Post #2640244
2026-04-23 20:22 UTC
OK, I am running some more iterations - I'm going to do 5 each for each name. It's not all as stark as the comparison in those first two! But I still feel that the vibe is more positive in John's.
To verify this, I'm going to run each of these feedbacks through a deterministic engine (not an LLM) so that it can produce a score of how positive the feedback sounds, and get some real stats in.
Replies (2)
-
@being@social.ornella.xyz 2026-04-23 21:32
QT: hold your horses. The first two looked real bad, but repeated iterations show this may be more about randomness than bias. (which is still an issue ofc) I used vaderSentiment (https://vadersentiment.readthedocs.io/en/latest/index.html) to do sentiment analysis and then pandas and scipy to check for statistical significance. So far, the sentiment difference is NOT statistically significant (sentiment is very slightly higher for Ornella in fact) I'll do a proper writeup since it sounds like folks are interested!
-
@feld@friedcheese.us 2026-04-23 22:18
@being@social.ornella.xyz curious how it responds to common gender-neutral names