Many sophisticated AI models work in ways that are difficult or impossible for humans to comprehend, which is why transparency and explainability are important when deciding whether to accept recommendations made by these models. People must not blindly trust models they cannot fully understand.

Existing prognostic tools such as nomograms have many advantages in predicting cancer patients’ chances of survival and ability to enjoy a good quality of life, but also tend toward inflexibility and inability to factor in new treatments, new data, and changes in patients’ conditions. AI, however, promises to remedy these shortcomings.