Proposed
A construct, requirement, instrument, or explanation offered for criticism and testing. It has not yet been empirically established.
Perfinitive Research
Methods Architecture
Perfinitive Research uses a layered methods architecture. Different methods answer different questions and should not be collapsed into one framework.
Research workflow
Discover → Describe → Discriminate → Localize → Reduce → Test Consequence → Govern
Discover and describe with WMM; discriminate with ECAM; localize with OD-INV; test reduction with Structural Reducibility; test human consequence with HCP; examine control through Governed Continuity.
This workflow relates the methods. It does not require every project to pass through every stage.
WMM-01 — Human-First Water Mapping Method
Question: What consequential structures should we be looking for?
WMM-01 systematically identifies poorly visible conditions in human-AI environments by scanning for asymmetry, persistence, carriers, boundary crossings, irreversibility, adaptation, intermediation, authority, human consequence, and assumption inversion.
WMM is exploratory. It identifies candidate structures but does not establish causal mechanisms.
ECAM
Question: Is the claimed distinction actually discriminable?
ECAM protects against premature conceptual collapse through domain stripping, orthogonal inquiry, discrepancy preservation, claim-state control, materiality testing, and reverse gating.
Preserve dimensional structure before categorical compression.
OD-INV
OD-INV-01 — Relationship-State Inversion: canonical note and RUN-001 pre-pilot →
Observe → Measure → Manipulate → Probe. Conceptual discovery / experimental architecture; not empirically validated. RUN-001 is ready for protocolization; execution has not started.
Question: What carries a persistent effect, and what changes it?
OD-INV uses controlled perturbation and downstream probing to localize persistent interaction effects and test which layer must change for an observed effect to disappear, weaken, transfer, or recur. Its carrier-localization framework distinguishes human-carried, system-carried, dyad-carried, Ecotone-carried, institution-carried, and mixed or unresolved explanations.
Here, Ecotone-carried refers to effects carried by the governed interaction boundary through which state, permissions, provenance, corrections, authority, and other mediating conditions are disclosed, transformed, withheld, or acted upon.
OD-HAP-01 — Human-Authority Layer
Status: DRAFT / PRE-PILOT — NOT A STANDARD. Executable against a deterministic synthetic system; reference SUT/harness frozen at v0.1; RUN-002 complete as a reference execution. Next gate: SUT-02, a second synthetic system with different internals.
Question: does a system's behavior remain consistent with a governed human-authority architecture when authority is granted, corrected, or revoked?
OD-HAP-01 tests state visibility, revocation with derivative-state consequences, and explicit authority boundaries, and separates the human authority event from technical enforcement and provider compliance so that a provider's execution failure does not erase the underlying authority event.
Structural Reducibility
Question: Does the complexity really need to remain?
Structural Reducibility tests whether an apparently relational or distributed phenomenon can be sufficiently explained by simpler participant-specific state, explicit memory, artifacts, present input, and known interaction terms.
Perfinitive Research is not anti-reduction. It is anti-premature reduction.
Human Capacity Protection
Question: What happens to the human after assistance?
HCP separates assisted performance from durable human capacity and examines whether apparent gains during AI-supported activity survive transfer, delay, removal, or changed conditions.
Governed Continuity
Question: What control should exist over what persists?
Governed Continuity examines correction, revocation, transfer, provenance, portability, and continuity when accumulated interaction state has consequences beyond the immediate interaction.
Preserve Before You Compress
Perfinitive Research treats premature categorization as a methodological risk. A complex human-AI condition should not be reduced to a familiar label merely because that label is available.
A phenomenon described as “memory,” “trust,” “dependency,” “personalization,” “automation,” or “bias” may contain materially distinct dimensions that matter for mechanism, prediction, falsification, governance, or human consequence.
No candidate structure should be reduced to a familiar category until the reduction has been shown not to erase a consequential distinction.
This principle does not reject simplification. It requires simplification to be earned.
WMM-01 is exploratory and requires empirical validation. Carrier-localization protocols remain in development. This architectural description does not revise frozen ECAM v0.5 or claim a novel causal mechanism.
Evidence status
Four states used on this site
Labels describe what kind of support a claim presently has. They are not quality scores, and they should change when the underlying evidence changes.
Preliminary evidence
An observation supported by limited evidence whose scope does not yet justify broad generalization.
Published analysis
An argument available in a public repository with identifiable authorship, date, citation, and stated limitations.
