Sunday, January 08, 2017

Adventures in Toxicology

We can, I hope, all agree that sexual harassment poisons organizations. That is all the more reason to be accurate in our measurements.

To take the literal analogue, suppose you read a newspaper article that reported that researchers have found that over 75% of sampled tap water in your city contains toxins of a particular kind. Concerned, you track down the study and find that the real value is 57%. Looking more closely at a table in the report, however, you realize that this number is actually an adding mistake. Only 32% of the samples actually contained the toxin in question; all the other samples were completely free of it.

Looking still more closely, you realize that the study didn't just detect the presence or absence of the toxin but actually measured the concentration of the toxin and classified it according to health risk. The levels were "none" (68%), "low" (19%), "moderate" (11%) and "high" (2%). "Low" here means that, while trace amounts could be detected, no action needs to be taken to protect yourself from the toxin. "Moderate" means that something should be done to bring the level down (to "low") within a few months (i.e., only continuous exposure over many years constitutes a health risk) and "high" means that the water should not be used for human consumption unless boiled first.

Moreover, it turns out that the 2% of samples that contained a "high" concentration of the toxin were all localized to a particular neighborhood, suggesting a common source, and therefore a straightforward solution. It's even possible that the samples with lower concentrations are all "downstream" from this source will therefore have the same origin.

My point in making this comparison is to counter those who would dismiss my "pedantry" about the CSWA workplace climate survey as missing the larger point that people did, in fact, report harassment. What difference does it make, these critics say, whether it's 75% or 57% or 32% that report it? Or whether they report that it happens "rarely" or "often"? Surely we have to do something about the 2% who report experiencing verbal harassment often?

Yes, of course. But a report that over 75% of the tap water in the city is "poisoned" in some unspecified sense might well lead everyone to boil their water before using it. This constitutes not just a moderate inconvenience for individuals, but an enormous energy cost for the city as a whole. If the real solution is to find and extract one dead rat from a water tower somewhere, and only 2% of households actually needed to boil their water, then the exaggerated report will have wasted a lot of resources. These are resources that could have been used more effectively elsewhere.

Similarly, there seems to be a willingness on the part of administrators in the sciences to do something about sexual harassment. It is very important that they don't make their decisions on the basis of inaccurate measurements of the problem they are trying to fix. In the case of the CSWA survey, it is very possible (given what we know about the study) that the occasional behavior of a handful of bad actors accounts for virtually all of the reported harassment. We can't know for sure until the report is made public.

No comments: