Anger-prone individuals are volatile and frequently dangerous. Accordingly, inferring the presence of this personality trait in others was important in ancestral human populations. This inference, made under uncertainty, can result in two types of errors: underestimation or overestimation of trait anger. Averaged over evolutionary time, underestimation will have been the more costly error, as the fitness decrements resulting from physical harm or death due to insufficient vigilance are greater than those resulting from lost social opportunities due to excessive caution. We therefore hypothesized that selection has favored an upwards bias in the estimation of others' trait anger relative to estimations of other traits not characterized by such an error asymmetry. Moreover, we hypothesized that additional attributes that i) make the actor more dangerous, or ii) make the observer more vulnerable increase the error asymmetry with regard to inferring anger-proneness, and should therefore correspondingly increase this overestimation bias. In Study 1 (N = 161), a fictitious individual portrayed in a vignette was judged to have higher trait anger than trait disgust, and trait anger ratings were more responsive than trait disgust ratings to behavioral cues of emotionality. In Study 2 (N = 335), participants viewed images of angry or fearful faces. The interaction of factors indicating target's formidability (male sex), target's intent to harm (direct gaze), and perceiver's vulnerability (female sex or high belief in a dangerous world) increased ratings of the target's trait anger but not trait fear.
Assessing others' personality traits is a key adaptive problem that social cognition evolved to address. Understanding people's personalities allows us to predict others' future behavior and facilitates navigating complex social interactions (Ross, 1977). However, because personality is invisible, it is difficult to assess. Past behavior may reveal underlying traits, but inferences about them (especially from a single observation) are highly uncertain, for two reasons. First, behaviors are produced not only by enduring dispositions, but also by fleeting situations. Proper discounting of situational influences requires repeated observations of an individual across multiple situations (Kelley, 1972), and this cannot always be achieved. Second, people strategically manage their behaviors, at times actively inhibiting the expression of negative traits and compromising observers' ability to discern personal characteristics.
Here, we explore the hypothesis that assessments of an individual's propensity to become angry are adaptively biased. Given that i) conspecifics were a primary source of danger for ancestral humans (Keeley, 1996), and ii) anger motivates violence (Fessler, 2010, Frank, 1988 and Sell, 2009), an important adaptive challenge was predicting an individual's enduring inclination to become angry (i.e., trait anger), a process we term “anger attribution”. Importantly, anger attribution is inherently imperfect, making complete accuracy unlikely, if not impossible.
1.1. Adaptive rationality and error management
The “adaptive rationality” approach contends that the mind was shaped by selection to enhance fitness in ancestral environments rather than to yield accurate judgments (Haselton et al., 2009; see also Funder, 1995, and Krueger & Funder, 2004). Therefore, human cognition can manifest seemingly irrational biases that are, in fact, “adaptively rational.” Anger attribution is one domain in which this might occur. Perceivers can commit one of two errors: underestimate an individual's trait anger (false negative) or overestimate it (false positive). On average, underestimations will have been costlier than overestimations in ancestral populations: assuming that an anger-prone individual was temperate placed the perceiver at risk of assault, whereas assuming that a temperate individual was anger-prone merely led to foregoing potentially profitable interactions. Thus, overall accuracy (i.e., committing false negative and false positive errors with equal frequency) did not maximize fitness over evolutionary time. Rather, in line with error management theory (Haselton and Buss, 2000 and Haselton and Nettle, 2006), we hypothesize that selection favored a biased tendency to commit the less costly false positive — overestimating trait anger. Although the same logic applies to the estimations of state anger, our predictions focus squarely on trait anger because traits predict future behavior, and it is costly to underestimate an individual's anger not only in the moment, but also in future interactions.
Absent objective baselines, investigating a hypothesized bias in judgment requires points of comparison; we employed other negative emotional dispositions, for which we predicted either no biases, or reverse biases (trait underestimation). For instance, in the case of fear directed toward the perceiver, there is no clear asymmetry in the costs of underestimating or overestimating another's propensity to experience fear. Therefore, we do not expect an evolved bias for perceptions of trait fear. If a target displays fear or disgust toward something or someone other than the perceiver, it was likely to have been adaptive to over-attribute their emotions to the situation (and underestimate the corresponding trait), since this enhances alertness to potential hazards. More formally:
