Causal Discounting Research Articles

When one model casts doubt on another: A levels-of-analysis approach to causal discounting.

Discounting is a phenomenon in causal reasoning in which the presence of one cause casts doubt on another. We provide a survey of the descriptive and formal models that attempt to explain the discounting process and summarize what current models do not account for and where room for improvement exists. We propose a levels-of-analysis framework organized around 2 types of models of causal discounting: computational and algorithmic models. Theories of causal discounting at the computational level attempt to provide normative, prescriptive explanations for discounting behavior, and they build on other normative frameworks like formal logic and probability theory. However, they tend not to focus on how those computations are carried out. Theories of discounting at the algorithmic level focus on the functions and representations from which discounting behavior emerges (i.e., they examine how problems are solved). We use this framework to identify gaps in the current literature and avenues for future model development.

Causal discounting in the presence of a stronger cue is due to bias

People use information about the covariation between a putative cause and an outcome to determine whether a causal relationship obtains. When there are two candidate causes and one is more strongly related to the effect than is the other, the influence of the second is underestimated. This phenomenon is called causal discounting. In two experiments, we adapted paradigms for studying causal learning in order to apply signal detection analysis to this phenomenon. We investigated whether the presence of a stronger alternative makes the task more difficult (indexed by differences in d') or whether people change the standard by which they assess causality (measured by beta). Our results indicate that the effect is due to bias.

Investigations in spontaneous discounting

Oppenheimer's (2004) demonstration that causal discounting (when the presence of one cause casts doubt on the presence of another) can happen spontaneously addressed the standing concern that discounting was an artifact of experimental demands, but these results could have resulted from memory inhibition. The present studies rule out this alternative using the same surname frequency estimation paradigm. In Study 1, individuals discounted surname familiarity even when it could be attributed to semantic meaning; in Study 2, participants under cognitive load discounted less; in Study 3, participants who were promised a prize for accuracy discounted more. All three results conform to a spontaneous causal discounting account better than to the inhibition alternative.

Belief-based and covariation-based cues affect causal discounting.

Causal discounting occurs when the perceived efficacy of a putative cause is reduced by the presence of a stronger causal candidate. Previous studies of causal discounting have defined the strength of causal candidates in terms of the degree to which the cause and the effect covary (e.g., Baker, Mercier, Vallee-Tourangeau, Frank, & Pan, 1993). In contrast, in the present study, causal strength was defined in terms of both covariation- and belief-based cues. Seventy-two participants made causality judgments for a fictional causal candidate both in isolation and when paired with either a stronger or a weaker cause. The results demonstrated that the degree to which a causal candidate is discounted depends not only on the degree to which an alternative cause covaries with the effect, but also on whether the alternative is a believable or unbelievable candidate. Indeed, it was observed that a highly believable alternative will produce the discounting effect, even if it is a weaker covariate than the original candidate. These findings suggest the need to incorporate both belief-based and covariation-based cues into models of causal attribution.