AI Ethics

Ethics becomes concrete the moment a model makes a decision about a person: who gets the loan, whose CV is read, whose face is matched. A model trained on historical decisions will reproduce the pattern of those decisions, including the parts nobody would defend out loud.

Practical work here means measuring error rates separately across groups, documenting what the training data represents, keeping a human in the loop where the stakes justify it, and being able to explain a decision after the fact.