Quantum of Proof in University Sexual Assault Investigations

When a university is deciding whether to expel, suspend, or otherwise discipline a student for an alleged sexual assault, how much proof should the university proceeding require? Should the student’s guilt be shown by “clear and convincing evidence”? By a “preponderance of the evidence,” which is what the Department of Education’s Office for Civil Rights has demanded, under its interpretation of Title IX? Beyond a reasonable doubt? I’m inclined to say — contrary to quite a few people whose judgment I generally much respect — that preponderance of the evidence would likely be the right standard, at least for claims of sexual assault and not just offensive words. Let me briefly explain why.

First, let’s think of preponderance of the evidence, clear and convincing evidence, and proof beyond a reasonable doubt as probability thresholds. Preponderance of the evidence means that the university should expel or discipline the accused student if there’s just a bit more than a 50% chance that the student is guilty. Clear and convincing evidence might be seen as requiring a 75% or 80% probability, or thereabouts. Proof beyond a reasonable doubt might be seen as requiring a 95% probability. These are oversimplifications, to be sure, but they are probably the most helpful way of looking at these standards.

And each of these probabilities might correspond to a number n in the statement that “better that n students guilty of sexual assault remain at the university, with no discipline imposed, than one innocent student be expelled or otherwise disciplined.” Proof beyond a reasonable doubt would be something like “better that ten students who have committed sexual assault remain at the university unpunished than one innocent student be expelled.” Proof by clear and convincing evidence would be something like “better that three guilty students (but no more) remain at the university than one innocent student be expelled.” Proof by a preponderance of the evidence would be something like “better that one guilty student (but no more) remain at the university than one innocent student be expelled.”

Of course, in criminal cases the standard is proof beyond a reasonable doubt. In criminal cases, a person’s liberty or life is usually at stake, so the legal system does say, “better than ten guilty people go free than one innocent person go to prison (or be executed).” But that this is true in criminal cases doesn’t tell us whether the same should be true when something less serious — both in terms of tangible loss and in terms of public condemnation — is at stake.

That’s why in civil cases, the more likely than not standard is routine. If you sue me, alleging I breached a contract or injured you in some way, you can prevail under this standard, even if the result is that I am financially ruined. That’s because we don’t say “better that 10 [or 5 or 2] culpable defendants are found not liable (and thus the plaintiffs are not compensated) than one nonculpable defendant be found liable.” Generally speaking, the risks of false positives and false negatives in civil litigation are treated very similarly, though we slightly err in favor of false negatives (since if the evidence is precisely balanced, defendant wins). There are exceptional cases in which “clear and convincing evidence” (often thought to mean something like a 75% or 80% probability of guilt) is required, for instance as to “actual malice” in libel cases, and as to punitive damages in most states; but those are unusual.

Likewise, where the issue is a government job, I expect that it’s quite routine for government employers to fire employees based on a 51% probability of wrongdoing. Would we say, for instance, “better that 10 corrupt/incompetent/dangerous government employees stay on the job than one innocent employee be fired”? Would we say, “If there isn’t enough to send someone to prison, there’s not enough evidence to fire him?” I doubt it.

Indeed, if, say, a postman is accused of rape or bribery or drunk driving, and we think there’s only a 66% chance that he’s guilty, I think we’d say that he shouldn’t be imprisoned (since there’s no proof beyond a reasonable doubt) and probably that he isn’t even guilty by clear or convincing evidence, but that he should be fired. (I set aside the question whether some civil service statutes or union contracts set a higher standard in some jurisdictions; I’m talking about what ought to be done, and what constitutionally may be done.) Indeed, in some jobs it might make sense for the employer to fire the employee based on a less than 50% probability that the employee is guilty, though we can set that aside for now.

