Please enable JavaScript.

Coggle requires JavaScript to display documents.

What kind of knowledge is necessary to be able to evaluate the validity…

- - - - ProPublica's analysis of Northpointe, finding that about 45% of African Americans deemed "high risk" never reoffended and about 50% of white americans deemed "low risk" did reoffend
  - - - What if they had to release it?
        
        The Formula's are unimportant because they do not explicitly ask about race
        
        RLS
        
        Facebook's add formula was found to target ads to specific groups despite reducing the amount of control advertisers had over targeting their adds.The way their formula worked still showed ads to targeted audiences by race despite not giving advertisers the option to target who the add was shown to by race and despite not collecting data regarding race.
        
        Similar to the Northpointe risk assessment which does not explicitly ask about race but asks questions such as education and occupation that result in racially biased predictions because race factors into those questions. This results in black people being given higher risk scores based on them not having a job or as good of an education as a white person, however because of systematic oppression, on average, black people are less likely to have a good education or job.
- - - - Northpointe disputes ProPublica's analysis and states that their methodology is inaccurate.