Interrogation and Disclosure of Evidence

In recent years there has been a shift to less confrontational interrogation methods. One argument in favor of these methods is that they reduce the scope for false confessions. This paper explores this idea from an information perspective with rational agents. It studies a model that extends the plea-bargaining model of Baker and Mezzetti (2001) to one that is suitable for analyzing interrogation methods with evidence implying a continuum of probabilities that the suspect is found guilty at trial. In a benchmark model, in which police only interrogate suspects that the prosecutor is potentially willing to take to trial, the non-confrontational method weakly dominates the confrontational method. A similar result holds in a richer model, in which police may interrogate some suspects for whom the evidence will not induce the prosecutor to go to trial. The key intuition for why the non-confrontational method performs better in some cases in both settings is that the non-confrontational method is able to induce a confession from some guilty who the prosecutor would not take to trial following non-confession, and for some distributions this can result in a strictly higher expected payoff for society and the prosecutor.