| Sumario: | After all, the Four Utilities View is sensitive only to the utilities of the four possible trial outcomes and to the factfinder's credence that that defendant is responsible; thus, there is no obvious dependence (as there was for the Error Comparison View) on statistical facts about the distribution of responsible individuals in the population of individuals facing trial.[31] However, there are also several deep theoretical difficulties for the Four Utilities View, which arise as well for the Error Comparison View. The use of these tools in the contexts of policing, pretrial release determinations, sentencing, and parole is highly controversial; in particular, the use of these tools raises serious Due Process and Equal Protection concerns, especially as related to minority members of a population.[27] But almost nobody thinks that their use is appropriate during the guilt phase of a trial as evidence of a defendant's criminal conduct, or as evidence of a civil defendant's illegal conduct; even if we stipulate that the algorithm's output is statistically relevant to the defendant's conduct, the intuitive thought is again that the illegal conduct of other members of a defendant's groups should not (at least in general) influence the determination of the defendant's liability. And since this is so regardless of which party is the meritorious party, it is natural to expect the requirement of sworn testimony to reduce both the likelihood that (meritorious) non-responsible defendants will be found liable, and also the likelihood that (non-meritorious) responsible defendants will be found non-liable, and hence to lower both the FPL and the FNL. Moreover, even where a civil plaintiff does enjoy a large power or resource advantage over a particular defendant, it is not typically the case that well-resourced plaintiffs have the same incentives as a prosecutor's office to obtain settlements from large numbers of similarly-situated defendants.
|
|---|