This is the correct take. We're simply proving what we expected. And of course we don't know anything about why it chooses female over male, just that it does

This is the correct take. We're simply proving what we expected. And of course we don't know anything about why it chooses female over male, just that it does so very consistently. There are of course very subtle differences between male and female cognition, so the next hard experiment is to reveal if this LLM bias is truly seeing past the test or is a training problem. https://en.m.wikipedia.org/wiki/Sex_differences_in_cognition https://en.m.wikipedia.org/wiki/Sex_differences_in_cognition