AI healthcare chatbots fail users with cognitive disabilities, study finds
A new study reveals that conversational agents designed for symptom checking in public healthcare are inaccessible to people with cognitive impairments, using confusing language and triggering distrust. The findings highlight a critical gap for healthcare systems deploying AI tools without adequate accessibility testing—a challenge that could expose organizations to both patient safety risks and regulatory liability.
Originaltitel: A collaborative approach to the evaluation of cognitive accessibility of a conversational agent for public healthcare
<p>Conversational agents are increasingly being introduced in healthcare to enhance access to healthcare and address limited healthcare resources. The aim of this study was to explore experiences of a conversational agent for symptom checking and triage tool assessing the urgency and appropriate level of care implemented in public healthcare from a cognitive accessibility perspective. With a Participatory Action Research approach, this study includes: (1) A Participatory Cognitive Barrier Walkthrough in workshops with people with cognitive impairment (n = 7), product owner representatives (n = 4), and researchers (n = 5), followed by a workshop for improvement suggestions; (2) a paired interview with healthcare system specialists (n = 2). The data from the study were analyzed using reflexive thematic analysis, resulting in five themes: 'The design of the eHealth service created expectations that were not met during interaction'; 'The eHealth service uses a difficult language'; 'The eHealth service triggers negative emotions and reactions'; 'The layout and graphic design of the eHealth service is difficult and sometimes misleading'; and 'The eHealth service lacks in trustworthiness' .The findings show the importance of including people with lived experience of cognitive impairment, to expose and find solutions for cognitive accessibility issues.</p>