[Draft posted at 12:08. Final version at at 15:40.]
severity[seriousness] of the problem of sexual harassment in science requires an understanding of both the range[severity] of the offending behavior and the prevalence of that behavior.*** Individual case histories can inform the first, while surveys and other data can inform the second. In both cases, it is essential that we interpret the facts carefully in order to get an accurate sense of what is going on.
I've recently had two lengthy exchanges on Twitter that have left me a bit despondent about the possibility of such accuracy. The first was with Grant, who had suggested that Jason Lieb was being justly punished by the court of public opinion for his "pretty rapey" behavior.
He was, of course, referring to the widely quoted remark in Amy Harmon's New York Times story that a University of Chicago investigation had determined that Lieb had engaged in "sexual activities" with a student who was "incapacitated due to alcohol and therefore could not consent." That sentence seemed to me to be carefully crafted to invoke a definition of sexual assault that is often used in Title IX cases without making any allegations that a crime took place (so that the police might be a more relevant authority). Indeed, given the way the story of Lieb's behavior is being told, it's not inconceivable that the woman in question did not feel violated at all, only that witnesses to the behavior felt uncomfortable with what they saw happening between her and Lieb.
I don't know what happened, of course. Nor does Grant. But suppose, for the sake of argument, that it turned out that I'm more or less right. Let's suppose, that is, that Lieb got very drunk at a party and was seen making out with a graduate student who was also very drunk. Though embarrassed, let's imagine, she did not feel he had failed to obtain consent and let's say she filed no complaint against him. She was willing to take responsibility for her part in the affair, let's say. A number of witnesses, however, found his behavior unseemly, and reported it to the Title IX office, whose coordinator determined that the young woman "could not consent," even if she herself says it's no big deal. This sort of situation is not at all a stretch in Title IX cases, as far as I can tell.
Now, suppose I confronted Grant with this version of events. (Remember, we're imagining that this is what the investigation actually found.) "Surely," I say, "it is unfair to describe what happened as an assault and Lieb as a rapist?" And now suppose Grant says, "Yes, but it's still unacceptable behavior." Indeed, it probably would be. It's entirely fair for an institution to have rules against getting drunk and making out with students, even against getting drunk with students in the first place. And such behavior could even be grounds for dismissal. But that does not make it rape. In this and other cases, there seems to be a presumption that if someone does something "unacceptable" or "inappropriate" then you can call them whatever bad names you like.
And here's an important additional point. Since the story I've made up is entirely consistent with what the NYTimes article says the investigation found, we actually don't have a very good reason to think of Lieb as a perpetrator of sexual assault at this point. Certainly, there is what a court of law would call "reasonable doubt". And that's why I refuse, for now, to describe him as a rapist (or even a "pretty rapey" guy, which I think is a hideous phrase on many levels anyway.)
In the court of public opinion, however, different standards apply. The NYTimes article (and probably the letter that Amy Harmon had obtained) was designed to elicit precisely this sort of verdict. After publishing that sentence, Grant's "pretty rapey" interpretation was almost certain to follow in social media, as were posts confidently asserting that "Jason Lieb is now publicly exposed as a sexual predator." Those who wanted to shame Lieb could depend on someone running with the insinuation. It's a powerful way of making the problem seem severe enough to warrant institutional action, as in Grant's memorable suggestion that, "We're not punishing the professors who sleep with their students; we're punishing the ones who rape them."
A similar effect can be produced when it comes to assessing the prevalence of sexual harassment. Here the relevant distinction is not between a court of law and the court of public opinion, but between a scientific study and a piece of popular science writing. This became clear to me last night (forcing me to vent with this post, actually.) I've told part of this story before, but it's worth thinking about some more.
In early January, Christina Richey presented the results of a survey sponsored by the Committee for the Status of Women in Astronomy at the annual meeting of the American Astronomical Society. Her talk appears to have been partly a presentation of research findings and partly a call to action, brought together in the strong claim that "We have a problem."** There are lots of issues with the presentation, but I want to focus on just one: the prevalence of gender-based verbal harassment that was reported by the respondents.
Slide 8 tells us that 57% reported that they had been verbally harassed based on gender characteristics.** This figure was quickly reported (and Tweeted) by Miriam Kramer at Mashable. Strangely, it was also reported by Ethan Siegel as "greater than 75%" in Forbes (which was then Tweeted with fitting exclamation marks by Vanessa Janek.) I was shocked enough at the >75% that I contacted Siegel through Twitter and, ultimately, Richey by email to see if that could be right. Sure enough, it wasn't, and Siegel's article was corrected to the 57% that Kramer had reported, citing Richey and her slides. Strangely, Siegel refused to thank me for pointing out the error, and Richey has still not responded to my mail.
