Bias is persistent : sequencing case information does not protect against contextual bias in criminal risk assessment

Abstract

Purpose: A large body of research indicates that bias is an inherent part of human information processing. This way, bias affects all disciplines that rely on human judgements, such as forensic psychological assessment, including criminal risk evaluation. Although there is a lack of empirical studies, scholars recommend considering case information sequentially beginning with the most relevant information to reduce the effect of potentially biasing task-irrelevant contextual information.

Methods: We ran a preregistered experimental study to test, first, whether task-irrelevant information results in bias effects when people use criminal risk assessment tools, and second, whether such bias could be reduced by sequencing case information according to its prognostic relevance. We collected data of 308 informed lay participants instructed to apply an empirical actuarial risk scale based on a case vignette.

Results: Results showed that task-irrelevant information biased risk assessment. Yet, sequencing case information did not protect against it.

Conclusions: Considering various boundary conditions (e.g., overconfidence in the accuracy of one's own assessment and other sources of bias), we discuss challenges to mitigate the biasing effect of task-irrelevant information.