HORTRAME: Accelerating Epistemological Alignment
REVIEW APPROACH · PUBLIC WORKING SYNTHESIS

Epistemethics and revision history

How HORTRAME relates evidence, adversarial review, ethical judgment and accountable revision.

Working synthesis · public process record v07

Reading guide (summary)

The complete Epistemethics v1.00 constitution is authoritative for this reference. This page is a condensed reading guide and historical process record, not a replacement text. Epistemic and ethical evaluation are coupled, not sequential. Review addresses the claim and its reasoning, not the person. Read the complete English constitution (PDF).

No novelty is claimed for this framework. No component, combination or presentation is offered as original. It is a working synthesis of established epistemic, scientific-review and ethical approaches, selected and combined according to what HORTRAME currently judges most pertinent to its needs. The selection is provisional and may change with evidence, context and use.

Epistemology limits what the available evidence permits us to claim. Adversarial review tests whether those claims survive serious challenge. Ethical evaluation governs what may or should be done under the resulting uncertainty and consequences for affected parties. None substitutes for the others.

Evidence and claims

Keep definitions, proofs, simulations, software tests, observations, interpretations and hypotheses distinct. Separate intended, implemented and observed behavior. Preserve provenance, assumptions, alternatives and uncertainty in the form in which they matter.

Challenge and comparison

Reconstruct the strongest defensible claim, then test the inference for hidden assumptions, proxy substitution, confounding, circularity and overclaim. Compare realistic alternatives, including simpler mechanisms and inaction. Confidence follows evidence, not agreement or prestige.

Ethical judgment

Keep empirical claims and normative commitments distinguishable but coupled. Assess purpose, means, consequences, distribution, agency, consent, coercion, privacy, power, reversibility and long-term effects. Severe harm, rights and unequal burdens are not erased by an aggregate score.

Action and revision

When action is required under uncertainty, compare acting, delaying, narrowing, learning and doing nothing. Raise the evidential burden with severity, irreversibility, systemic reach and unequal distribution. Preserve attributable corrections, proportionate access and scrutiny of the evaluator itself.

The workflow is continuous: evidence constrains claims; review applies counterpressure; ethical judgment constrains action; outcomes become new evidence. Additional review or computation is warranted only when it is expected to reduce relevant uncertainty, expose meaningful failure or protect against consequential harm.

Research context

The research originated in a modular QBP implementation developed in Simulink/QUARC during a first internship in the second year of an engineering program. Shokoufeh Naderi’s supplied centralized MATLAB simulation was the practical algorithmic reference. Maude J. Blondin supervised the internship, but did not approve the unproved extrapolation to real-time agent addition and removal. She knew the high-level claim and demanded proof; she did not know all additional implementation choices.

The implementation exposed a communication-delay problem. Naderi developed the subsequent refinement, which Maude relayed. Whether that final refinement was installed before the internship ended remains unresolved. The original QBP implementation ran in simulation and on physical QBot robots before the refinement; the current membership formulas and browser implementation have not been tested on physical robots.

The membership exploration arose from system implementation without specialist mathematical or multi-agent expertise. Its exploratory motivation and additional engineering choices do not establish independent invention of the assigned algorithm or priority over the research literature.

What changed in the scientific claim

ENDPOINTS

Membership support does not imply an unchanged result.

Exact counterexamples ruled out unrestricted endpoint preservation. The question became when the prescribed probability-state maps return the original mean after one agent arrives, mixes and leaves.

CHARACTERIZATION

Separate the general theorem from its executable specialization.

Theorem 4.1 allows arbitrary nonnegative row-stochastic mixing and characterizes feasibility, uniqueness and zero newcomer influence. Corollary 4.2 adds equal retention and double stochasticity. The controlled browser trial uses the consensus-power specialization, with a prescribed number of updates. Standard final-epoch consensus gives the endpoint interpretation.

PRIMARY SOURCES

Compare state semantics and mechanisms.

Naderi’s priority and delay studies use fixed membership. Zhou et al. already change the length of a global state vector while individual states remain fixed-dimensional. Deplano et al. formulate open operators between changing labeled spaces; variable dimension and state maps in general are not novelty claims here. Hu et al. use mass/count accounting and departure handoffs under their communication assumptions. Here each row allocates probability across active objective identities and departure conditions retained coordinates.

The article’s scope and primary references →

EXPERIMENTAL SCOPE

Distinguish transport, state dynamics and motion.

