Hallucination
When an AI generates content that is fluent and plausible but wrong or never happened, such as a measurement that was never taken or a finding never discussed. In clinical documents this is a patient-safety issue, not a quirk.
A discharge summary drafted by a language model states that an echocardiogram showed an ejection fraction of 55%, for a patient admitted with breathlessness who had no echo this admission. The sentence is well formed and sits where an echo result usually goes, so the primary care physician reads it as left ventricular systolic function checked and normal. Check every number and named investigation against the record before signing, and favor systems that link each statement to its source entry so that check is quick.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →