Established literature provides components, not confirmation of the ARM-3 architecture.
ARM-3 · Foundation v1.1.3 · Research Status
What the Core claims — and what can disconfirm it.
ARM-3 is a theoretical research prototype, not a validated psychometric or diagnostic system. Its scientific value depends on whether the constructs can be operationally identified, whether the central Activation-by-Protection prediction survives preregistered causal tests, and whether the architecture adds value beyond simpler models.
Testable statement of Foundation v1.1.3.
Simpler explanations must be tested actively.
Failure of central tests should lead to simplification, merging, or rejection.
Empirical anchors
Research traditions ARM-3 builds on.
Motivated Reasoning & Motivated Skepticism
Motivational goals can influence search, evaluation, argumentation, and belief formation. ARM-3 cannot claim these phenomena as novel discoveries.
Belief Updating & Evidence Evaluation
Belief change can be separated from upstream processes such as search, source acceptance, interpretation, and evidence weighting.
Effort Allocation & Cognitive Control
Cognitive effort is costly and context-sensitive. ARM-3 adds the question of whether mobilized cognition changes epistemic direction under protection pressure.
AOT, Intellectual Humility & Metacognition
Alternative consideration, fallibility recognition, monitoring, and control are important comparison constructs against which CF and EMC must show incremental value.
Canonical Core hypotheses
H1–H7 of Foundation v1.1.3
Capacity-Activation Dissociation
C and state CA should separate empirically; CA must add beyond ability and state/trait engagement.
Commitment-Revision-Cost Differentiation
MC and MRC should show discriminant and incremental validity; otherwise merge or simplify them.
Threat-Appraisal Path
TE should affect TA under matched conditions; TA may transmit part of the effect and TS may moderate TE → TA.
Model-Protection Asymmetry
Motivational alignment should predict directional asymmetry beyond priors, source reliability, knowledge, and task difficulty.
Activation-by-Protection
Positive preregistered Z_A × Z_P ITT interaction on EWA, conditional on successful independent first-stage checks.
Elaborated Rationalizans
Under high protection pressure, independently validated activation should increase selective-reasoning complexity beyond general response complexity.
Regulatory Moderation
CF and/or EMC should reduce translation of protection-relevant conditions into MEA if they are distinct regulatory constructs.
Incremental Validity
ARM-3 should outperform preregistered simpler models under a complexity-sensitive comparison; better in-sample fit alone is insufficient.
Canonical causal test
H5a is not confirmed by just any correlation.
Randomize protection pressure without changing the evidential payload; where feasible, experimentally create model ownership or motivational alignment.
Randomize activation encouragement independently of protection pressure and estimate the Z_A × Z_P ITT interaction as the primary causal contrast.
Preregister first-stage criteria: Z_A must sufficiently change an independent CA indicator; Z_P an independent protection-manipulation indicator.
Premeasure priors and experimentally specify or independently measure source reliability.
Preregister EWA as the canonical primary H5a outcome, its direction, SESOI, and equivalence region.
Replicate the core interaction in at least two independently preregistered, first-stage-valid paradigms, including at least one independent research team.
Legacy Instrument v1.0
Earlier scores and 16 reference profiles are not part of the reconstructed Core Theory.
Epistemic Integrity, Epistemic Resilience, Defensive Pressure, earlier Rationalization Power, Rational Stability Index, shift metrics, and prototype similarity may remain for backward compatibility and exploratory research. They are not validated latent constructs, population norms, or normative individual classifications of Foundation v1.1.3.
reference profiles / prototypes
current scoring
from v1.1.3 measurement architecture
No retroactive validation
The new Core Paper does not retroactively validate earlier scores. A future instrument must independently operationalize capacity, activation, antecedents, regulation, and outcomes and undergo new psychometric evaluation.
Disconfirmation
When ARM-3 should contract or fail.
If capacity and state activation cannot be separated reliably, the distinction must be simplified or abandoned.
If the two constructs lack discriminant and incremental validity, they should be merged.
If the preregistered core interaction does not replicate in valid canonical tests, the strongest specific ARM-3 claim is unsupported.
If a strong state-engagement × directional-motivation rival explains equally well and no discriminating prediction favors ARM-3, support is limited to integrative/organizational value.
The expanded mechanism architecture is defensible only if MC, MRC, TA, CF, and EMC add value beyond the simpler Core model.
Selected literature
Scientific connections
This selection situates the Core Paper and is not a systematic literature review.
- Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498.
- Taber, C. S. & Lodge, M. (2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political Science, 50(3), 755–769.
- Shenhav, A., Botvinick, M. M. & Cohen, J. D. (2013). The expected value of control. Neuron, 79(2), 217–240.
- Fleming, S. M. & Lau, H. C. (2014). How to measure metacognition. Frontiers in Human Neuroscience, 8, 443.
- Stanovich, K. E. & Toplak, M. E. (2023). Actively Open-Minded Thinking and Its Measurement. Journal of Intelligence, 11(2), 27.
- Sommer, J., Musolino, J. & Hemmer, P. (2024). Updating, evidence evaluation, and operator availability. Psychological Review, 131(2), 373–401.
Research status
Foundation v1.1.3 is a non-peer-reviewed theoretical research prototype. No proposed ARM-3 mechanism has yet been established as an ARM-3-specific validated latent construct. Current instrument scores must not be used for clinical, legal, insurance, employment, or other high-stakes decisions.