Is the goal of the AI model to predict crime rates in a hypothetical world where everyone has equal rates of lead exposure?
What a terrible example. There are many other features being omitted that would predict crime rates. If anything, this is an example of enforcing your own bias on the model by not including all relevant features.
Great then go ahead and include additional features. The problem is, recidivism rates in real life are not equal so as the model increases in accuracy it will inevitably produce disparities in its results.
What a terrible example. There are many other features being omitted that would predict crime rates. If anything, this is an example of enforcing your own bias on the model by not including all relevant features.