|Event Starts:||Thursday, 21st January, 2016 3:15pm|
|Event Ends:||Thursday, 21st January, 2016 5:00pm|
|Location:||Neil MacCormick Room|
Legal probabilism, when applied to criminal trials, is the view that a high probability of guilt is legally sufficient to convict (Kaye, 1986; Tillers and Gottfried, 2007). This is a plausible view. Since the certainty of guilt is unattainable, one can argue that a high probability should be enough to convict, for if it is not, conviction itself would be unattainable. Certain hypothetical scenarios, however, present situations in which the probability of the defendant’s guilt is high, and yet a conviction seems inappropriate (Cohen, 1977; Nesson, 1979; Thomson, 1986; Stein, 2005; Red- mayne, 2008; Enoch et al., 2012). This has sparked a debate between the legal probabilists and their opponents. Here I articulate a hybrid account of the standard of proof according to which a high probability of guilt is a necessary but not a sufficient condition for a conviction. This account should appeal to both sides of the debate.
I begin by considering two goals of the trial system. The first is to ensure that as many guilty defendants as possible are convicted and that as many innocent defendants as possible are acquit- ted. In other words, the first goal is to keep decisional errors as low as possible (call this overall error reduction). Further, given the inevitability of some errors, the second goal is to ensure that er- rors are distributed in a socially desirable way. This typically means having proportionally fewer wrongful convictions even at the cost of more wrongful acquittals (call this desirable error distribu- tion). Interestingly enough, a high threshold probability, such as >0.99, promotes a desirable error distribution but does not serve well the goal of error reduction. In fact, error reduction is better served by a lower threshold, such as >0.5 (Kaye, 1982; Dekay, 1996). This has lead the legal prob- abilists to think that error reduction and a desirable error distribution are incompatible goals. As a result, their account assigns to the criminal standard of proof the function of distributing errors but neglects error reduction.
By drawing from Signal Detection Theory (Wickens, 2002), I propose an account of the crim- inal standard of proof in which both error reduction and error distribution are accommodated. The account consists of three requirements: (1) high probability threshold; (2) completeness of the evidence; (3) resiliency of the prosecutor’s case. Requirement (1) is in line with legal prob- abilism; (2) develops a suggestion by Kaye (1986); and (3) applies to criminal trials the idea of ‘resilient probability’ by Skyrms (1980) and Leitgeb (2013). I show that (1) promotes a desirable error distribution, while (2) and (3) promote error reduction. So, the three requirements, taken together, constitute an account of the standard of proof in which both error distribution and error reduction are accommodated.
My threefold account is conciliatory in nature. For one, requirements (1), (2) and (3) develop insights that come from the applications of probability theory to epistemology and law, so they should appeal to the legal probabilists themselves. On the other hand, the account offers an explanation for why a high threshold probability is not enough to convict—an explanation that is normatively grounded in the trial goals of error distribution and error reduction.
Discussant: Professor Duncan Pritchard