I figure you’d audit it by examining the results, and if bias isn’t detectable in the results

I was just asking for more info on how you’d examine the results for bias since it would need to be pretty extreme like the example you gave to be identifiable as something worth investigating.

All good if you don’t have a more specific answer, I was just curious what your personal thoughts were here.

Right, that’s why I was trying to ask you for your opinion on what the threshold for “investigation worthy” results

That would be a pretty extreme bias to have, so yeah that would make sense. If it’s not so drastic it might be harder to spot by just looking at the results.

What would unbiased results look like?

Looks similar to Confluence but more of a pain to get set up.

Can you share the lemmy communities?