not under classification two after controlling for relevant covariates. Update calculated as first minus second estimation for desirable trials and the reverse for undesirable trials (positive values therefore indicate a move towards the information presented). Error bars are SEM; *indicates statistical significance at a threshold of P<0.05, two tailed after controlling for all relevant covariates. Trials classified as desirable when the participant overestimated the probability of the event occurring and undesirable when the participant underestimated the probability of the event occurring: (a) to themselves relative to the provided base rate; (b) in the population relative to the provided base rate.

Figure 3. Regression coefficients predicting update from estimation errors.

Estimation Error of self risk (i.e. the difference between a participants' estimate of self risk and provided base rate) significantly predicted update of self risk both for trials in which subjects received desirable and undesirable information, and under both classifications. Estimation Error of base rate (i.e. the difference between a participants' estimate of base rate and provided base rate) significantly predicted update of base rate both for trials in which participants received desirable and undesirable information, and under both classifications. Error bars are SEM; * indicates statistical significance at a threshold of P<0.05, two tailed paired sample t-test after controlling for relevant factors; ** indicates significantly different to a mean of 0, one sample t-test (P<0.05).