AI scribes may fail to capture patients' experiences

A new review raises concerns that artificial intelligence (AI) scribes could compromise the nuances of patient-physician interactions, potentially overlooking critical details during consultations. These AI tools, known as ambient AI scribes, integrate speech-to-text technology with AI to assist clinicians in capturing patient notes during medical appointments.

However, research conducted by the University of Edinburgh indicates that while AI can expedite administrative duties, it may lead to deskilling among clinicians and deny them the chance to develop clinical reasoning skills.

The study, which analyzed 27 scientific and medical articles, found that despite AI tools' ability to transcribe patient speech, they often fail to capture essential elements like facial expressions, gestures, and emotional states. Clinicians frequently rely on these nonverbal cues to better understand a patient's experience and to foster trust.

AI-generated summaries tend to prioritize clinical information, which might overshadow patients' personal stories and accounts of their illness. Moreover, patients may hesitate to discuss sensitive issues like substance abuse, domestic abuse, or mental health struggles when they suspect the consultation is being recorded and processed by AI.

The review also highlights that manual note-taking plays a crucial role in clinicians' clinical reasoning and reflection. Using AI scribes could result in clinicians delegating mental effort—known as cognitive offloading—which may have benefits such as enabling more meaningful conversations with patients and allowing more time for complex tasks.

However, these tools may also hinder memory recall and skill development in clinicians. Instances of clinicians forgetting their own notes or struggling to recognize patients upon subsequent visits were reported.

The researchers emphasize that AI scribes must preserve patients' voices rather than overwrite them. They also stress the need for further research to comprehend the medium- and long-term use of these AI tools in specific health care settings and potential emerging risks, particularly in countries with healthcare systems distinct from those where the AI tools were developed.

The findings were published in BMJ Digital Health and AI. Dr. Lucas Seuren, a GAIL Fellow and research fellow at the University of Edinburgh's Centre for Biomedicine, Self and Society, cautioned that while many clinicians are enthusiastic about ambient AI scribes for their potential to reduce paperwork, the experiences of patients should not be overlooked, as they risk losing valuable narratives that could further disadvantage marginalized individuals in health and social care services.