We know our data is being used.
But what happens after?
Created By: Eric Huang
Transforms a set of inputs to outputs. Faciliates human decision making ("United States").
Used when need to make many decisions where not possible for human to do so. For example: predicting crime or scheduling appointments (Larrazabal et al.) (Heaven).
Data is an integral part in creating ADS's. Need a lot of data to ensure the ADT knows what it is doing (Ng).
Past medical algorithm used on 200 million Americans was found by researchers to discriminate against African Americans (Ledford)
These medical algorithms can determine the health of entire communities (Samorani et al.)
17.7% of African American patients received extra care, compared to the significantly higher figure of 46.5% had it been unbiased (Ledford)
You are probably familar with automated decision systems or some variant of the technology.
Of course, the idea of such process is easy to grasp.
But the consequences they bring may not
be as well understood and recognized.
Low income minority neighborhoods impacted by predictive policing: different behavior in police officers > leads to agitation in officers (Heaven)
“Inherently discriminatory” as "decisions are made on the basis of subjects falling within so-defined groups” (B. Goodman and Flaxman).
Leads to cycles of discrimation: data obtained by biased behavior from officers fed into models used by officers again (Heaven).
Scroll Up