The problem is that those thoughts… become actions… we can usually only see as systematic trends. Just because someone does not say “women are incompetent” does not mean they aren’t underestimating women’s competence and e.g. hiring them less than he should.
The flip side of it being hard to discern whether people have bad thoughts and act biasedly except by drawing inferences from broader patterns is that it’s also hard to discern whether people actually do have bad thoughts and acted biasedly from those broader patterns. (c.f. the many fields where women dominate men in terms of prevalence and performance, as well as EAs many other demographic biases which don’t receive the same treatment e.g. a 14:1 left-right bias, and a 4:1 20-35:any age over 35 bias).
The flip side of it being hard to discern whether people have bad thoughts and act biasedly except by drawing inferences from broader patterns is that it’s also hard to discern whether people actually do have bad thoughts and acted biasedly from those broader patterns. (c.f. the many fields where women dominate men in terms of prevalence and performance, as well as EAs many other demographic biases which don’t receive the same treatment e.g. a 14:1 left-right bias, and a 4:1 20-35:any age over 35 bias).