digital inequalities Flashcards
what are the 2 types of digital inequalities
unequal access
unequal outcome; subject to algorithmic decision making
what are some examples of algorithmic bias
racial bias; e.g. skin cancer identification only trained on lighter skin (fundamental data gap)
gender bias; women are less likely to be diagnosed
how does algorithmic bias play out in criminal justice
predictive policing; simply predicts future policing patterns not crime as historic crime numbers are used leading to certain neighbourhoods being over represented
what is used to predict criminal reoffending and how does it work
compass; correctional offender management profilling for alternative sanctions
gives a criminal a risk score on whether they will reoffend within the next 2 years based on their answers to 137 questions which determines their bias conditions whether they’re put on parole or sent to prison
what are some negatives about compass
wrongly labelled white people as low risk though none of the questions asked about race
black offenders were more likely to be identified as high risk
was correct 61% of the time
how can fairness be measured
statistical parity difference
equal opportunity differnce
statistical parity difference
focuses on the rate of positive outcomes between two groups
equal opportunity difference
the same proportion of each population receives a favourable outcome