Hypervisible: “Research has shown that some emotional AIs disproportionately attribute negative emotions to the faces of black people, which would have clear and worrying implications if deployed in areas such as recruitment, performance evaluations, medical diagnostics or policing.”

“Research has shown that some emotional AIs disproportionately attribute negative emotions to the faces of black people, which would have clear and worrying implications if deployed in areas such as recruitment, performance evaluations, medical diagnostics or policing.”

Are you 80% angry and 2% sad? Why ‘emotional AI’ is fraught with problemswww.theguardian.com AI that purports to read our feelings may enhance user experience but concerns over misuse and bias mean the field is fraught with potential dangers

Dr. Rhue's 2018 paper they link to -- using basketball players’ pictures -- is a classic. phys.org/news/2019-01... is her excellent blog post on it.

Emotion-reading tech fails the racial bias testphys.org Facial recognition technology has progressed to point where it now interprets emotions in facial expressions. This type of analysis is increasingly used in daily life. For example, companies can use f...

this, plus it really fucks over autistic folks (some whom are also Black of course). what my face is doing at any specific moment does not necessarily correlate to my true emotions at that moment.

Replacing the human racism that currently does this

my only qualm with that is that it should read "...some emotional AIs are programmed to disproportionately attribute negative emotions..." they aren't doing this of their own free will. this is the end result of how they are programmed to process input and generate output

And the general terribleness of the inputs. Consistently labeled training data is expensive, and The Open Internet is not well labeled.

My blood ran cold when I read "policing".

Post