Skip to content

R0043/2026-03-28/Q001/SRC09/E01

Research R0043 — Sycophancy Vocabulary
Run 2026-03-28
Query Q001
Source SRC09
Evidence SRC09-E01
Type Analytical

Healthcare acquiescence problem and related terminology

URL: https://www.mdpi.com/2076-3417/16/2/710

Extract

Healthcare-specific terminology: - Acquiescence problem: "Current AI systems passively confirm rather than challenge clinicians' hypotheses, reinforcing cognitive biases such as anchoring and premature closure" - Automation bias: Used extensively in healthcare CDSS literature for inappropriate reliance on AI recommendations - Alert fatigue: Clinician desensitization from excessive automated alerts, leading to inappropriate override of ALL alerts - Deskilling: Loss of clinical diagnostic reasoning abilities due to overreliance on AI tools - Commission errors: Following incorrect automated advice - Omission errors: Failing to notice when automation does not flag a problem

The Dialogic Reasoning Framework proposes three roles to counter acquiescence: 1. Framework Coach: Guides structured reasoning 2. Socratic Guide: Asks probing questions 3. Red Team Partner: Presents evidence-based alternatives

JUDGMENT: Healthcare has the richest vocabulary for the phenomenon after AI safety. The "acquiescence problem" is the closest domain-specific equivalent to "sycophancy" — it describes system behavior (passive confirmation) rather than just human behavior (overreliance). However, it frames the AI as passive (failing to challenge) rather than active (deliberately agreeing), which is a meaningful distinction from sycophancy.

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Supports Healthcare has multiple domain-specific terms
H2 Contradicts Rich vocabulary exists
H3 Supports "Acquiescence problem" is the closest to system-side framing but still treats it as passive failure rather than active behavior — maintaining the asymmetry thesis

Context

The healthcare context is particularly important because patient safety makes the stakes of sycophancy-adjacent behavior concrete and measurable. The 26% increase in error rates from automation bias (Parasuraman & Manzey) demonstrates that the vocabulary gap has real consequences.