Can Prediction Markets Reduce Sexual Harassment?

Recently the @litenitenoah twitter account observed that:

Prediction markets for sexual harassment accusations don't seem very ethical, but I bet they would successfully expose a lot of real harassers.

— Probably not Noah Smith but you never know (@latenitenoah) May 11, 2018

I'm not certain what ethical objections not-Noah has in mind, and suspect that I probably don't care about them. If prediction markets in sexual harassment (henceforth PMSHs) have the effect of reducing sexual harassment, then this is good and it will take a lot to convince me that the markets are overall not worth having.  (There used to be a prediction market in terrorism, which was shut down after outrage from politicians.) That said, it remains an open question as to whether or not PMSHs actually will have this effect.

After a couple of weeks of on-and-off thinking about this, I want to suggest that any PMSH will have both some specific advantages and some specific disadvantages. At present my fear is that the disadvantages win out; however,  The size of these effects will of course depend upon the precise way in which these prediction markets are implemented. One of my aims with this post, then, is to open up discussion about how exactly these markets can be designed so as to maximise the good and minimise the bad.

It is also worth stating, as a preliminary, a couple of limitations on all of this. Firstly, prediction markets are means of aggregating information, but they are not by themselves a means of governance. They can function as part of a government mechanism, as in Robin Hanson's futarchy, but only as a part. What this means is that while PMSHs may give us a reasonable idea of which men are abusers, it does not in itself provide a means towards actual trying men who may be guilty: any trial will require a concrete accusation from a concrete victim. This does not mean PMSHs can't reduce harassment, however, as we will shortly see.

Second, it is typically assumed in discussions of prediction markets that the existence of and odds given by markets do not affect the outcome being predicted. This may well not hold in this case - a victim might be emboldened to speak out against her harasser if the prediction market says he is probably a harasser, or might alternatively conclude that someone else is likely to come forward and there is no need to subject herself to examination in court. The fact that prediction market odds can affect the outcome is not by itself a problem - one might imagine a prediction market for individuals' health and life expectancies, with individuals buying bullishly on themselves so as to have a financial incentive to eat well and exercise - but it can cause problems, which we will discuss later.

Lastly before getting onto the ins and outs: I shall be proceeding on the assumption that prediction markets are basically efficient at aggregating information. If you disagree with this premise, please take that up elsewhere with Robin Hanson or someone, and accept it for the sake of argument in this post.

The case in favour

In my view, there are two large advantages which any PMSH would have, and two other advantages which PMSHs might have depending upon their design and size, and one other advantage whose size is difficult to gauge.

Firstly, there already exist informal whisper-networks, mostly though by no means entirely between women, about which men are not to be trusted or enabled. These networks can enable women to reduce their vulnerability to potential harassers, and can enable concerned third-parties to jump in to head off and stop harassment at an early opportunity. The effect of a prediction market would be to make this information, in an admittedly less-finely-detailed format, available to all concerned. Women should not have to change their behaviour to avoid being harassed, but since in some cases they can having access to PMSHs would give them a better idea of when this is necessary; concerned friends, similarly would be in a better position to know which men ought not to be left alone with young women for significant lengths of time, and which men really are harmless.

Second, harassers are frequently enabled by the institutions in which they work or serve. Larry Nassar, the former medic at Michigan State University and USA Athletics, was able to abuse over 300 women and girls because of silence surrounding his activities which had been going on since the 1990s. Such silences can only be maintained because institutions and the people within them have plausible deniability about whether they were truly aware of abuse going on. PMSHs would remove that deniability: having a high predicted odds of being accused of harassment would be an instant red flag that would make it much harder for institutions to engage in the kind of motivated ignorance which allows abuse to continue over extended periods of time.

