Katarina Kovacevic, Jonathan Kominsky, Francesca Bonalumi and Christophe Heintz
Affiliation: Department of Cognitice Science, Central European University
Category: Psychology
Keywords: counterfactual reasoning, responsibility ascription, ignorance, epistemic states
Date: Tuesday 2nd of September
Time: 15:00
Location: Room 154 (154)
View the full session: Blame & the Knobe Effect
When something negative happens, we tend to think of alternative events that would lead to a different outcome, or, in other words, we engage in counterfactual simulations (Kahneman & Tversky, 1982). Many researchers agree that people engage in counterfactuals reasoning when they make causality, responsibility and blame judgments (Kominsky & Phillips, 2019; Kirfel & Lagnado, 2021; Alicke et al., 2008; Gerstenberg, 2024). While recent research investigates which counterfactual are perceived as relevant in certain situations (e.g. Icard, Kominsky, & Knobe, 2017), less is known about the causal relationship between counterfactual selection and moral judgements. Previous research has shown that prompting people to consider one counterfactual alternative over the other influences causality judgments (Phillips, Luguri, & Knobe, 2015; Kominsky & Phillips, 2019). In these studies, vignettes included two agents who equally contributed to the outcome, or an agent using the artifact. In the current set of experiments, we further investigated the relationship between counterfactual reasoning and responsibility ascription. We employed a vignette study approach. The scenarios followed a similar structure: an introduction describing the context with main character A doing B (general); information about epistemic states and actions that we manipulated depending on the experimental condition; main character doing S; and finally, consequences. In each experiment, participants were recruited via Prolific and assigned to one of the conditions. They first read a story, after which they answered different questions, among which the measure of responsibility in a form of “To what extent do you agree with the statement: "Perpetrator P is responsible for the consequence C." In Experiment 1, we asked people to spontaneously produce counterfactual alternatives to outcome of the story they read. Main agents in the stories were either knowledgeable (Knowledge condition), fully ignorant (No knowledge), ignorant because it was too effortful to get informed (Hard effort), or because they simply did not take the opportunity to learn (Easy effort). We coded participants’ answers according to the following scheme: a) External (non-epistemic) factors - factors not directly related to the agents' actions or thoughts; b) External epistemic factors - better availability of the information, that is not the consequence of agent’s actions, but the environment; c) Perpetrator’s epistemic actions - actions that lead to learning; d) Perpetrator instrumental actions - actions not related to getting informed; e) Victim actions - actions that victim could have taken to prevent the negative outcome. These categories were the levels of the dependent variable Counterfactual type. Results showed that people conceive different counterfactuals, depending on the reasons behind ignorance, Chi sq.(12) = 232, p < .001, Cramer’s V = 0.52. In No knowledge condition, people most often considered External epistemic factors (45,7%). In Hard effort condition, they considered Perpetrator’s instrumental actions (37%). In Easy effort they most often mentioned Perpetrator’s epistemic actions (86%), while in Knowledge condition, Perpetrator’s instrumental actions were by far the most common answer (81,2%). In addition, we conducted a mediation analysis with condition as a predictor, counterfactual type as a mediator and responsibility rating as an outcome. Apart from the direct effect of condition on responsibility (b = 0.520, SE = 0.067, z = 7.728, p < .001), indirect effect of condition on responsibility ratings through counterfactual type was significant too (b = 0.160, SE = 0.032, z = 5.067, p < .001). These results suggest that the effect of condition on response is partially mediated by counterfactual type. Figure 1 shows that when participant mentioned counterfactuals related to perpetrator’s actions, they were more likely to ascribe responsibility, in comparison to when mentioning victim related counterfactuals. In Experiments 2 and 3, we explored the causal connection between certain counterfactual types and responsibility ascription. As a starting point for the counterfactual types, we used responses people provided in Experiment 1. We used only one of the ignorance conditions (High effort). In Experiment 2, we guided people to consider either what the agent could have done to get the relevant information (Perpetrator’s epistemic action), or how the information could have been more available (External epistemic factors). We make one the counterfactual alternative salient by presenting people with the unfinished sentence they need to fill in. For example, in the Perpetrator’s epistemic action condition this sentence was: "If only Ian ran to the bus so he could ask that woman if the money belongs to her, _____, and the woman would have been able to pay her bills that day." Relationship between the counterfactual condition and perpetrator’s responsibility ratings was tested by employing Cumulative link mixed models. The tested model included the main effect of the condition and random effect of the scenario nested on the condition. The model did not show a better fit of the data compared to the nullChi sq.(1) = 1.09, p =.29. In Experiment 3, we again guided people to consider Perpetrator’s epistemic action, but in the other condition we guided them to consider what the other agent in the story, or a victim, could have done differently. In this case, the model showed a better fit of the data compared to the null Chi sq.(1) = 6.47, p < .05. This means that guiding people to consider what perpetrators could have done differently to inform themselves led participants to more likely ascribe responsibility to the perpetrator, than when we guided them to consider how the victim could have acted differently to prevent the negative outcome. Our results show people consider different counterfactuals when agents produce harm, depending on what agents knew before acting. We also show it is possible to affect people’s responsibility assessment to an extent, by making perpetrators or victim’s actions salient.