#m=image%2Fgif&dim=220x162&blurhash=%23DG7f3RR01%25%7CP98%7BWB.7Myu6xHxZwc%23%2BJ-w%5D%24gRjB%3DKO%252m-rX-%3AaxRPxtGGs%3As%3AWVf*bcRkNaofAvxGt6NHaeS%7Dsor%5DkB%24*R%2BafS%24OE%24MX8S%23s9nOslXRSPWYsmbINxn%24&x=a5a8ee0af4588487f9f5a49520aad4baf35055ee8d59ac872abcee7c7b8d271fIf it rights your worldview, the median was a 6/7, which matches the AI rating.
There's a difference between median and range, and the more atypical she looks, the wider the range, but the median might stay the same.
Discussion
Very true, but it means there are other sources of variability, and the 95% confidence band will cover most of the chart. If you posted on a Sunday with no major sporting events, for example. Or right after a female celebrity said something ridiculous and widely publicised.
Yes, but that is always the case and can be solved for statistically.
They weighted the ratings of the different men, for instance, and threw out extreme outliers, and they didn't show any rating until a certain number of men had responded.
And there's the same effect IRL.
Deleting outliers is highly controversial. I do it too, professionally, but there are good arguments that it is introducing at least as much error as it removes. (Where this lies on balance depends on one's confidence in other stages of the workflow).
You are correct that this can be ameliorated statistically if we are only reporting the mean, but mostly only by increasing the sample size.
They're doing the best they can with the resources they have...
It seems to work, seeing as how the AI matches and most of my bros were okay with the median I got. 🤷♀️
I have a few superlative features and the rest is underwhelming. That's polarizing.
The AI is trained on the SAME data, does NOT fill me with confidence!
Your bros, though, are MUCH more credible.
Probably about right in your case, then, but I wouldn't recommend anyone take anonymous troll ratings to heart!
I'm relatively confident on the AI because it correlates closely with professional/trained raters. This is their business, so they try to do a good job.
Interesting is that they're all trying to figure out how to correctly account for eyes, but no AI has done it, yet. Humans still seem better at it.
Hence the value in "bro ratings".
LOL re eyes. Human males fail at that too, that's why mascara works! Interesting that AI is failing harder. I can haz link?