When @ticobas looked closely at the relevant slide, however, he noticed that the numbers didn't add up to 57%. It was more likely to be somewhere around 33%. I wrote to Richey by email again to ask about this, and again heard nothing. (Indeed, I'm blocked on Twitter by both Richey and her co-author Kathryn Clancy, for no interaction other than my questions pertaining to this study.) Then I noticed that Michael Brown was circulating the slide on Twitter under the #AstroSH hashtag, and I pointed out that it probably contained an error. To his credit, Michael did thank me. But what happened next says a great deal about how partisan this discussion is.
Michael contacted "the authors" (presumably Richey and Clancy) and was told that the slide does indeed contain an error. They did not, it seems, tell him what exactly the error consists in, nor what the right (i.e., corrected) figure would therefore be, but they assured him that in the final, published version all would be set right. In the meantime, I guess we're free to think the figure is anywhere between 1/3 and 3/4's.*
I find this very frustrating. After making the effort to find and point out an error in their work, which may well have gone unnoticed before being published if we hadn't, @ticobas and I have been cut completely out of the loop. And not even an "ally" like Michael has been told what the correct figure will be after bringing the problem to their attention.
Michael doesn't think this is a big deal. "Regardless if the number is 20%, 40%, 60% or 80%," he says, "it is still unacceptably high." This is a bit like saying "Even if it wasn't rape, it's still unacceptable behavior."**** When I asked what the point of doing a survey is, if you're going to judge 20% to be indistinguishably "unacceptable" from 80%, he said something very plausible, but more damning of Richey and Clancy than I think he intended. "The survey," he said, "provides an estimate of the scale of the harassment problem, and some people won't act without such data."
It's altogether possible that the CSWA survey's main purpose is to generate "data" in order to force people to "act". It is not intended to actually gauge the severity of the problem, which no one seriously doubts exists. We can see this by breaking down the 33% that it now seems likely actually represents the proportion of respondents that had experienced verbal harassment at all: 19% had experienced it "rarely" and 11% experienced it "sometimes". (The survey asked respondents to focus on the past five years.) Only about 2% had experienced verbal harassment frequently (like I say, at some point during the past five years). And we don't even really know what constitutes "verbal harassment" to the respondents. Perhaps they're as quick to call someone's drunken indiscretions "rapey" as Grant?
In Lieb's case, his "sexual activity" is turned into a "sexual assault" by playing a carefully constructed phrase into the hands of the media. In Richey's case, the idea is the same: carefully construct a number for maximum rhetorical impact, and count on the press not to break it down into its less dramatic components. And count on them not to look at the slide closely enough to spot an obvious error too. If an adding mistake or a typo happens to inflate the number by over 20%, after all, that's just gravy! Now you just have to wait until it's been disseminated far and wide before publishing the correct number. The important thing, it seems, is to get people to "act", not to help them develop an accurate perception of the problem.
*This is almost funny. About an hour after I tweeted the final version of this post, Miriam Kramer added the following to her original 57% tweet:
This number is actually 32% due to a numerical error reported by the survey's authors. Thanks to @PlanSciCRichey for correction.— Miriam Kramer (@mirikramer) February 9, 2016
As in the case of Siegel. No acknowledgement of my work in bringing this to light. No thanks. Doesn't even tag me in.
**Update (10/02/16): I just noticed that Richey has provided a new set of slides at the previous link, correcting the 57% error, and removing the slide that says "We have a problem." [I've discussed this change here.]
***Michael Brown found my use of the word "severity" confusing here. I see his point. Severity and prevalence are normally considered two dimensions of behavior, which are used to determine whether or not it counts as harassment. I was talking about the severity of the general problem, and the prevalence of the behavior in the community. I've changed this sentence so that it now applies severity and prevalence (in the community) to the behavior alone, in an attempt to gauge the "seriousness" of the problem.
****Apparently some people (Michael and Sue) think this is an "awful" "rape analogy". Please note that it is a reference back to my critique of Grant's use of "pretty rapey" above, which I characterized as "a hideous phrase on many levels". I'm not comparing rape to anything, nor "confusing severity and incidence". I'm comparing (Grant's) imprecise talk about severity with (Michael's) imprecise talk about incidence.