Tuesday, September 9, 2014

Fed Up with Sexual Harassment II: Information Escrows

Today’s guest blogger is Mordecai-Mark Mac Low. Mordecai is a curator in the Department of Astrophysics at the American Museum of Natural History, where he leads a research group studying the formation of planets and stars and the structure of the interstellar gas, and has curated the Space Shows "Journey to the Stars" (w/Rebecca Oppenheimer) and "Dark Universe."

After Eliot Rodger's rampage, the hashtags #notallmen and #yesallwomen swept Twitter, expressing the reality that although most men do not engage in sexual assaults or harassment, the ones that do tend to be serial offenders (e.g., sec 5.6 in this federal report from 1981, "Sexual Harassment in the Federal Workplace: Is It a Problem?", or Lucero et al. 2003, "An Empirical Investigation of Sexual Harassers: Toward a Perpetrator Typology" Hum. Rel., 56, 1461), ensuring that almost all women have had to deal with such problems at some level. (Although male victims may be slightly less common, and female perpetrators more so, both appear to suffer from even more overwhelming underreporting than the usual narrative.) 

However, in the adversarial system of justice in the US, each single victim must prove the circumstances of her (or his) harassment against resistance from the harasser.  As a result there are strong reasons for any individual victim to not make a complaint that might end up as a test of credibility against a perpetrator.  If other victims make follow-on claims confirming a pattern of abuse, though, the credibility of the case will be substantially enhanced.

Even if such follow-on claims appear, though, the first mover is likely to suffer stronger attacks, while the follow-on claims will suffer from accusations of being copycat accusations without a basis in fact. Furthermore, without the certainty of follow-on claims, the first mover cannot be sure of having the increased credibility that such claims would bring.

To simultaneously address both the first-mover disadvantage and the loss of credibility of follow-on claims, Ian Ayres & Cait Unkovic (2012, Michigan Law Review, 111, 145) propose applying the game theoretic concept of an information escrow to sexual harassment and assault cases. They argue that in this case, withholding information can sometimes lead to a more positive equilibrium state than immediate transmission. 

They suggest that the option be given to victims of filing a detailed accusation with an escrow agent who will hold the report in confidence, unless one or more other accusations are made against the same perpetrator, at which point the multiple allegations would be forwarded together to the appropriate university or police authority. This has the advantage that any response would start with effectively simultaneous, independent accusations, enhancing the credibility of the accusers, and at the same time diffusing their risk.

The immediate objection to this idea is that unmatched accusations are lost forever in the escrow, effectively giving perpetrators a free pass.  Therefore, one must consider whether the risk of accusations never being acted on outweighs the net gain in number (and quality) of accusations.  The less likely a victim would be willing to be the first mover in the absence of the escrow mechanism, and the larger the fraction of repeat offenders, the stronger the case is for escrows.  Ultimately, the benefit for the escrow depends on factors including how often perpetrators act, how many accusations would never have been made without the option of escrow, versus how many escrowed complaints would have been made directly if the escrow had not been available. Although documented false accusations remain rare, ensuring that the escrow system does not support them is also important to its credibility. One option outlined would be to allow people who believe they may be accused falsely to deposit defences into the escrow prophylactically.

Ayres & Unkovic analyze quantitatively whether more accusations will see the light of day if an escrow is available, using a spreadsheet model.  (They are unable to analyze the qualitative question of how much additional benefit can be gained from the greater credibility of matched accusations, though.)  They find that escrows enhance the number of accusations as the fraction of serial perpetrators increases, as the probability of direct complaints in the absence of the escrow declines, and as the number of victims only willing to make an accusation to escrow increases. However, for the fairly conservative estimates they make for the model parameters (based on asking colleagues and students for estimates), the result ends up being positive, but not hugely so.  More extreme parameter estimates appear equally likely to me, though, improving the case for setting up escrows.

Their spreadsheet model is available online (not, unfortunately, the URL given in the article).  This zip file contains Excel spreadsheets reproducing the three tables in the article, as well as one named "You Choose the Parameters" that is documented to allow the user to easily enter the model parameters (fraction of serial harassers and so forth) themselves.

Given the prevalence of underreporting of sexual harassment and assault, implementation of a credible escrow might improve deterrence, particularly among the frequent offenders who would be the most at risk of being confronted with multiple independent accusations, and likely do the most harm.

 For more information about this series:

Fed Up with Sexual Harassment II: The Solutions Series