I don’t think its confidence will necessarily match the degree to which a trans person has transitioned.

But indeed it would be interesting to see how trans people are treated at all.

It depends on what you mean, as that community has a pretty vast and complex taxonomy.

Estrogen levels affect the eyes, so given enough samples and context it may be surprising accurate for people getting that type of therapy.

I think the problem with that line of thinking is that even after fully transitioning, trans people still tend to have strong biological markers of the sex they've transitioned from.

For a ML model to correctly guess gender identity, it'd need other cues that indicate a person would prefer to be referred to by a certain pronoun, such as clothing or facial hair.

It really depends on the bio marker. Most bio markers are effected: fat distribution, hair texture (curly vs. straight), and even immune system function. I’d think that learning whether this model works on people who’ve taken hormones for gender transition could give some insight into what it’s detecting. Perhaps there are retinal features whose function or form are impacted by the presence of testosterone, for instance, or maybe it’s a different that forms before birth.