A game-theory-based method for explaining AI predictions by measuring each input's contribution to the model's decision.