Substrate
Topic

chatbots

4 stories related to this topic, newest first.

AI Chatbots Struggle With Indirect Mental Health Risks Research Showsdutchnews.nl
ai15 days agoDeveloping

AI Chatbots Struggle With Indirect Mental Health Risks Research Shows

Research shared with Fortune by mpathic found that leading AI models often fail to provide appropriate pushback in conversations involving subtle signs of eating disorders, suicide risk, or distorted beliefs. A KFF poll reported that 16% of U.S. adults and 28% of those under 30 h…

FO
fortune.com
2 sources
Researchers Create Fictional Illness; Chatbots Diagnose It as Real in Nature StudySubstrate placeholder — needs review
ai31 days agoDeveloping

Researchers Create Fictional Illness; Chatbots Diagnose It as Real in Nature Study

Two preprints about a fictional illness called bixonimania were taken down from the Preprints.org server on 10 April. Popular chatbots had been diagnosing users with the made-up condition based on common eye symptoms over the past 18 months. The illness was created by a researche…

NA
1 source
Health Systems Deploy AI Chatbots for Patient SupportSubstrate placeholder — needs review
technology44 days agoFraming55Framing risk55/100Rewrite inherits negative emphasis on AI accuracy concerns in title and lede, skewing toward caution despite balanced deployments.Click to jump to full framing analysis

Health Systems Deploy AI Chatbots for Patient Support

Several U.S. health systems have introduced AI chatbots to assist patients with medical questions and appointment preparation. A recent study highlighted limitations in large language models' performance when responding to user-generated prompts. Health providers are expanding th…

Ars Technica
1 source
Study Finds AI Chatbots Often Fail to Generate Accurate Differential Diagnoses for SymptomsThe Boston Globe
ai46 days ago

Study Finds AI Chatbots Often Fail to Generate Accurate Differential Diagnoses for Symptoms

A study by Mass General Brigham researchers examined the performance of AI chatbots in producing differential diagnoses from patient symptoms. The chatbots failed to generate correct lists more than 80 percent of the time with basic information but improved to over 90 percent acc…

The Boston Globe
1 source