> To examine whether these demographic traits were driving the algorithm’s accuracy, Kosinski recomputed the analysis comparing only pairs of faces with the same gender, age, and ethnicity. Accuracy only dropped by around 3.5% — suggesting that many facial features beyond demographic traits were providing cues to political orientation. Crazy
CountChoculahh
Wow what if we automated phrenology
russ226
> Even modestly accurate predictions can have tremendous impact when applied to large populations in high stakes contexts, such as elections This is exactly the only use for such a system though.
x_interloper
What processing was done on images to ensure that the training set only contained faces? Did they, for example, remove pins or badges from clothes? Remove flags or patriot paraphernalia from the backgrounds? It would be funny if this was a repeat of the “our AI believes yellow rulers are a sign of lung cancer” story.
manicdee33
As someone who has opened a stats book, precision score is also important. Accuracy might just be meaningless.
cindimimim