Jeffrey Polzer at Harvard Business School posed a deceptively simple question in a case about hiring: If you had an algorithm that predicted reasonably well who is likely to succeed in a job, and you had a hiring manager, armed with those data, who nevertheless thought that another reasonably identical candidate with a lower score on the algorithm was better … which would you choose?
My experience posing this question to executives, confirmed by that of my colleagues, is that the overwhelmingly number of them want to go with the hiring manager’s preferences.
You might say, that’s not surprising; maybe they just don’t trust the algorithm. But, if you adjust the story to make its predictive power reasonably good, it doesn’t change the outcome. Nor does it help to explain how algorithms work. If instead, we switch up the story so that it is not an algorithm per se—which the participants may not completely understand—and instead just say that one candidate had higher scores on tests that predict performance, the results don’t seem to change, either.
Related: Minimizing pain points in hiring
You can tell by now that I have really been pushing to pin down what the explanation is here. When I embellish the account to make it clear that the manager’s preference for one of the candidates is based on his or her judgment, not from objective criteria, most of the participants still think we should go with the manager’s preference.
So I push harder and remind them about interviews and how bad most of us are at them, how we tend to prefer people like us and how the way we usually do them (“unstructured,” where each manager asks candidates whatever they want) just leads to bias. Yes, that makes the class uncomfortable, but it doesn’t sway their conclusion much. No matter what I do to the story, the manager’s preferences win out.
I can’t say I know for certain why that is, but I don’t think in the end people are willing to go with the manager’s view because they think it will lead to a better hire. You can stack the story in such a way that it is difficult to believe that could be the case. I believe the thinking is that we should let the hiring manager pick who will work for them, that it is something like their right to be able to have people they prefer work for them.
If this is the case, it seems troubling to me. Yes, we all might prefer to work with people we like, but is that really more important than hiring someone who might likely perform better in the job? I think for some people it seems to be that way. One explanation for this outcome might be that we think of the hiring decision as being more about the hiring manager than about the organization: We should give them what they want, as if this is something like picking out office furniture. You should get to pick out the one you like.
Maybe some of the willingness to let hiring managers choose whom they like also comes from the decentralization of business that has gone on over the past generation. Before then, recruiting departments actually made the hiring decisions. Line managers did not get any say in the process. New hires were seen as assets of the company, not of a division, and surely not of a particular office. Given that, it didn’t make much sense for hiring to be done at the local level.
It seems pretty clear to me that the power in hiring decisions has to go back to the center, that is, if we actually care about the quality of hires. A simple reason for that is because line managers don’t do enough hiring to get good at it, and we don’t train them on how to hire—so how could they get good at it? But doing so at this point is going to require overcoming a lot of resistance.
We started the process of pushing a lot of HR tasks off to line managers a few years ago as a way of cutting back on the costs of having large, centralized HR departments. It seemed cheaper just to add this task onto what operating managers were doing, and while they don’t like to do most of these tasks, do they do want to control whom they hire? Except they aren’t good at it, and that’s the problem.