Professor Menna El-Assady shines a torch into the black box of generative AI: How does a model arrive at decisions? And how can we correct them?