Bayesian Manipulation of Litigation Outcomes

TLDR: Guerra-Pujol et al. as mentioned in this paper presented a general Bayesian model of the litigation process and concluded that regardless of the operative rules of procedure and substantive legal doctrine, litigation outcomes [are] a highly reliable indicator of a defendant's [actual] guilt.

Abstract: In our previous work (Guerra-Pujol, 2011), we presented a general Bayesian model of the litigation process and concluded that "regardless of the operative rules of procedure and substantive legal doctrine, litigation outcomes [are] a highly reliable indicator of a defendant's [actual] guilt." By contrast, economists Eric Kamenica and Matthew Gentzkow recently presented a hypothetical example of Bayesian manipulation of litigation outcomes in their 2011 paper titled "Bayesian Persuasion." The remainder of our paper is thus organized as follows: In part one, we restate Kamenica and Gentzkow's example of Bayesian manipulation in litigation outcomes. Next, in part two, we evaluate their hypothetical example through a Bayesian lens and identify a crucial defect in their analysis: although the Sender in their model is required to tell the truth, he is not required to send an accurate or reliable signal. Thus, because the Sender can choose the level of accuracy or reliability of his signal, we would expect a rational judge to take this possibility into account by updating two separate probabilities -- not only the prior probability of the defendant’s guilt -- but also the prior probability that the prosecutor's signal is accurate or reliable. In part three, to illustrate our Bayesian analysis of the litigation process, we restate a thought-experiment that appears in Thomas Bayes' 1763 essay on inverse probability and draw an analogy between Bayes' 1763 thought-experiment and Kamenica and Gentzkow’s model. Part four concludes.