Large scale statistical analysis of words used in performance reviews is, to put it in the article's words, very unactionable feedback. How could you go about r

Large scale statistical analysis of words used in performance reviews is, to put it in the article's words, very unactionable feedback. How could you go about resolving this? The impression you get from a person is a very difficult thing to change. If you were to treat each race and group differently, being mindful of this data when writing reviews you will probably end up with a system that is more racist than the one you started with. It is noted, for instance, that Asians received a higher incidence of "genius" feedback; they also have a higher IQ on average. Would it be more or less racist not to tell a gifted person that they are gifted based on their race? Certainly there is a stereotype that may have biased you, but there is a truth beneath that stereotype. It forms a very complex feedback loop, where a reality creates a self-enforcing perception. The mind is pattern seeking, but once it sees a pattern it will start to find it in places it wasn't present, and use these incidences to justify the pattern as a whole. If you just try not to mention these things in reviews, you haven't solved the problem. You just aren't telling the person about it to their face. I would rather hear someone's biased opinion of me, than have them hold it behind my back. The most pressing issue to me is that there is no way to detect which of these things are genuine insights and which are biases. The only way would be a control group who don't know each other's race. You could partially achieve this through online interactions, but the less of a person you experience, the less you can learn what they are like. The only way to hold a truly unbiased opinion is to never have talked to someone or seen them. A review being potentially accurate is inextricably linked with it being potentially biased. Fundamentally, you can't bring your perception of your coworkers in line with reality given that: 1 — you don't know what reality is; 2 — you can only discern reality by perceiving it, yet your flawed perception is the issue you're trying to solve; and 3 — your perception of your perception is also subject to biases.