Slang, typos can cause AI to give incorrect health assessments
Common human typing errors, slang and missing gender references can artificial intelligence programs reviewing health records to give faulty assessments, a study found. Conducted by Massachusetts Institute of Technology researchers and presented at a recent conference of the Association for Computing Machinery in Athens, Greece, the study found mistakes or language choices increased the likelihood AI would recommend a patient self-manage their health problems rather than seek medical attention. The problems were more likely to change AI treatment recommendations for women, causing them to advise incorrectly that patients should not seek medical care. (HealthDay , 6/26/25)