zlacker

>>pseudo+(OP)
They are complaining about "racial bias" in predictive policing products. This "racial bias" is composed of data collected from crime statistics. If reality doesn't support the ideology, reality must be ignored. This is the sorry state of academia nowadays.

>>lazyjo+ys
I don't think that's accurate at all. These algorithms are only as accurate as the data being input into them. The current system of policing is deeply flawed and as a result the "data collected from crime statistics" will be skewed by these inherent, systemic biases. Mathematical optimization against a flawed dataset of this kind is highly likely to create a feedback loop of increasing bias with terrible human consequences.

>>krohli+Av
This seems like a plausible criticism, but it also seems like a plausible criticism of all data and derivates about human beings (and probably other things).

>>perl4e+DF
It _is_ a plausible criticism of all data and why we want such large (and diverse) sets of them.

source: work with data and am not terrible at my job