> Yes, underrepresentation of the population is literally the definition of bias.
It's literally not.
Men account for 90% of convictions for homicide and other serious forms of violence in the US, and the overwhelming majority of countries abroad. Is that evidence of overwhelming misandry and bias in criminal justice? There's been demonstration of some bias, IIRC men tend to get 40-60% longer sentences. But that's nowhere near enough to account for this disparity. The reality is men do commit more violence than women and the disparity of men in prison for violence so this overrepresentation of men is not the product of bias, but rather men's own decisions.
Similarly, women make up the majority of professions like speech pathology, education, publishing with some of these over 90% female. Do we conclude that this is because of rampant bias against men? Perhaps, but in order to prove that we'd have to prove that these institutions prefer female applicants over male applicants. A non-bias explanation is that 90% of applicants are female, because 90% of people who want to go into these jobs are women.
Not only is underrepresentation not evidence of bias, the lack of underrepresentation does not absolve a group of bias. For instance, say a tech company decided it wanted to have an engineering workforce representative of the general population with respect to gender. 20% of the applications it gets are from women, so the company randomly throws away 75% of the male applications to make it so that an equal number of people go on to the phone screen, and subsequent on-site. This produces an engineering workforce that is perfectly representative of the population's gender distribution. But it is overwhelmingly biased against men and in favor of women. Obviously real-world tech companies are rarely this extreme - I've personally only seen OKRs calling for 33% at most. The point is, force equal or closer-to-equal outcomes in spite of unequal decisions made by people often entails the creation of bias, rather than its elimination. Do you really think that a company can achieve 50% women in tech without biasing their interview process in favor of women? The only company that even gets close to 50/50 practices explicit quota hiring:
https://en.wikipedia.org/wiki/ThoughtWorks#Hiring_and_divers...
Ultimately, at this point I think your conception that bias has a 1:1 relationship with representation relative to general population is fundamentally different from mine in a way that makes it difficult to productively explore the value of affirmative action in the hiring process.