Personalization is great. Netflix shows me the movies that I want to watch. Amazon shows me the products that I want to buy. As personalization increases, customer satisfaction should go up.

Personalization in education and medicine can work in the same way. If you have an AI-enabled tutor that understands the best way to present material to you, then that should help you learn.

Where it may get tricky, however, is when you consider the probabilistic aspect of today’s large language models. Injecting randomness into personalization in an attempt to improve outcomes cannot come at the expense of accuracy and fidelity.