An advantage which I think would be real, but can only speak for anecdotally, would accrue to men with prediction markets on their own odds of being accused of harassment. I do not wish to harass women; being of imperfect social intelligence, however, I frequently struggle to identify which behaviours will be taken as playful flirting or everyday platonic compliments, and which will be experienced as threatening by the women at whom they are directed. Of course I try to err on the side of safety, but I can hardly pretend that I have always succeeded here. Having an external evaluation of how threatening I am seen as would allow me to better calibrate my behaviour - was that girl giving of signs of distress that I didn't pick up on and the other guy did, or did he just want her to dance with him instead? Do I need to reduce the amount of alcohol I consume when going out on the town? Certainly I'm not alone in asking myself these questions - more than one male friend has expressed similar concerns in private conversations.

I wish to mention two other ways in which PMSHs might - might - serve to reduce sexual harassment. One of the biggest problems in tackling sexual abuse is that victims are, entirely reasonably, unwilling to publicly accuse their abusers because doing so will mean exposing deeply personal aspects of their lives to strangers. Whether you consider this to be the Patriarchy in action, an unfortunate but unavoidable consequence of having a well-functioning justice system, or a bit of both, this is the constraint within which we have to work. PMSHs would allow women to provide information about their abusers anonymously, by buying bets that the abuser will in fact be accused.

The advantage I am most doubtful about - and which I think a PMSH would ultimately have to jettison - is that it may provide some material compensation to women who do expose their abusers. A woman who has bought bets on the man who harassed her may stand to make money by actually going public, which may make her more likely to go public and/or may alleviate her loss of privacy, for example by allowing her to spend a while in a new location without running down her savings.

The case against PMSHs

There are two issues with this, however, which I suspect mean that a well-functioning PMSH would have to prevent women from financially benefiting by accusing men. Firstly, it is not clear that this incentive would only affect cases where abuse actually did occur. This may therefore create cases where men are falsely accused of harassment by women who want to make money out of the accusation.

This is unlikely to be an especially widespread problem - while false accusations of rape do occur, they are at most a small minority of actual accusations. That said, the prospect of such accusations means that there will be an obvious new brush with which genuine victims can be tarred - any man accused of harassing women may simply claim that his accusers are mercenaries trying to destroy his reputation for money. This will both create extra stress for genuine victims, and may lead courts to wrongly fail to convict a higher proportion of genuine abusers.

It is possible that we may come up with a way to prevent false accusers from financially benefiting from their accusations. Suffice it to say, however, that I have not yet thought up such a way, and this is my greatest worry as to why PMSHs may ultimately be unworkable or counterproductive.

A second major concern is that rich abusers may be able to cover up perceptions of their threat level by buying all bets on their being exposed. This is not the absolute worst possible scenario - it would at least mean that they would pay some price for their misdeeds - but it might allow them a pretence of harmlessness which the informal whisper-networks would have quickly dissipated. We all know stories of rich artists who have raped young women and got away with it; while it might be better that they were in prison, at least their reputations provide a warning to other young women who fall into their orbits. These men might be able to counteract or upend these reputations by betting financially on their not being accused.

There are other, smaller, objections, mostly of the form that PMSHs do not go far enough or are insufficient - that they would only take into account abuse of women with money, or that only men who are already in some way notable would have PMSHs surrounding them. These objections might well be correct, but they are not reasons to oppose PMSHs, merely to think that they must serve as part of a whole package of measures we might take to reduce abuse.

Conclusion

My current suspicion is that the disadvantages win out - that PMSHs might well, on balance, make it easier for men to get away with abuse. There are ways to combat this - for example, by preventing men from betting on their own behaviour, and by preventing people from both holding bets that a person will be accused and accusing that same person. If these are even achievable, however, they may undermine the advantages that are supposed to make PMSHs useful.

This should not be the final word. I would welcome any suggestions as to how PMSHs can be designed so as to avoid incentivising false accusations - and as importantly, to avoid giving the impression of incentivising false false accusations - and as to how they can prevent rich abusers from rigging their own reputations. But it seems clear to me that such suggestions are sorely needed before PMSHs can serve as a tool for making women safer.