Expectations affect physical causation judgments

Abstract

When several causes contributed to an outcome, people often single out one as "the" cause. What explains this selection? Previous work has argued that people select abnormal events as causes, though recent work has shown that sometimes normal events are preferred over abnormal ones. Existing studies have relied on vignettes that commonly feature agents committing immoral acts. An important challenge to the thesis that norms permeate causal reasoning is that people's responses may merely reflect pragmatic or social reasoning rather than arising from causal cognition per se. We tested this hypothesis by asking whether the previously observed patterns of causal selection emerge in tasks that recruit participants' causal reasoning about physical systems. Strikingly, we found that the same patterns observed in vignette studies with intentional agents arise in visual animations of physical interactions. Our results demonstrate how deeply normative expectations affect causal cognition.

A counterfactual simulation model of causal judgments for physical events

“…For example, if participants learn that one ball generally prevents another from going through the gate, while another ball generally tends to make it go through, one could expect asymmetric judgments in a situation in which both balls jointly caused another one to go through the gate. It will also be interesting to see whether the effects of normality that have thus far been demonstrated in vignette studies will generalize to the domain considered here (see Gerstenberg & Icard, 2019, for evidence that this is the case). One of the major drawbacks of vignette studies is that much about what actually happened is left implicit, and different participants may fill in the gaps differently.…”

“…It doesn't say how good of a cause it was. To remedy this situation, structural equation models have been supplemented with considerations of normality (Halpern & Hitchcock, 2015) that accommodate the fact that we tend to favor abnormal over normal events as causes (Gerstenberg & Icard, 2019;Hilton & Slugoski, 1986;Hitchcock & Knobe, 2009;Icard, Kominsky, & Knobe, 2017;Kominsky et al, 2015;Samland & Waldmann, 2015). It has also been proposed that the causal responsibility of a variable increases the closer a situation was in which the variable would have made a difference to the outcome (Chockler & Halpern, 2004;Gerstenberg & Lagnado, 2010;Lagnado, Gerstenberg, & Zultan, 2013;Zultan, Gerstenberg, & Lagnado, 2012).…”

“…This way, one cannot be sure whether participants' causal judgments merely reflect demand effects, or whether they would have spontaneously sought out the relevant counterfactual information. Furthermore, it can be challenging to communicate certain scenarios verbally (Gerstenberg & Icard, 2019). For example, the situation of "double prevention" shown in Figure 3b is not easily described (particularly if one tried to do so without using counterfactuals), but the causal structure of the situation is readily apparent when viewing the clip.…”

“…Recently, it has been proposed that causal judgments further play the role of highlighting those aspects that not only made a difference in the actual situation, but that are also likely to continue to make a difference in other situations as well (Danks, 2013;Hitchcock, 2012;Lombrozo, 2010;Nagel & Stephan, 2016). A causal judgment now, may help to pinpoint a useful place for intervention in the future (Bramley, Gerstenberg, & Tenenbaum, 2016;Bramley, Mayrhofer, Gerstenberg, & Lagnado, 2017;Gerstenberg & Icard, 2019;Girotto, Legrenzi, & Rizzo, 1991;Meder et al, 2010).…”

When do we think that X caused Y?

“…The focal variable is 'drawing green' while the alternate variable is 'drawing blue'.Both the human data and the model exhibit a well-known effect of statistical normality on causal judgment in disjunctive structures. This effect is abnormal deflation : as "drawing green" (the focal variable) becomes less likely, causality ratings for "drawing green" decrease[1,7,21]. They also exhibit a trend, reverse supersession, that had not been identified prior to the study by Morris et al[2]: as "drawing blue" (the alternate variable) becomes less likely, causality ratings for "drawing green" increase.…”