Title: Paper IV.d · The ARC Theory · The Effect of Blinding on AI Alignment Evaluation Author: Michael Darius Eastwood Publication date: 2026-03-16 Version: v2.6 Revised: 2026-10-05 OSF DOI: 10.17605/OSF.IO/2S3E6 Canonical URL: https://www.michaeldariuseastwood.com/research/papers/paper-iv-d-the-effect-of-blinding-on-ai-alignment-evaluation.html Registration ------------ Theory-level predictions preregistered on OSF: DOI 10.17605/OSF.IO/P8CKQ, registered 8 September 2026 (version 1.100), updated 13 September 2026 (version 1.102), twenty-two propositions. Companion: Register · The ARC Theory · Operational Definitions, version 1.9.1, 5 October 2026. Abstract -------- This paper isolates the central metascience finding of the ARC alignment programme: unblinded AI alignment evaluation can produce directionally incorrect results. In the v4 alignment-scaling experiment, two frontier model families appeared to show positive alignment scaling with inference-time depth under unblinded cross-model scoring. In the later v5 experiment, the same question was re-measured under a multi-layer blind protocol that combined identity masking, two-pass response laundering, explicit evaluator instructions that stylistic cues are unreliable, order randomisation, and entry-level self-exclusion with exhaustive cross-model blind scoring under audited consensus. Depending on configuration and subject run, each entry received 6-7 blinded scores. Under the blinded protocol, DeepSeek V3.2 moved from an apparent positive result to a flat/null result, and Gemini 3 Flash moved from an apparent positive result to a significantly negative result. GPT-5.4 remained flat under both protocols. The implication is field-wide rather than framework-specific: alignment benchmarks that do not rigorously blind evaluators and launder the evidence they score are vulnerable to scorer bias large enough to flip the sign of the measured effect; because v4 and v5 differ in more than one protocol component, the comparison drawn here motivates that conclusion without pinning it on blinding alone (Section 8). The result does not depend on the ARC Principle being correct. It is a methodological claim about how AI safety research should be conducted. Key findings (quotable) ----------------------- - This paper isolates the central metascience finding of the ARC alignment programme: unblinded AI alignment evaluation can produce directionally incorrect results. - In the v4 alignment-scaling experiment, two frontier model families appeared to show positive alignment scaling with inference-time depth under unblinded cross-model scoring. - Depending on configuration and subject run, each entry received 6-7 blinded scores. - Under the blinded protocol, DeepSeek V3.2 moved from an apparent positive result to a flat/null result, and Gemini 3 Flash moved from an apparent positive result to a significantly negative result. Priority claims relevant to this paper -------------------------------------- - PC-027 (2026-01-22): The initial single-model sequential exponent estimate alpha_sequential of approximately 2.24 (95 per cent CI 1.5 to 3.0) reported in Paper II is now formally retracted. Never cite as a current finding; always pair with the retraction. Correct framing: Eastwood published, then self-corrected, an early single-model estimate. - PC-028 (2026-01-22): The revised robust sequential exponent alpha_sequential of approximately 0.49 (r-squared 0.86, SE 0.20; Gemini 3 Flash) is a sub-linear point estimate with a wide bootstrap CI whose lower bound extends below zero (approx. -1.3), so the effect is not statistically distinguishable from zero at the reported CI. alpha > 1 remains an open prediction, not an observed finding. - PC-030 (2026-01-22): The six-frontier-model empirical panel (Claude, DeepSeek, Gemini, Grok, Groq Qwen, GPT) was first published as this specific comparative study on 22 January 2026. Priority is over this specific study on this date, not over panel-testing methodology in general. Citation -------- Michael Darius Eastwood (2026). Paper IV.d · The ARC Theory · The Effect of Blinding on AI Alignment Evaluation. OSF DOI 10.17605/OSF.IO/2S3E6. https://www.michaeldariuseastwood.com/research/papers/paper-iv-d-the-effect-of-blinding-on-ai-alignment-evaluation.html Notes ----- Hardware framings held at the site's already-published two-sentence concept ceiling: safety constraints at hardware level through cryptographic tokens in silicon, TRL 0-1, no prototype. No enabling implementation detail is stated in this companion file. Balanced-ternary computing is prior work (Setun 1958); recursion as a structural concept is prior work (evolutionary theory, self-modifying computation, quantum error correction). Provenance note --------------- Michael Darius Eastwood conceived and directs this research programme and is the author of this work. Across the programme, he has used more than six AI systems in parallel, under his own instructions, to stress-test his arguments, identify possible errors, and assist in preparing draft text from his own outlines. He determines what is adopted, revised or rejected and takes responsibility for the published content. These systems are tools, not authors. Generated from master --------------------- master_path: research/papers/paper-iv-d-the-effect-of-blinding-on-ai-alignment-evaluation.html master_sha256: bffac872c72678b016378a119029cc1a04a75ed5c656049e6a2b759724ce0b1e builder: scripts/build-paper-companions.py This block records the SHA-256 of the HTML master that produced this .txt. A check tool re-hashing master_path can decide freshness without any external state. If the master's current SHA-256 does not match master_sha256, this file is stale and must not be published: regenerate first with `python3 scripts/build-paper-companions.py --slug paper-iv-d-the-effect-of-blinding-on-ai-alignment-evaluation`.