by Claude Opus 5.5
Standards of proof distribute error costs. Given the respective harms of false positives and false negatives in campus adjudication, can the choice between preponderance and clear-and-convincing be settled on principle, or only politically?
The modern understanding of standards of proof owes much to decision theory. John Kaplan’s 1968 article “Decision Theory and the Factfinding Process” showed that a standard of proof can be represented as a probability threshold, and that the rational threshold depends on the relative disutility of the two kinds of error. If wrongly finding a respondent responsible (a false positive) is k times as bad as wrongly clearing him (a false negative), the decision-maker should find responsibility only when the probability of responsibility exceeds k/(k+1). Preponderance, interpreted as a threshold of just over 0.5, corresponds to treating the two errors as equally bad. Clear and convincing evidence, often glossed as roughly 0.7 to 0.75, corresponds to treating false positives as two or three times as bad.
On this analysis, the question in campus adjudication becomes: what is the ratio of the costs of the two errors? If it can be determined on principle, the standard can be derived. If it cannot, the choice is political.
Can the error ratio be fixed on principle?
Arguments for treating the errors as roughly equal emphasise symmetry between the parties. Both are students; both have interests in continued education; Title IX frames the issue as equal access to education. A false negative leaves a complainant sharing campus with someone who assaulted her; a false positive expels an innocent student. Each is grave, and neither party has a stronger claim.
Arguments for weighting false positives more heavily emphasise asymmetries. A finding of responsibility for sexual assault carries stigma that may follow the respondent permanently, affecting graduate admissions and employment, and is imposed by the institution, an exercise of power analogous, if not identical, to state punishment. The liberal tradition, expressed in Blackstone’s ratio, treats institutional condemnation of the innocent as a distinctive wrong, not merely a cost.
Arguments for weighting false negatives more heavily are rarer but exist: the harm of continued exposure to an assailant, the deterrent effect on future reporting, and the systemic consequences of under-enforcement in a domain where offending is common and convictions rare.
Why principle runs out
These arguments do not settle the ratio, for at least three reasons.
First, the costs are incommensurable. Comparing the harm of wrongful expulsion with the harm of continued exposure to an assailant requires a common metric that neither welfare nor rights theories straightforwardly supply.
Second, the costs vary across cases. The harm of a false positive depends on the sanction; the harm of a false negative depends on the risk the respondent poses. A single standard applied across all cases will be wrong for many of them. Some have proposed varying the standard with the sanction, which is principled but administratively complex.
Third, the expected error distribution depends on base rates and on the quality of procedures, not only on the threshold. Larry Laudan has argued that focusing on standards of proof neglects the much larger effects of evidential rules and investigative quality. A clear-and-convincing standard applied after a poor investigation may produce more errors of both kinds than preponderance applied after a good one.
Beyond decision theory
There are also reasons to doubt that the decision-theoretic frame captures everything relevant. Judith Jarvis Thomson’s work on individualised evidence suggests that what matters is not only the probability of responsibility but the kind of evidence that supports it. Procedural values, such as respect for the parties and the legitimacy of the process, may also bear on the standard independently of error costs. And the expressive dimension of a standard, what it communicates about how the institution regards each party, may matter as much as its consequences.
Is the choice therefore merely political?
Not merely. Principle constrains the range of defensible standards. It rules out beyond reasonable doubt as inappropriate to a non-penal process, and it rules out standards below preponderance as inconsistent with the seriousness of the finding. Within that range, the choice between preponderance and clear and convincing reflects contestable value judgements about the relative weight of harms. Such judgements are properly made through accountable processes, which is to say politically. The difficulty arises when political processes oscillate, as Title IX rules have between 2011, 2020 and 2024, so that the standard applied depends on electoral timing rather than on any settled judgement.
The honest conclusion is that principle narrows the choice, but politics completes it, and the legitimacy of the result depends on its stability and transparency as much as on its content.