How banks use AI to catch criminals and detect bias

Data science and analytics teams at banks must find the right balance where their AI algorithms can ferret out fraudulent transactions without infringing on anyone's rights. Developers of AI systems make sure to avoid including problematic variables such as gender, race, and ethnicity in their models. But the problem is that other information can stand as proxies for those same elements, and AI scientists must make sure these proxies do not affect the decision-making of their algorithms. For instance, in the case of Amazon's flawed hiring algorithm, while gender was not explicitly considered in hiring decisions, the algorithm had learned to associate negative scores to resumes with female names or terms such as "women's chess club."