From Candidate Hierarchy to Experimental Dependency Map: Results from Five Preregistered Adversarial Tests of Maintenance, Heredity, Learning, Agency and Self-Modelling.
- 1Independent Researcher
Abstract
This paper synthesises the first completed preregistered experimental cycle of the Theory of Everything Research Programme and reconstructs the programme's dependency architecture in light of the confirmatory evidence from Experimental Blocks I–V.
Five experimental blocks tested candidate relations involving dynamical persistence and maintenance, heredity and selection, retention–feedback–learning, structural generalisation and model-based agency, and reflective self-monitoring. The programme used matched controls, direct intervention, rescue, held-out task variants, serious bypass architectures, preregistered decision rules and reproducible analysis.
The results do not validate a compulsory hierarchy. Block I retained internal causal maintenance while reducing reciprocal-topology necessity. Block II retained heredity and selection within an artificial ecology while leaving robustness to alternative reproducing-unit definitions deferred. Block III retained separable causal roles for retention, feedback and learning while preserving successful informational and control bypasses. Block IV retained a causal role for world-model content and counterfactual rollout in registered structural-novelty tasks. Block V retained active self-monitoring, impairment-sensitive recovery, misinformation/correction rescue and self-specific capability tracking, restricted episode-specific monitor correspondence, and reduced endogenous self-model necessity because externally supplied calibrated self-relevant information reproduced the registered downstream control function.
The evidence-driven reconstruction therefore favours a conditional dependency network containing scoped effective mechanisms, bypass routes and unresolved transition problems rather than a single ladder from physics to consciousness. In particular, the transitions from dissipative organisation to self-maintenance and from self-maintenance to reproduction remain open; broad intelligence is not identified with one architecture; and phenomenal consciousness remains an independent open problem.
This synthesis introduces no new confirmatory dataset. It consolidates the archived OSF preregistrations, Zenodo software releases, confirmatory datasets and decision records from the completed five-block programme. Cross-block judgements are synthesis-level inferences and are not presented as retroactively preregistered hypotheses.
Related
- From Candidate Hierarchy to Experimental Dependency Map: Post-HTP1.6 Updated Edition.
- Preregistered Adversarial Batteries: A Falsifiable Protocol for Security-Testing Autonomous AI Systems
- Why are some neurons spared in neurodegeneration? Cellular subtype identity maintenance as a candidate threshold-setting mechanism
- The Riemann Hypothesis with MCH: Research Audit of the "Three Gates" Operator Route to Weil Positivity (Adversarial Reconstruction, Independent Computation, and a Minimal Candidate Proof Spine)
- Food-Derived Umami Peptides: Linking Candidate Discovery, Experimental Validation, and Food Applications
BibTeX
@misc{craggs2026from,
title = {{From Candidate Hierarchy to Experimental Dependency Map: Results from Five Preregistered Adversarial Tests of Maintenance, Heredity, Learning, Agency and Self-Modelling.}},
author = {Kenneth Craggs},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23413881},
url = {https://zenodo.org/records/23413881},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - From Candidate Hierarchy to Experimental Dependency Map: Results from Five Preregistered Adversarial Tests of Maintenance, Heredity, Learning, Agency and Self-Modelling.
AU - Kenneth Craggs
PY - 2026
DA - 2026/10/16
PB - Zenodo
DO - 10.5281/zenodo.23413881
UR - https://zenodo.org/records/23413881
AB - This paper synthesises the first completed preregistered experimental cycle of the Theory of Everything Research Programme and reconstructs the programme's dependency architecture in light of the confirmatory evidence from Experimental Blocks I–V. Five experimental blocks tested candidate relations involving dynamical persistence and maintenance, heredity and selection, retention–feedback–learning, structural generalisation and model-based agency, and reflective self-monitoring. The programme used matched controls, direct intervention, rescue, held-out task variants, serious bypass architectures, preregistered decision rules and reproducible analysis. The results do not validate a compulsory hierarchy. Block I retained internal causal maintenance while reducing reciprocal-topology necessity. Block II retained heredity and selection within an artificial ecology while leaving robustness to alternative reproducing-unit definitions deferred. Block III retained separable causal roles for retention, feedback and learning while preserving successful informational and control bypasses. Block IV retained a causal role for world-model content and counterfactual rollout in registered structural-novelty tasks. Block V retained active self-monitoring, impairment-sensitive recovery, misinformation/correction rescue and self-specific capability tracking, restricted episode-specific monitor correspondence, and reduced endogenous self-model necessity because externally supplied calibrated self-relevant information reproduced the registered downstream control function. The evidence-driven reconstruction therefore favours a conditional dependency network containing scoped effective mechanisms, bypass routes and unresolved transition problems rather than a single ladder from physics to consciousness. In particular, the transitions from dissipative organisation to self-maintenance and from self-maintenance to reproduction remain open; broad intelligence is not identified with one architecture; and phenomenal consciousness remains an independent open problem. This synthesis introduces no new confirmatory dataset. It consolidates the archived OSF preregistrations, Zenodo software releases, confirmatory datasets and decision records from the completed five-block programme. Cross-block judgements are synthesis-level inferences and are not presented as retroactively preregistered hypotheses.
ER -