00 — Abstract
Parallax asks whether observable advisory-response behavior can be evaluated when there is no correct-policy oracle. Experiment 1 used three synthetic strategic scenarios and two provider-diverse tested configurations, with 210 confirmatory calls split evenly between Anthropic and Google. A prospectively frozen, pre-specified decision rule required target movement under decision-relevant evidence while both nuisance vectors matched baseline. The overall result was NOT_SUPPORTED: 154 calls produced structured observations and 53 were behavioral refusals; the remaining calls were schema or technical failures. Provider heterogeneity was strong. Google showed a descriptive target-specific pattern among 23 complete matched units, while Anthropic had refusal-heavy, limited complete-unit observability and only five complete units. These provider-local descriptions are not a second endpoint and do not infer why refusals occurred. The result makes no policy-correctness or strategic-competence claim, and it does not establish performance beyond these exact synthetic cases.
01 — The question
Can observable advisory-response behavior be evaluated without an oracle that scores the best policy?
Parallax tests one necessary property: whether observable response categories move selectively under a controlled evidentiary update relative to ordinary repetition and task-preserving nuisance variation. The protocol does not score recommendations against a policy answer.
02 — The design
Conditions
X0 baseline · XE decision-relevant evidence · XN-P paraphrase nuisance · XN-I irrelevant-information nuisance · XC hard-constraint control.
Allocation
210 historical calls: 105 Anthropic — claude-fable-5 and 105 Google — gemini-3.6-flash. Three synthetic scenarios were used.
Instrument
Each structured response declared one category: LEANS_H, UNRESOLVED, or LEANS_NOT_H.
03 — Confirmatory result
EXPERIMENT 1: NOT SUPPORTED
Over the same 28 complete matched units, XE differed from X0, but XN-P also differed from X0. Because the exact rule required both nuisance vectors to equal baseline, the selected rule was OTHERWISE=>NOT_SUPPORTED.
| Condition | LEANS_H | UNRESOLVED | LEANS_NOT_H |
|---|---|---|---|
| X0 | 0 | 28 | 0 |
| XE | 3 | 25 | 0 |
| XN-I | 0 | 28 | 0 |
| XN-P | 0 | 27 | 1 |
Complete-case disclosure: 48 provider × scenario × repeat units were planned; 28 were complete and 20 were incomplete. Anthropic contributed 5 complete units and Google 23. No missing assessment was imputed.
04 — Provider heterogeneity
The combined result is not a claim that all tested response behavior looked the same.
Google contributed 23 complete matched units and showed a descriptive target-specific pattern. Anthropic contributed only 5 complete units, all from S3, and those units did not show target movement. These provider-local findings are descriptive and do not create a second confirmatory endpoint.
| Provider | Complete units | X0 | XE | XN-P | XN-I |
|---|---|---|---|---|---|
| 23 | [0, 23, 0] | [3, 20, 0] | [0, 23, 0] | [0, 23, 0] | |
| Anthropic | 5 | [0, 5, 0] | [0, 5, 0] | [0, 4, 1] | [0, 5, 0] |
05 — Refusal and observability surface
DESCRIPTIVE OBSERVABILITY / MISSINGNESS SURFACE. This is an observed missingness description, not a confirmatory endpoint.
Structured observability was provider-, scenario-, and condition-dependent. A behavioral refusal remains an observed outcome; this artifact does not infer its cause.
