The felony fallacy

Abstract

This article is concerned with criminal trials where the defendant is a convicted felon with a prior conviction for the same offence. The article investigates a probabilistic argument that I will call the felony argument. According to the felony argument, the fact that the defendant has been previously convicted for the same offence has probative value in the present case, and increases the probability that the defendant is guilty. The article analyses the felony argument in Bayesian terms to pinpoint under what circumstances it is correct, and under what circumstances it comes to the wrong conclusion. This analysis shows that the felony argument is correct if the probability that a guilty defendant has been previously convicted for the same offence is higher than the probability that an innocent defendant has been previously convicted for the same offence. The problem with the felony argument is that the latter probability is not as low as one might think. It is not unlikely that an innocent defendant has been previously convicted for the same offence. In cases where the defendant is innocent, the cause behind the wrongful prosecution is often related to prior conviction, e.g. when photos of people with a prior conviction for the same offence are presented to an eyewitness and the witness mistakenly identifies a person who has nothing to do with the present case. Prior conviction for the same offence is used by the police as a criterion for selecting suspects, and this selection bias increases the probability that a defendant who is innocent has a prior conviction for the same offence. The defendant in a criminal trial is a person who belongs to a very special sample, the sample of people selected by the police as suspects and then selected for prosecution. An assessment of the probability that he/she is guilty, given a prior conviction for the same offence that does not take into account that he/she belongs to this sample, commits an error, that I call the felony fallacy. Language: en