Lagnado, D. , Harris, A. & Cullen, V.
University College London
The extent to which people speculate from absent evidence is an important issue for legal theory and practice. It also presents challenges to psychological theories of causal reasoning. This paper proposes a Bayesian Network (BN) analysis of inference from the absence of evidence. We claim that the inferences people draw depend on their causal models of the case, and their explanations for the absence. Thus the same information about absence can be treated as incriminating, exonerating or neutral depending on which factors are considered as most likely explanations for that absence. An empirical study supported this analysis. Sixty participants were given an identical murder case, and saw the same incriminating evidence. They were all informed of potential eyewitnesses to the crime who were not presented at court. The reasons for this absence were manipulated in three between-subject conditions: participants received ‘incriminating’, ‘exonerating’ or neutral explanations. As predicted, judgments of guilt were modulated by the explanations given for the absence of eyewitnesses: judgments of guilt increased with incriminating reasons and decreased with exonerating reasons. Moreover, BN analyses based on participants’ verbal explanations matched their probability of guilt judgments. These findings have implications for psychological models of causal reasoning, and for legal decision making.