Not so simple! Causal mechanisms increase preference for complex explanations

Zemla, Jeffrey C.; Sloman, Steven A.; Bechlivanidis, Christos; Lagnado, David A.

doi:10.1016/j.cognition.2023.105551

Cited by 5 publications

(1 citation statement)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One of their findings was that scenario wording influenced participants' assumptions about the conditional independence of the effects predicted by competing explanations, and thus the direction of explanatory simplicity effects. Similarly, work by Zemla and colleagues (see, e.g., Sloman, Zemla, Lagnado, Bechlivanidis, & Hemmatian, 2019;Zemla et al, 2017Zemla et al, , 2023 may be taken to suggest that the degree to which reasoners adhere to different explanatory virtues (e.g., simplicity or abstractness) also may depend on the pragmatic "purpose" (Sloman et al, 2019, p. 14) an explanation is assumed to have; they contrast the example of a policymaker who might favor more abstract and generalizable explanations with one of a private investigator who might demand as much detail about a case as possible). Future work may continue looking into this.…”

Section: Discussionmentioning

confidence: 75%

Revisiting the Narrow Latent Scope Bias in Explanatory Reasoning

Stephan

2023

Preprint

View full text Add to dashboard Cite

Humans are capable explainers and lay people tend to share the same explanatory virtues held in high regard by philosophers and scientists. However, a recent line of studies found a striking deviation from normativity in lay people's explanations, termed the "narrow latent scope bias". When competing explanations with identical a priori probabilities fit observed evidence equally well -- but differ in the number of unobserved pieces of evidence they predict (latent scope) -- reasoners seem to prefer explanations that predict fewer unobserved pieces of evidence (narrow latent scope). This tendency has been described as a robust explanatory reasoning bias. The present paper empirically demonstrates across six experiments (N = 2200) that this bias is less robust than has been claimed, and influenced by nuanced pragmatic inferences on the side of participants. Pragmatic factors shown to influence the bias are assumptions about how easily an unobserved piece of evidence should have been observed if it was present ("feature diagnosability"), and the formulation of the test question being asked. Across studies, genuine narrow latent scope biases resulting from fallacious reasoning were found only in a fraction of participants. It is also demonstrated that the magnitude of the bias depends on response options: it is stronger if participants are forced to commit an error, but at best weak if they are allowed to give the correct answer.

show abstract

Section: Discussionmentioning

confidence: 75%