25
That moment I realized I was training AI on biased data without even knowing it
I was building a resume screening tool for a small HR company in Portland... using my own hiring history as the training set. About 3 months in, a friend pointed out my dataset had like 80% male candidates from the same two universities. I never even thought about it, I just grabbed what was easy. Has anyone else accidentally baked their own blind spots into a model like that?
3 comments
Log in to join the discussion
Log In3 Comments
margaret_taylor4212d agoMost Upvoted
Used to think bias didnt matter much. Thought hey the data is what it is. Then I read a post from @lisak26 about how it can ruin a tool and make it useless for real people. That hit home. Now I check everything twice and pull from way more sources. It's honestly embarrassing how blind I was.
10
sagew501mo ago
Still worked well enough for what it was doing, though.
8
lisak261mo ago
It did the job, but barely. Loading times were brutal and it crashed at least three times for me during a single session. The interface felt stiff and unresponsive, like it was designed for a computer from 2005. Just because something technically works doesn't mean it's acceptable. Have you tried any other options that run smoother?
3