The question, then, is whether we should analogize university disciplinary proceedings to (a) criminal trials, in which case we ought to require proof beyond a reasonable doubt, (b) ordinary civil liability or government employee dismissals, in which case a more-likely-than-not standard would suffice, or (c) punitive damages civil liability, in which case we ought to use clear and convincing evidence. That’s not an easy question, but my inclination is to opt for (b), which is what the Office for Civil Rights says.

Should we say “better that 10 rapists be allowed to stay at the university than one innocent student be expelled”? I don’t think so. “Better that 4 rapists be allowed to stay at the university than one innocent student be expelled” (the clear and convincing evidence standard)? I doubt it, partly because we don’t say this even when a lawsuit risking a defendant’s entire life savings is involved, or when the loss of a job and perhaps a career is involved. I don’t think we ought to so strongly favor protecting people from erroneous expulsions over protecting people from future sexual assaults (again, just as we don’t so strongly favor protecting people from erroneous firings over protecting customers or coworkers from future assaults, or even the employer from future thefts). “Better that one rapist be allowed to stay at the university than one innocent student be expelled, but not better that two rapists be allowed to stay” — i.e., “If we think there’s a 66% chance that the defendant is a rapist, that’s not enough to imprison him, but it is enough to decide that our campus is better off without him”? That’s probably about right, and that’s the more-likely-than-not standard that OCR promotes.

Of course, there are other criticisms of university disciplinary proceedings, such as that the adjudicators are biased, or that the defendant doesn’t have adequate opportunities to cross-examine the witnesses against him. Those are serious objections, but I think they warrant changing those aspects of the proceedings, rather than trying to compensate with a clear-and-convincing-evidence standard (which would be a highly over- and underinclusive compensation for those other problems). There’s also the question whether the OCR has the statutory authority to interpret the statute as requiring a particular quantum of proof; I set that aside here, though, since I don’t think this debate is about that.

The main problem with this approach, I think, is that it gives malicious complainants a great deal of power to badly harm classmates just by accusing them. A swearing match between two people, with no corroborating physical evidence or testimony from impartial witnesses, will often not lead to a conviction under a beyond-a-reasonable-doubt standard (though sometimes it does, and not just in sex crime cases). But it often would lead to a finding of guilt under a more-likely-than-not standard, especially given most adjudicators’ plausible assumption that the accused has more of a motive to lie in denying guilt than the accuser does in asserting guilt. This means that if even a few people are willing to falsely accuse someone, this could lead to a good deal of harm.

Yet while the risk of such false accusations is serious, so is the risk of false denials of real guilt. If someone who is “merely” 66% likely to be a rapist is allowed to stay on campus, you’re exposing the rapist’s classmates to a very serious danger. All the quanta of proof accept the possibility that there would be both false convictions and false acquittals; the question is how much of a risk of letting the guilty go free we should run in trying to prevent the innocent from being convicted. And where what is at stake is expulsion from university — a very serious matter, to be sure, but not materially more so than firing from an important job or loss of most of one’s assets — I think it’s reasonable to run a reasonably substantial risk of false conviction.

Finally, two more thoughts. First, none of this affects the criticisms of the Office for Civil Rights’ attempt to pressure universities into instituting speech codes aimed at restricting a broad range of speech that’s labeled “sexual harassment.” I’m speaking here of actual assaults, not just offensive words.

Second, it’s certainly possible to imagine different quanta of proof required depending on the gravity of the punishment (e.g., more likely than not for reprimands or even suspensions, clear and convincing evidence for expulsion) or on the gravity of the offense (e.g., more likely than not when someone is accused of a sex crime, clear and convincing evidence when someone is accused of minor vandalism or starting a not very dangerous fight). Generally speaking, that’s not the way our legal system has mostly developed within the area of civil liability or criminal liability, but such distinctions are in some measure present even there, and one can imagine similar distinctions in university administrative proceedings. My sense is that this isn’t what the debate is about right now, but maybe it should be.

UPDATE: Joe Cohn of the Foundation for Individual Rights in Education responds.