Causal responsibility and counterfactuals

Cognition, 2012

Attributions of responsibility play a critical role in many group interactions. This paper explores the role of causal and counterfactual reasoning in blame attributions in groups. We develop a general framework that builds on the notion of pivotality: an agent is pivotal if she could have changed the group outcome by acting differently. In three experiments we test successive refinements of this notion – whether an agent is pivotal in close possible situations and the number of paths to achieve pivotality. In order to discriminate between potential models, we introduced group tasks with asymmetric structures. Some group members were complements (for the two to contribute to the group outcome it was necessary that both succeed) whereas others were substitutes (for the two to contribute to the group outcome it was sufficient that one succeeds). Across all three experiments we found that people’s attributions were sensitive to the number of paths to pivotality. In particular, an agent incurred more blame for a team loss in the presence of a successful complementary peer than in the presence of a successful substitute.

Psychology and Marketing, 1997

The role of counterfactual reasoning and the assignment of blame within a context involving product failure, personal injury, and luck was examined. In the first study, it was determined that directing attention to the focal individual in an event increased the perceived mutability of that individual's actions in an exceptional circumstance but not in a common circumstance. Study 2 explored how the availability of different counterfactual alternatives influenced assignments of blame for a negative outcome. The results of this experiment showed that presenting information that directed attention to the focal individual increased the mutability of that individual's actions, which in turn, increased the blame observers assigned to that individual. However, this assertion needs to be qualified; when attention was already focused on the focal individual, directing further attention to that person seemed to have little additional impact.

Journal of Artificial Intelligence Research, 2004

Causality is typically treated an all-or-nothing concept; either A is a cause of B or it is not. We extend the definition of causality introduced by Halpern and Pearl (2004a) to take into account the degree of responsibility of A for B. For example, if someone wins an election 11-0, then each person who votes for him is less responsible for the victory than if he had won 6-5. We then define a notion of degree of blame, which takes into account an agent's epistemic state. Roughly speaking, the degree of blame of A for B is the expected degree of responsibility of A for B, taken over the epistemic state of an agent.

How much are people’s responsibility attributions affected by intended versus actual contributions in group contexts? A novel experimental-game paradigm dissociated intended from actual contributions: good intentions could result in bad outcomes and bad intentions in good ones. Participants acted as external judges and attributed responsibility to computer players engaging in a repeated game. On each round, three players formed a group and each chose to roll one of three dice that differed in terms of price and probability distribution. The team won if the sum exceeded a certain threshold. The results showed that both intended contribution, reflected in the choice of die, and actual contribution, reflected in the outcome of rolling the die, were determinants of participants’ responsibility attributions. However, contrary to previous evidence (Cushman, Dreber, Wang & Costa, 2009), more participants based their attributions on the intention rather than the outcome.

Legal and prescriptive theories of blame generally propose that judgments about an actor's mental state (e.g., her knowledge or intent) should remain separate from judgments about whether the actor caused an outcome. Three experiments, however, show that, even in the absence of intent or immorality, actors who have knowledge relevant to a potential outcome will be rated as more causal of that outcome than their ignorant counterparts, even when their actions were identical. Additional analysis revealed that this effect was mediated by counterfactual thinking-that is, by imagining ways the outcome could have been prevented. Specifically, when actors had knowledge, participants generated more counterfactuals about ways the outcome could have been different that the actor could control, which in turn increased causal assignment to the actor. These results are consistent with the Crediting Causality Model , but conflict with some legal and moral theories of blame.

2012

Abstract Social causality is the inference an entity makes about the social behavior of other entities and self. Besides physical cause and effect, social causality involves reasoning about epistemic states of agents and coercive circumstances. Based on such inference, responsibility judgment is the process whereby one singles out individuals to assign responsibility, credit or blame for multi-agent activities.

In order to be held responsible, a person’s action has to have made some sort of difference to the outcome. In this paper, we propose a counterfactual replacement model according to which people attribute responsibility by comparing their prior expectation about how an agent was going to act in a given situation, with their posterior expectation after having observed the agent’s action. The model predicts blame if the posterior expectation is worse than the prior expectation and credit if it is better. In a novel experiment, we manipulate people’s prior expectations by changing the framing of a structurally isomorphic task. As predicted by our counterfactual replacement model, people’s prior expectations significantly influenced their responsibility attributions. We also show how our model can capture Johnson and Rips’s (2013) findings that an agent is attributed less responsibility for bringing about a positive outcome when their action was suboptimal rather than optimal.

How do people assign responsibility to individuals in a group context? Participants played a repeated trial experimental game with three computer players, in which they counted triangles presented in complex diagrams. Three between-subject conditions differed in how the group outcome was computed from the individual players’ answers. After each round, participants assigned responsibility for the outcome to each player. The results showed that participants’ assignments varied between conditions, and were sensitive to the function that translated individual contributions into the group outcome. The predictions of different cognitive models of attribution were tested, and the Structural Model (Chockler & Halpern, 2004) predicted the data best.

Social Cognition, 1985

This paper introduces a new family of cases where agents are jointly morally responsible for outcomes over which they have no individual control, a family that resists standard ways of understanding outcome responsibility. First, the agents in these cases do not individually facilitate the outcomes and would not seem individually responsible for them if the other agents were replaced by non-agential causes. This undermines attempts to understand joint responsibility as overlapping individual responsibility; the responsibility in question is essentially joint. Second, the agents involved in these cases are not aware of each other's existence and do not form a social group. This undermines attempts to understand joint responsibility in terms of actual or possible joint action or joint intentions, or in terms of other social ties. Instead, it is argued that intuitions about joint responsibility are best understood given the Explanation Hypothesis, according to which a group of agents are seen as jointly responsible for outcomes that are suitably explained by their motivational structures: something bad happened because they didn’t care enough; something good happened because their dedication was extraordinary. One important consequence of the proposed account is that responsi-bility for outcomes of collective action is a deeply normative matter.