The last few weeks I’ve been in Heidelberg as a visiting professor (if I had been Paulian Cornelis, I would have written a column on this wonderful word and other wonderful characteristics of the German language). I ended up in all kinds of meetings, like an interdisciplinary seminar Monday night on how to use randomness to better deal with uncertainty. One idea that came to my mind was that sometimes you can’t measure things accurately and drawing lots might be more fair than choosing based on measurements filled with uncertainty. But how do you do that in practice?
We have given an example to work on in groups: Suppose you can choose fifty candidates, but there are two thousand. How can you use arbitrariness in selection procedures? You can consider job applications, grant applications, or students who are enrolling for the desired study. Or something else where you only have fifty places and two thousand people who want that place.
She began working on a set with a historian, anthropologist, and computer scientist. The historian argued that drawing lots for important tasks was more natural, but that we came to see ourselves as ‘rational’ and arbitrariness was no longer appropriate. The computer scientist has argued that it makes sense to draw lots if the differences are too small to measure properly. The anthropologist wondered when lottery tickets are fair to people. I brought up an anecdote about the recruiter who gave half of the cover letters “because he didn’t want to hire people out of luck.” We unanimously believed that this was not a good measure. But how was it supposed to be?
In the end we came up with a combination of evaluation and sweepstakes. The committee had to divide two thousand characters into three categories: a small group of excellent candidates were “definitely accepted”, a middle group received “maybe yes, maybe no” and a third group withdrew from the judgment “definitely not”. In the middle group, we wanted to make a draw for the remaining places, but how are we going to do that? Did everyone get the same chance? Or did we want a weighted draw in which the better candidates stand a better chance? The anthropologist began to look into the glass. The computer scientist suggested dividing the middle group randomly into a number of groups where there are still empty places and then choosing the best for each group. Then both the historian and the anthropologist seemed glassy.
Back at the seminar, I was surprised that the other groups came up with very different solutions. For example, one club had the idea of withdrawing the criteria on the basis of which candidates could be judged among the members of the committee. Then, one of the committee members had to evaluate all the proposals related to this single criterion, and then the evaluations were combined in the classical way. One economist objected that this was too ineffective because members of the committee still had to carefully evaluate all the candidates. A computer scientist asked if you didn’t retain the measurement uncertainty problem that you wanted to solve with a lottery item.
We couldn’t find a solution in this seminar, but even more from the different ideas about lottery drawings, it took me to the Netherlands how good it is to have these kinds of discussions with – whether chosen at random or not – experts from different backgrounds.
