Reasonable Doubt: Quantifying the Risk of Police Interrogation Tactics in False Confession Wrongful Convictions

Abstract

False confession wrongful convictions (FCWCs) remain among the most serious failures in the criminal justice system, yet efforts to quantify their prevalence have been hampered by methodological and inferential limitations. Existing estimates often rely on anecdotal case series, experiments of uncertain real-world relevance, or data that select on the very outcome under scrutiny. We develop a Bayesian inverse-probability framework to estimate the risk that currently lawful interrogation tactics produce FCWCs. Using observational and survey estimates of confession prevalence, conviction error rate, and the share of wrongful convictions involving false confessions, we construct three hierarchical models and simulate one million posterior draws. We embed these base-rate estimates in probabilistic scenarios incorporating plausible ranges of sensitivity and specificity, reflecting how well tactics identify the guilty while avoiding false positives. An acceptability curve then allows policymakers and scholars to evaluate interrogation practices against normative thresholds, such as Blackstone’s ratio or other harm-weighting assumptions. Findings suggest the estimated FCWC rate associated with current techniques clusters below one percent. This revised risk profile has implications for both policy and scholarly focus. Rather than settling the debate, our analysis reframes it: if risk is low, further reform may require new empirical evidence or clearer normative commitments.