Send message to Author

Paper short abstract

How do physicians make sense of AI outputs? This study shows how vocal biomarkers reshape perception, redirect clinical attention, and contribute to distributed clinical reasoning in AI-assisted teleconsultations.

Paper long abstract

This study explores how AI-generated vocal biomarkers reconfigure clinical attention and contribute to distributed clinical reasoning during primary care teleconsultations. AI-based vocal biomarker analysis has been presented as a promising technique for the early detection of Parkinson's disease. In this study, video-recorded simulated teleconsultations in which family physicians interact with an AI system providing vocal biomarker analyses are examined. Post-simulation interviews are also used to investigate how AI outputs are integrated into clinical reasoning and decision-making.

Preliminary findings suggest that AI-generated vocal biomarkers reshape what clinicians attend to, how they interpret vocal and clinical signs, and how diagnoses evolve after dashboard consultation. Clinical reasoning emerges as a distributed accomplishment involving physicians, patients, and AI-generated vocal biomarkers, whose outputs are continuously interpreted, negotiated, and situated within the clinical encounter. Four key dynamics were identified:

1. Cross-validation: AI outputs reinforce physicians' clinical assessments.

2. Clinical arbitrage: When AI assessments differ from physicians' judgments, clinicians engage in interpretive work to evaluate the relevance and implications of data generated.

3. Sensory reframing: AI-generated biomarkers reveal discrepancies between physicians' sensory assessments of patients' speech and algorithmic analyses. The system reconfigures how vocal cues are interpreted by creating tensions between clinicians' perceptions and algorithmic analyses.

4. Attentional reframing: AI-generated biomarkers redirect diagnostic attention toward aspects that were previously overlooked or considered peripheral, inciting physicians to explore new lines of inquiry and consider alternative diagnostic hypotheses.

By simultaneously reconfiguring perception and attention, this AI system shapes what clinicians notice, question, and ultimately consider to be relevant.