We investigate how learning that an established type-level causal relationship is implemented by human agency affects people’s conceptualization of this relationship. In particular, we ask under what conditions subjects continue to perceive the original root cause as appropriate explanation for the resulting effect, and under what conditions they perceive the mediating intentional action as alternative explanation instead. Using a new experimental paradigm, we demonstrate that mechanisms involving intentional action lead to intuitions of causal intransitivity, but only when these actions are norm-violating. Potential generalizations and implications for scientific theory construction are discussed.

When A causes B and B causes C, under what conditions isA a good explanation for the occurrence of C? We proposethat distal causes are only perceived to be explanatory if thecausal mechanism is insensitive to inessential variations ofboundary conditions. In two experiments, subjects first ob-served deterministic A → B → C relationships in a single ex-emplar of an unknown kind. They judged A to be crucial forC by default. However, when they subsequently learned thatthe causal mechanism fails to generate the A → C dependencyin other exemplars of the same kind, subjects devalued A asa crucial explanation for C even within the first exemplar. Werelate these findings to the idea that good explanations pick outportable dependency relations, and that sensitive causes fail tomeet this requirement.

The communicative goal behind a causal claim like “Smok-ing causes heart attacks” is to inform recipients about the ex-istence of a causal link between the factors mentioned in theproposition. Different terminologies can be used to accomplishthis goal. Sometimes people use formulations of the form “Ccauses E”, like in the tobacco warning above, and sometimesthey use other formulations, such as modal propositions like“C can cause / lead to E.”, or statements like “C increases therisk of E.”. We investigate the hypothesis that different causalstructure claims, by means of different terminologies, not onlycommunicate the existence of a causal link but also implic-itly elicit intuitions about that link’s strength. Experiment 1revealed that claims like “C causes E” imply a stronger linkthan, for example, modal formulations like “C can cause E”.Experiment 2 tested implications of this finding for researchon causal structure learning.

Causal scope, the number of different effects a cause can pro-duce, is a salient feature of causes. In the present research, weaddress the question whether reasoners use causal scope as adiagnostic cue to infer the strengths of individual causal links.In three experiments, we manipulated the number of effects ofa cause, and asked subjects to assess the causal strengths ofsingle causal links. The results document a clear influence ofcausal scope on perceived link strength. In particular, subjectstended to display a “dilution” effect. They perceived a causallink to be weaker if that link belonged to a cause that is capa-ble of producing additional effects. This dilution effect can beexplained by a dispositional notion of causality according towhich a cause possesses a certain amount of causal power orcapacity that it distributes across its different causal pathways.

Queries about singular causation face two problems: It needsto be decided whether the two observed events are instanti-ations of a generic cause-effect relation. Second, causationneeds to be distinguished from coincidence. We propose acomputational model that addresses both questions. It accessesgeneric causal knowledge either on the individual or the grouplevel. Moreover, the model considers the possibility of a co-incidence by adopting Cheng and Novick’s (2005) power PCmeasure of causal responsibility. This measure delivers theconditional probability that a cause is causally responsible foran effect given that both events have occurred. To take uncer-tainty about both the causal structure and the parameters intoaccount we embedded the causal responsibility measure withinthe structure induction (SI) model developed by Meder et al.(2014). We report the results of three experiments that showthat the SI model better captures the data than the power PCmodel.

Causal queries about singular cases are ubiquitous, yet thequestion of how we assess whether a particular outcome wasactually caused by a specific potential cause turns out to bedifficult to answer. Relying on the causal power approach,Cheng and Novick (2005) proposed a model of causal attribu-tion intended to help answering this question. We challengethis model, both conceptually and empirically. The centralproblem of this model is that it treats the presence of sufficientcauses as necessarily causal in singular causation, and thus ne-glects that causes can be preempted in their efficacy. Also, themodel does not take into account that reasoners incorporateuncertainty about the underlying causal structure and strengthof causes when making causal inferences. We propose a newmeasure of causal attribution and embed it into our structure in-duction model of singular causation (SISC). Two experimentssupport the model.

Dependency theories of causal cognition, like causal Bayes net theory, postulate that the strength of causal links is independent of the causal structure in which they are embedded. We propose a new theory that postulates that people's concept of causality is richer, and that predicts an influence of causal structure on strength intuitions. According to the theory, people's concept of causality involves the idea that causal powers behave like phenomena studied in fluid dynamics: Continuous causes are assumed to spread their capacity across causal pathways, akin to fluids running through pipe systems. The theory predicts a perceived dilution of causal strength. A series of experiments (N = 3,733) and a meta-analysis corroborate the theory. For common causes, people perceive the strength of individual links to decrease with the number of links. In causal chains, people perceive a link strength gradient. This dilution effect disappears for genuinely binary causal variables.

Most psychological studies focused on how people reason about generative causation, in which a cause produces an effect. We study the prevention of effects both on the general and singular level. A general prevention query might ask how strongly a vaccine is expected to reduce the risk of contracting COVID-19, whereas a singular prevention query might ask whether the absence of COVID-19 in a vaccinated person actually resulted from this person's vaccination. We propose a computational model answering how knowledge about the general strength of a preventive cause can be used to assess whether a preventive link is instantiated in a singular case. We also discuss how psychological models of causal strength learning relate to mathematical models of vaccination efficacy used in medical research. The results of an experiment suggest that many, but not all people differentiate between preventive strength and singular prevention queries, in line with the formal model.

Singular causation queries require an assessment of whethera singular co-occurrence of two events c and e was causal orsimply coincidental. The current study builds on our previ-ous research (Stephan & Waldmann, 2018) in which we pro-posed a computational model of singular causation judgments.The model highlights that singular causation judgments needto take into account the power of the target cause C and ofalternative causes A, as well as the possibility of preemption.What was missing was a detailed model allowing us to esti-mate the probability of preemption of a target cause by thealternative causes. The present research fills this gap by elab-orating the temporal assumptions that might enter assessmentsof singular causation. We focus on assumptions about tempo-ral precedence between target and alternative causes, with aspecific focus on assumptions about causal latency. We reportthe results of two new experiments supporting the model.

Consider the following causal explanation: The ball wentthrough the goal because the defender didn’t block it. Thereare at least two problems with citing omissions as causal ex-planations. First, how do we choose the relevant candidateomission (e.g. why the defender and not the goalkeeper). Sec-ond, how do we determine what would have happened in therelevant counterfactual situation (i.e. maybe the shot wouldstill have gone through the goal even if it had been blocked).In this paper, we extend the counterfactual simulation model(CSM) of causal judgment (Gerstenberg, Goodman, Lagnado,& Tenenbaum, 2014) to handle the second problem. In two ex-periments, we show how people’s causal model of the situationaffects their causal judgments via influencing what counterfac-tuals they consider. Omissions are considered causes to theextent that the outcome in the relevant counterfactual situationwould have been different from what it actually was.