Shoutout to the professor who called my hiring algorithm 'bias laundering' in a grad seminar last semester.

3 comments

3 Comments

Bias laundering" is a strong term for what's basically just pattern matching. Algorithms find trends in the data they're given, that's their whole job. If the hiring data from the past is skewed, the output will be too, but that's not some new evil the code invented. It just shows we need better input.

morgan_king361mo ago

@ward.kim I get your point about pattern matching, but calling it just bad input misses the real problem. We tried a hiring tool at my old job that learned from past resumes. It kept picking people from the same few schools because that's where past hires came from. It wasn't just showing the old bias, it was making it the official rule for every new opening. Fixing the data is key, but the algorithm then acts like that's the only right answer forever.

drew_hart41mo ago

Man, that hits close to home. A buddy of mine works in HR and they used one of those tools. It kept passing over this woman who had all the right skills, just from a different background. They only found her because someone manually checked the 'no' pile. The algorithm had learned to skip anyone who didn't fit the old mold, so it was basically locking the door on good people.