Synthetic message delays use a loss-free central barrier and preserve synchronous logical mixing. The controlled restoration trial is separate from ordinary random arrivals. Robot motion uses a global epoch target and is illustrative. None establishes an asynchronous delayed/open protocol or physical validation of the current formulas.

The latest received review round, R5, evaluated author v05. The current v07 article incorporates subsequent framing, prior-art and disclosure revisions. These records are not journal acceptance or independent human proof approval.

A conditional application, not a fault-management result

Admissible membership transitions need not wait for priority consensus. They can define state changes for a policy that turns sustained communication loss into deliberate departure and later reconnection into a declared arrival. Detection, coherent membership decisions, connectivity, state transfer and continuing failures remain separate obligations. Restoration still requires its stated conditions; the browser has no loss detector or such fault-management policy.

Use of AI

AI assisted coding during the original internship implementation; research direction and implementation decisions were not delegated to AI. Later assistance included archive and literature analysis, mathematical formulation and proof drafting, counterexamples and tests, adversarial review, manuscript preparation, simulation development and the bilingual companion. The supplied algorithm and original implementation records were inputs to that later process.

The paper places the research-process summary and AI acknowledgment after the mathematics. Responsibility for its claims remains with the author. Numerical checks establish only their tested scope, and the selected venue’s disclosure requirements must be checked at submission.

Consistency is a separate release check

A current PDF and passing engine tests do not guarantee that every translated page, explanatory card, download and version record is current. This revision checks those surfaces separately and generates the repeated mathematical explanations from one reviewed source-bound record. The ordinary recurrence remains background; the restoration characterization leads the current presentation.

An initial missing-v06-history finding was withdrawn when a complete inspection found the entries in the current-records category. The corrected record distinguishes a partial view from actual absence. Automated bindings and release checks support this process; they do not replace semantic review or mathematical proof.

Reviewer companion — Lemma 2.1

Reviewer-produced reconstruction of Lemma 2.1; not the original cited supplement or independent human mathematical validation.

Versioned process records

The current public edition is v07, bound to local source-method v08. These are separate lineages. Earlier public editions remain immutable snapshots and may contain superseded wording.

Work in progress — view the current process records

Author · public v07

Concise bilingual record. The full methodology is not yet presented as finished research.

SHA-256 · 0982818e6a7eb77525dd6e46bc62db4f75361fc69ad628645d3bf18403a346fe

Reviewer · public v07

Concise bilingual record. The full methodology is not yet presented as finished research.

SHA-256 · fb4df73120fdbfc4a32bf6f058e8316f9e25dc3190bcc2b20f2e51de8c055320
v06 · Archived — Future method and conditional applications

Preserved as published at this stage; read the current edition for the latest assessment.

Author · Markdown · ZIP80179a807243fce2d238dff70f14beaf5eacd85110de870f02958f5baf3ad750

Reviewer · Markdown · ZIP9cd9e184ba54ac6e2c2c88d27103a4aa431ba23e7074ed0122c6a534fab963b1

v05 · Archived — Executable transport and restoration

Preserved as published at this stage; read the current edition for the latest assessment.

Author · Markdown · ZIPea3ae450a6509692487f1f277983f8a435b381a28c55b64d4b457dda09c3463b

Reviewer · Markdown · ZIP26de86d2298b63c2d6826b01841df2054d09ce38c25157a9ea4431f9a625757c

v04 · Archived — Supervision and approval

Preserved as published at this stage; read the current edition for the latest assessment.

Author · Markdown · ZIP68ae676b4a089a883d5f1aa9a6fccc69f74b43412860c33179bc3f9cf9d4388c

Reviewer · Markdown · ZIP9b9cd2c1fdc451e11ccd88c57a8b75b83e186043e7e6d092af6d0c6b5b358407

v03 · Archived — Source semantics and AI disclosure

Preserved as published at this stage; read the current edition for the latest assessment.

Author · Markdown · ZIP6a065084c334dcf128b23203181b0f6fa1364fc1e4a889f4b5385677f01faf9b

Reviewer · Markdown · ZIP85605554d8cc78fd117686a13466b8bca27a3d3bbc523f0057ee36f4bbe1499b

Public v01–v02 presentation was withdrawn. Its original bytes and detailed local source snapshots remain in the private project archive; their identities are retained in the metadata.

Complete categorized history and predecessor hashes

Article v07 · Interface v07 · Process v07 · Release record

Evidence types and revision records / types de preuves et révisions