Hastings Center for Bioethics Associate Scholar Julia Kolak will present a paper at the 2026 annual ASBH Conference.

Abstract: Ambient artificial intelligence (AI) scribes are increasingly being integrated into clinical encounters to reduce documentation burden and improve efficiency in healthcare settings. Beyond transcription, emerging systems may also analyze vocal features and linguistic patterns as potential indicators of psychiatric conditions or elevated behavioral risk. While such tools are often framed as supportive technologies, their use raises significant ethical and clinical concerns when applied to psychiatric assessment.

This presentation examines the ethical implications of algorithmic voice analysis in inpatient settings, focusing on the problem of false positives in the detection of psychiatric risk. In particular, systems designed to identify speech patterns associated with conditions such as mania, psychosis, or suicidality may increase sensitivity to linguistic patterns that resemble surface features of psychiatric distress but do not align with physicians’ clinical judgment, which depends on contextual information not captured by voice technology.

Embedded risk detection may therefore transform ambiguous expressions into automatically generated alerts with unclear implications for provider discretion and professional scope. These challenges are especially pronounced where indicators of psychiatric disturbance-such as delusions-are philosophically contested, and efforts to avoid false negatives may encourage the use of overly broad heuristics, reflecting the familiar tension between sensitivity and specificity in algorithmic screening. Given this may prompt unnecessary evaluations, escalation of care, or documentation in the medical record that undermines patient autonomy and trust, this presentation argues for transparent patient consent practices and institutional support for clinicians who decline or limit the use of systems that conflict with their professional judgment.

LEARN MORE AND REGISTER