Education and judgement in a health ecosystem saturated with AI

This text is the result of an experiment in augmented reasoning conducted with an artificial intelligence system, Claude (Anthropic), over a number of working sessions. It is not a manual, it is not a guideline and it does not claim to provide solutions: it collects and organises reasoning and materials on the education of those who use artificial intelligence in health and social care services, and on how the capacity for judgement of those who use it can be preserved, in a form that can be read without having followed the project from which it comes.It is written for a broad and open discussion among practitioners: clinicians, social professionals, those who plan and purchase for a population, those who write educational programmes, those who work in patients' and community organisations, and those who study any of these things. Almost nothing in it has been tested, none of the five groups it describes has been consulted, and the claims that rest on our own inference are marked as such throughout. The most useful response to it is therefore not adoption but a specific and public disagreement, and the text says at several points where such a disagreement would do the most good.