R0048/2026-04-01/Q001 — ACH Matrix¶
Matrix¶
| H1: Comprehensive coverage | H2: Widespread but shallow | H3: Perfunctory/checkbox | |
|---|---|---|---|
| SRC01-E01: NAVEX general "limitations" language | - | ++ | + |
| SRC02-E01: GSA 21-session program, broad themes | N/A | + | - |
| SRC03-E01: Deloitte governance framework approach | - | + | + |
| SRC04-E01: DOL names hallucinations explicitly | + | + | -- |
| SRC05-E01: NHS names automation bias explicitly | ++ | + | -- |
| SRC06-E01: EU AI Act broad mandate | N/A | + | - |
| SRC07-E01: Microsoft six principles | - | + | - |
Legend:
- ++ Strongly supports
- + Supports
- -- Strongly contradicts
- - Contradicts
- N/A Not applicable to this hypothesis
Diagnosticity Analysis¶
Most Diagnostic Evidence¶
| Evidence | Why Diagnostic |
|---|---|
| SRC04-E01 | DOL's explicit naming of hallucinations discriminates between H2 and H3 — it shows training goes beyond checkbox level for at least some frameworks |
| SRC05-E01 | NHS naming of automation bias discriminates between H1 and H2 — it shows some programs achieve specific failure-mode coverage, but only in one sector |
| SRC01-E01 | NAVEX's general language discriminates between H1 and H2 — commercial training's lack of specificity contradicts comprehensive coverage |
Least Diagnostic Evidence¶
| Evidence | Why Non-Diagnostic |
|---|---|
| SRC06-E01 | EU AI Act's deliberate flexibility means it supports both H2 (broad training satisfies it) and H3-contradiction (legal mandate exists) equally |
| SRC02-E01 | GSA training's existence contradicts H3 but its undocumented content prevents evaluating H1 vs H2 |
Outcome¶
Hypothesis supported: H2 — Training is widespread but covers limitations at a shallow level. Most programs address limitations generally; only government healthcare (NHS) and the DOL framework reach the level of naming specific failure modes.
Hypotheses eliminated: H3 — The DOL framework, GSA program, NHS framework, and EU AI Act mandate collectively eliminate the claim that training is purely perfunctory. However, H3 captures a valid concern about commercial training depth.
Hypotheses inconclusive: H1 — Cannot be fully evaluated because internal training content is not publicly visible. The NHS framework partially supports H1 for healthcare-specific training, but no evidence supports H1 across the general corporate landscape.