This still happens today. The easiest way for a computer programmer to see how is to consider the problem of building a non-racially biased hiring algorithm that takes resumes as input.
Heuristic one: Pick candidates with first names that are correlated with high performance. This works because certain areas have better math education, and first names help filter on parents’ incomes.
It is obviously racist. Eliminate it.
Heuristic 2: Look at educational experience. Public schools in the US are horribly racist/classist, so this doesn’t work either.
Heuristic 3: Look at work experience. Other employers used heuristic 1 and 2, so this inherits racism from them. Also, older hires are (hopefully) less diverse, assuming the field is improving.
Heuristic 4: Objective performance based whiteboard coding. People with imposter syndrome are more likely to choke in high-stress environments. Scratch this.
Heuristic 5: “Cultural fit” bonus points to fix up #4. Studies show this selects for things like drinking the same brand alcohol or enjoying the same sports. Strike it too.
Heuristic 6: Tailor the interview for the specific candidate. Studies show that women/minorities tend to get “softball” questions like “how did you handle this”, which accidentally puts them on the defensive. Men get alpha male planning questions “how will you grow this company”.
I personally prefer #3 graded on a scale from 1-4 with 0.5 increments. I add 0.5 to nervous people’s (women, or english as second language, mostly) scores and truncate the integer.
On average, this corrects the bias inherent in the interview setup and in my monkey brain. It pisses off almost everyone I know though.
If I controlled recruiters, I’d add women / minority quotas to the candidate stream, but not to the interview phase. I would also track job performance vs. ethnicity/gender, and loosen standards for any group that was outperforming the team.