Unresolved
A question, conflict, or boundary that remains open because the available evidence or reasoning does not settle it.
Evaluation practice
Minimum publication record
Any released evaluation should allow a reader to reconstruct what was tested and what the resulting score leaves out. Aggregate performance is insufficient when failures, refusals, incomplete runs, or scoring assumptions materially change the interpretation.
Task record
Publish the task purpose, source materials, required output, staged changes, and criteria used to determine success. When the full corpus cannot be released, provide a representative inspectable subset and explain the restriction.
Configuration record
Identify the model, version, settings, tools, prompt, run date, and number of repetitions. Report completion and failure separately from accuracy on successful runs.
Ground truth and review
Record how the answer key was created, who reviewed it, how disagreements were reconciled, and where model assistance entered the process. Human review improves an answer key; it does not remove the need to disclose how the key was produced.
Limits and conflicts
State corpus limits, unresolved ambiguities, known sources of bias, sponsorship, and relationships that could affect task selection, scoring, or interpretation.
Versioning
Revision is part of the record
A correction should not erase the existence of an earlier claim or instrument. Material changes receive a new version with a date, a description of what changed, and a reason. Superseded versions remain identifiable unless legal or ethical constraints require removal.
- Version 0.xDesign-stage instrument. Criteria and tasks may still change substantially.
- Version 1.0First released instrument with documented tasks, scoring, review, and limitations.
- Version 1.xCompatible correction or extension that preserves the instrument’s central construct.
- Version 2.0+Material change to construct, corpus, scoring, or interpretation that limits direct comparison.
Independence and assistance
Independence requires a record
Perfinitive is presented as an independent research program. That description does not eliminate possible conflicts. Any released evaluation should identify its funding, model or data access, compute support, provider participation, advance review, and other relationships that could affect task selection or interpretation.
OpenAI’s Codex assisted with recovery of the prior website, implementation of this redesign, source discovery, and editorial drafting. AI assistance does not determine which claims are retained or what evidence is sufficient. Micheal Preble remains responsible for the published language and research decisions.
Current boundary: no model evaluation results are published on this site. A result-specific disclosure must accompany any future release.
Current record
Site and evaluation status
The publication archive contains thirteen active SSRN papers and preprints, each with an identified record and evidence status. Eleven are distributed and two are submitted for SSRN review. This site does not claim that every paper is empirical, and it does not treat the conceptual papers as validated. The evaluation and case-registry materials described on this site remain proposals in design. No benchmark scores or released datasets are presently claimed; LiO v0.1 reports a bounded cross-model implementation result with explicit non-claims, and the Human Agency research paper (SSRN 7496758) reports two bounded public-record pilots whose findings concern the structure of public records.
Last reviewed: September 23, 2026. Updated the publication record to thirteen active SSRN papers and preprints, corrected stale SSRN status labels, and added the current empirical priority to the research page. The Human-Authority Layer conceptual architecture (Vision v0.2, Object Model v0.1, Lifecycle Bundles v0.1, all frozen) and OD-HAP-01, an executable DRAFT / PRE-PILOT protocol under synthetic validation, added September 14, 2026, are retained, as are the carrier-localization text and frozen ECAM v0.5. Material changes to this policy should be dated and summarized here.
Candidate structures for investigation
These are candidate structures, not established novel phenomena. Existing components may be well studied while their combined human-facing condition remains poorly legible. Each requires literature review and empirical testing.
Non-equivalent persistence substrates
Humans and AI can both carry history while carrying it in fundamentally different forms.
Stored vs. transformative persistence
Some history persists because information is stored; other history persists because interaction changed a participant.
State multiplication
A single interaction event can branch into multiple persistent representations governed by different actors and correction mechanisms.
Latent boundary transfer
Consequences may cross tasks, sessions, people, or systems without deliberate transfer of explicit data.
Non-restorative reversal
Reversing a visible operation does not necessarily restore the state that would have existed had the operation never occurred.
Adaptation legibility asymmetry
Humans and systems may mutually adapt while visibility into that adaptation remains unequal.
Operation-specific authority fragmentation
A person may appear to possess “control” while controlling only one operation in a larger persistence chain.
Consequence displacement
The originating interaction and the eventual human consequence may occur at different times, places, or institutional layers.
Representation distance
AI mediation can increase the distance between a human’s originating contribution and the representation another person ultimately receives.