2 ms·
How to provide fairness and minimize the bias when implementing AI algorithms
- josephcsible 4y ago> Even when these classes of attributes aren’t being used in machine learning models, discrimination may still exist due to some correlation. That discrimination can be unintentional (disparate impact) or intentional (disparate treatment). If income and credit score are correlated with race, then is it wrong under the disparate impact theory for lenders to discriminate based on income and credit score? If so, this seems like prima facie evidence that there's something wrong with that theory.