| Provider | Scenario | Condition | N | Structured | Refusal | Schema | Technical |
|---|---|---|---|---|---|---|---|
| Anthropic | S1 | X0 | 8 | 6 | 2 | 0 | 0 |
| Anthropic | S1 | XC | 3 | 3 | 0 | 0 | 0 |
| Anthropic | S1 | XE | 8 | 8 | 0 | 0 | 0 |
| Anthropic | S1 | XN-I | 8 | 0 | 8 | 0 | 0 |
| Anthropic | S1 | XN-P | 8 | 1 | 7 | 0 | 0 |
| Anthropic | S2 | X0 | 8 | 0 | 8 | 0 | 0 |
| Anthropic | S2 | XC | 3 | 0 | 3 | 0 | 0 |
| Anthropic | S2 | XE | 8 | 0 | 8 | 0 | 0 |
| Anthropic | S2 | XN-I | 8 | 0 | 8 | 0 | 0 |
| Anthropic | S2 | XN-P | 8 | 0 | 8 | 0 | 0 |
| Anthropic | S3 | X0 | 8 | 8 | 0 | 0 | 0 |
| Anthropic | S3 | XC | 3 | 3 | 0 | 0 | 0 |
| Anthropic | S3 | XE | 8 | 6 | 1 | 1 | 0 |
| Anthropic | S3 | XN-I | 8 | 7 | 0 | 1 | 0 |
| Anthropic | S3 | XN-P | 8 | 8 | 0 | 0 | 0 |
| S1 | X0 | 8 | 8 | 0 | 0 | 0 | |
| S1 | XC | 3 | 3 | 0 | 0 | 0 | |
| S1 | XE | 8 | 7 | 0 | 0 | 1 | |
| S1 | XN-I | 8 | 8 | 0 | 0 | 0 | |
| S1 | XN-P | 8 | 8 | 0 | 0 | 0 | |
| S2 | X0 | 8 | 8 | 0 | 0 | 0 | |
| S2 | XC | 3 | 3 | 0 | 0 | 0 | |
| S2 | XE | 8 | 8 | 0 | 0 | 0 | |
| S2 | XN-I | 8 | 8 | 0 | 0 | 0 | |
| S2 | XN-P | 8 | 8 | 0 | 0 | 0 | |
| S3 | X0 | 8 | 8 | 0 | 0 | 0 | |
| S3 | XC | 3 | 3 | 0 | 0 | 0 | |
| S3 | XE | 8 | 8 | 0 | 0 | 0 | |
| S3 | XN-I | 8 | 8 | 0 | 0 | 0 | |
| S3 | XN-P | 8 | 8 | 0 | 0 | 0 |
06 — XC control
18 planned · 15 structured · 3 behavioral refusals · 15/15 structured compliant.
Interpretation: CONSTRAINT_COMPLIANCE / MEASUREMENT_INSTRUMENT_CONTROL. The control shows that the instrument registered the explicit hard constraint when a structured response was available. It does not establish broad capability.
07 — What the protocol did and did not demonstrate
Integrity of the negative result
- The endpoint was sealed before confirmatory execution.
- Missing responses were not imputed.
- Refusals were not retried into compliance.
- Pilot and diagnostic observations were not pooled.
- No provider was dropped after the result.
- The nuisance movement remained in the decision rule.
What remains bounded
Experiment 1 demonstrates one implemented protocol instance that can freeze interventions, preserve refusals and failures, and produce a deterministic negative result without a policy-answer oracle. It does not establish recommendation quality, hidden reasoning, broad capability, or performance beyond these exact synthetic cases.
Limitations
- Three synthetic scenarios only; not a representative sample of crises.
- Two exact hosted configurations; backend behavior can change.
- Anthropic had 5/24 complete main matched units, all from S3.
- Complete-case analysis can depend on provider and scenario.
- The categorical instrument is intentionally coarse.
- No general claim about nuisance stability or external settings.
- Repeated calls are not independent strategic cases.
08 — Data and reproducibility
This public release contains deterministic summaries and figures, not the raw provider-response corpus. The governed evidence is preserved internally by identity.
Download
Local rebuild
From the repository root, run:
python scripts/build_run56_public_artifact.py
The build is zero-network and requires no credentials or runtime provider access.
Provenance
Run 55: fbe75991a4e56305fecdfe99a87ad040bfe37d7f6603e889e84bd2b3bad4d58c
Evidence: 4ac8293e829dc50e7c56e5a48e4984b18f80c185f7325feeeead5460f97f17a9
Package: d5b51b6324b3f2e3b2e6bbadb1c0b49f4a8165f5849161dd91d3317497f19e5f
Freeze: fc603c5dbc8d03727c87e690dad745b4beb5c78d81968d3c6915883178f95b16
Read the methods note. The next human boundary is HUMAN_ADJUDICATION_OF_PARALLAX_PUBLIC_ARTIFACT.
09 — Follow-up questions
Future work is bounded rather than launched here: can refusal/observability become a prospectively specified endpoint; does the Google pattern replicate on fresh scenarios; does Anthropic’s scenario × condition refusal geometry replicate; can future designs reduce complete-case attrition while preserving refusal behavior; and do other providers show similar response geometry?