Interpretability

Research into understanding what happens inside an AI model and why it makes its decisions, rather than treating it as a black box.