LAUSR

Artificial intelligence (AI) could be a great thing in medicine. It could make healthcare safer and faster. It could make medicine more satisfying to practise and less unpleasant to receive. But we must test a hypothesis before we roll it out to the public. The huge datasets being collected to use AI in diagnostic radiography, for example, are indicative of the amount and quality of work needed to make reliable, safe tools. AI must be subjected to the same criteria and testing we’d expect with other forms of evidence based medicine. NHS 24 in Scotland now has a symptom checker app. While there’s no published evaluation, it’s been based on NHS 111 algorithms, has undergone user testing, and remains open to improvements (on raising a concern about the …