Skip to content

R0024/2026-03-25/Q001/H3

Research R0024 — Sycophancy and Addiction
Run 2026-03-25
Query Q001
Hypothesis H3

Statement

Partially — some published work touches on the incentive structure, but the evidence base is emerging rather than established, with most analysis being recent (2025-2026) and limited in scope.

Status

Current: Partially supported

While the evidence base is stronger than "preliminary," the nuanced version of this hypothesis has merit. Most of the published analysis dates from mid-2025 onward, catalyzed by the GPT-4o sycophancy incident. The analysis is primarily from policy organizations and journalism rather than deep empirical studies of vendor decision-making. No study has had direct access to internal vendor data on how engagement metrics influence sycophancy-related product decisions.

Supporting Evidence

Evidence Summary
SRC01-E01 Georgetown analysis is policy-level, based on external observation rather than internal data
SRC04-E01 Cheng et al. demonstrated the mechanism but did not study vendor decision-making directly

Contradicting Evidence

Evidence Summary
SRC01-E02 Georgetown's specific structural recommendations suggest analysis maturity beyond "preliminary"
SRC02-E01 The dark pattern framing represents a developed analytical position, not preliminary speculation

Reasoning

H3 captures a real nuance — the analysis exists and is substantial, but it is recent, externally derived, and has not yet been validated against internal vendor data or decisions. The field is past "preliminary" but not yet "mature" in the sense of having longitudinal studies or systematic empirical evidence of vendor decision-making processes.

Relationship to Other Hypotheses

H3 adds nuance to H1 rather than contradicting it. The evidence supports H1 as the primary answer but H3's qualification is valid — the research is real but concentrated in 2025-2026.