Skip to content

R0044/2026-04-01/Q001/SRC04/E01
Research R0044 — Expanded Vocabulary Research
Run 2026-04-01
Query Q001
Source SRC04
Evidence SRC04-E01
Type Analytical

FDA CDS guidance addresses automation bias through human independent review, not system output constraints

URL: https://www.cooley.com/news/insight/2026/2026-01-20-automation-bias-and-clinical-practice-fda-makes-incremental-updates-to-clinical-decision-support-software-guidance

Extract

The FDA's January 2026 Clinical Decision Support guidance:

  • Relocated automation bias discussion to the fourth statutory criterion, which requires that "healthcare professional[s] independently review the basis for recommendations."
  • Focuses on human-side behavior: preventing clinicians from over-relying on software recommendations without independent verification.
  • Does not impose explicit system-side behavioral controls on AI output generation.
  • Establishes that software rendering physicians unable to independently evaluate recommendation foundations may be classified as a regulated medical device rather than exempt CDS.
  • Now permits singular recommendations to qualify for enforcement discretion, moving away from previous requirement for multiple alternatives — reducing one system-side mechanism (recommendation pluralism) that could have mitigated automation bias.
  • FDA "continues to support its concern about automation bias by citing only a 2004 journal article."

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Contradicts FDA has not produced system-side output behavior constraints
H2 Supports FDA recognizes automation bias as a concern but addresses it through human-side requirements
H3 Supports In the healthcare sector specifically, requirements are entirely human-side

Context

The FDA's approach is notable for what it does not do: it does not require AI systems to express uncertainty, challenge clinician assumptions, present alternative diagnoses, or avoid reinforcing user expectations. The regulatory mechanism is to ensure the human can independently review — not to ensure the AI behaves in a way that facilitates critical evaluation.

Notes

The relaxation of the multiple-alternatives requirement is significant — recommendation pluralism was one system-side design pattern that could have mitigated automation bias by forcing the system to present alternatives rather than a single recommendation.