Causal judgments about atypical actions are influenced by agents' epistemic states

Abstract

A prominent finding in causal cognition research is people's tendency to attribute increased causality to atypical actions. If two agents jointly cause an outcome ("conjunctive causation"), but differ in how frequently they have performed the causal action before, people judge the atypically acting agent to have caused the outcome to a greater extent than the normally acting agent. In this paper, we argue that it is the epistemic state of an abnormally acting agent, rather than the abnormality of their action, that is driving people's causal judgments. Given the predictability of the normally acting agent's behaviour, the abnormal agent is in a better position to foresee the consequences of their action. We put this hypothesis to test in four experiments. In Experiment 1, we show that people judge the atypical agent as more causal than the normally acting agent, but also perceive an epistemic advantage of the abnormal agent. In Experiment 2, we find that people do not judge a causal difference if there is no epistemic asymmetry between the agents. In Experiment 3, we replicate these findings for a scenario in which the abnormal agent's epistemic advantage generalises to a novel context. In Experiment 4, we extend these findings to mental states more broadly construed. We develop a Bayesian network model that predicts the degree of mental states based on action normality and epistemic states, and find that people infer mental states like desire and intention to a greater extent from abnormal behaviour. We discuss these results in light of current theories and research on people's preference for atypical causes.

The Pervasive Impact of Ignorance

“…More specifically, we both (1) attempt to replicate the effect of normality that has been previously demonstrated, and then (2) compare the replicated effect to minimally modified versions of the original materials that allow us to ask whether the effect of normality is sensitive to changes in the agents' epistemic states. Drawing on recent work on the impact of ignorance on normality effects in causal judgments (Kirfel & Lagnado, 2021a;Samland & Waldmann, 2016), we predict a similar effect of ignorant norm-violation on judgments of intentional action, freedom, happiness, etc. That is, we predict that the difference in judgments about an agent who does not violate a norm and an agent who unknowingly violates a norm (here analysed as effect sizes) will be smaller than the difference between an agent who does not violate a norm and one who knowingly violates a norm.…”

“…In addition, the motivated reasoning account still faces the challenge of explaining the similar impact of descriptive rather than prescriptive norms (Gerstenberg & Icard, 2020;Kominsky et al, 2015;Morris, Phillips, Gerstenberg, & Cushman, 2019): Agents who deviate from typical or usual behaviour are judged as having acted more intentionally , are seen as more causal (Icard et al, 2017;Kirfel & Lagnado, 2021a;Kominsky et al, 2015), are attributed more free will (Monroe & Ysidron, 2021) or negative emotions such as regret (Fillon, Lantian, Feldman, & N'gbala, 2019;Kahneman & Miller, 1986), and so on. Critically, more recently, the effect of ignorance has also been shown to moderate the effect of descriptive norm violations on causal judgments in an identical fashion (Kirfel & Lagnado, 2021a). Some have suggested that statistical norm-deviances and atypical behaviour might likewise influence responsibility judgements (Sytsma, 2020a;Sytsma, Livengood, & Rose, 2012).…”

“…In that sense, the attribution of causality to a normally or abnormally acting agent simply tracks people's responsibility judgements (Sytsma, 2020a(Sytsma, , 2021, often because such behaviours also co-vary in or give raise to inferences about more moral features of that action (Livengood et al, 2017;Sytsma, 2019Sytsma, , 2020a. Such accounts however, still face the challenge to explain the role of agent ignorance in case of descriptive norms, such as the finding that people do not increase causal attributions to atypically acting agents who at the same time are ignorant about the typicality of other people's behaviour (Kirfel & Lagnado, 2021a).…”

The role of causal representations in moral judgment

“…Furthermore, we hypothesized that differences in people's causal model representations of moral scenarios, be they on the level of causal structure or on the level of causal strength, do not impact moral judgments directly. It is well established that moral judgments depend not only on the objective relations between actions and their effects in the world, but also crucially on agents' presumed or actual mental states ; J. T. Kirfel & Lagnado, 2021a;. In the context of chains, outcome foreseeability seems particularly relevant.…”

“…Descriptively, people are also clearly sensitive to agents' mental states when it comes to making judgments about moral permissibility , blame , punishment , liability (J. T. , or agent causation Kirfel & Lagnado, 2021a;.…”