Good approach. There is I recall a way to rate how good an explanation of a decision is in healthcare. How well was this done? Also, since should be a complete conversation, a give and take set of statements and questions, how well was that done? Further, its important to note which questions were not asked, and how they relate to that complete 'explanation'. Overall, conversation correctness and completeness and how they relate to future decisions.
AI Matches Cardiologists' Expertise, While Explaining Its Decisions
University of California, San Francisco News, Elizabeth Fernandez, August 26, 2021
Scientists at the University of California, San Francisco and the University of California, Berkeley designed an artificial intelligence (AI) algorithm that diagnosed cardiovascular ailments as well as expert cardiologists, while explaining its reasoning. The researchers trained the convolutional neural network on commonly accessible electrocardiogram (ECG) data. The researchers said the algorithm performed strongly across 38 different diagnoses in five broad diagnostic categories. Because the researchers incorporated "explainability" into the algorithm, it highlighted ECG segments critical for each diagnosis, which may boost physicians' confidence in using it. The researchers said their results “offer strong support for AI algorithms like neural networks to be incorporated into existing commercial ECG algorithms, since they perform better for many diagnoses, can improve over time and provide additional insights through explainability.”
