R0042/2026-04-01/Q003 — ACH Matrix¶
Matrix¶
| H1: Enterprise documented anti-sycophancy private AI | H2: Anti-sycophancy at developers, not enterprises | H3: No anti-sycophancy work anywhere | |
|---|---|---|---|
| SRC01-E01: Anthropic's comprehensive anti-sycophancy program | -- | ++ | -- |
| SRC02-E01: Cognition Dynamics and AI Innovate reduction | - | + | -- |
| SRC03-E01: Georgetown vendor-obligation framework | -- | ++ | - |
Legend:
- ++ Strongly supports
- + Supports
- -- Strongly contradicts
- - Contradicts
- N/A Not applicable to this hypothesis
Diagnosticity Analysis¶
Most Diagnostic Evidence¶
| Evidence | Why Diagnostic |
|---|---|
| SRC01-E01 | Anthropic's program is the most comprehensive documented anti-sycophancy effort — and it is entirely a model development activity, not enterprise deployment. Strongly discriminates between H1 and H2. |
| SRC03-E01 | Georgetown's vendor-obligation framing explains WHY enterprises have not adopted anti-sycophancy as a deployment goal — because the policy conversation assigns this responsibility to vendors. |
Least Diagnostic Evidence¶
| Evidence | Why Non-Diagnostic |
|---|---|
| SRC02-E01 | Case studies are unverifiable and the organizations are AI research firms, making it unclear whether this represents enterprise or developer activity |
Outcome¶
Hypothesis supported: H2 — Anti-sycophancy work is active at AI model developers and researchers, but no enterprise deploying private AI has documented it as a design goal.
Hypotheses eliminated: H1 — No enterprise case study found. H3 — Extensive anti-sycophancy work exists at developer level.
Hypotheses inconclusive: None