The Problem
What an AI-assisted gain leaves behind, and the three questions that follow it.
Start here →Perfinitive Research
Human-first research across AI-mediated conditions
Perfinitive Research is the broader research home. Operant Dyad™ is a major program within it, focused on persistent human-AI interaction.
Start by question
Six ways into the same program
Each area below is a short routing page into the papers and project pages that develop it. They do not replace the research areas below, which remain the complete map.
Persistent Relationships
How a task-bound working relationship forms and what it retains.
Start here →Human Authority
What people must remain able to see, correct, govern, carry, and retire.
Start here →Human Capacity
Whether an AI-assisted gain becomes durable human capacity.
Start here →Consequence
What continues to shape later decisions after the interaction ends.
Start here →Evidence
What kind of support a claim on this site currently has.
Start here →Living register
Research questions we are thinking about
Open questions across 14 areas, from the definition and measurement of Relationship State to human capacity, governance, portability, and future extensions. Inclusion identifies a problem worth investigating; it does not establish an answer or claim ownership.
Active research queue
OD-INV-01-RUN-001 — Relationship-State Inversion Pre-Pilot
READY FOR PROTOCOLIZATION / PRE-PILOT · Experiment not started.
Establish matched histories, manipulate one governed interaction parameter, present an identical target task and downstream probe, then measure trajectory and terminal behavior. Apply correction or revocation and test recovery and propagation.
Next action: Turn the minimal experiment into a controlled, reproducible protocol.
Conceptual discovery / experimental architecture; not empirically validated. No causal claim and no ECAM v0.6 architectural change unless the effect is reproducible under controlled conditions. ECAM v0.5 stays frozen.
Foundational Research
Operant Dyad™ and Emergent Novelty examine persistent collaboration and pairing-sensitive explanations.
Human-Authority Layer
Conceptual architecture — Vision v0.2 frozen. OD-HAP-01 is an executable pre-pilot protocol under synthetic validation, separating human authority from provider enforcement and compliance.
Human Capacity and Measurement
HCP, Beyond Performance, and the Human-Development Conversion Constraint separate assisted performance from durable human capacity.
Persistent State and Governance
Governed Continuity and the legal research examine correction, revocation, provenance, transfer, and portability.
Methods and Measurement
WMM-01, ECAM, OD-INV, Structural Reducibility, and carrier-localization protocols answer different questions and preserve distinct evidentiary roles.
Policy and Institutional Work
The policy brief and economic measurement design examine how human-first measurement can inform institutional decisions.
Research areas
The work spans foundational research, human capacity and measurement, persistent state and governance, methods, and policy. These areas share methods without being forced into one theory.
Operant Dyad™ program · Methods architecture · Publications by research area
Research program
What continues, and what changes around it?
Persistent AI raises connected questions about governed relationship state, the conversion of assisted performance into human capability, and the economic value produced around that relationship.
The continuity problem
A correction that improves tomorrow’s answer can later become stale. A permission valid in one system can be treated as though it survived migration into another. These failures differ in their legal and technical details, but they share a structural source: relationship state has become consequential without an adequate account of its provenance, validity, correction, and authority.
My work examines the configuration required when retained state begins to shape later interaction. The central questions concern what must be recorded, what must remain contestable, what may travel, and what has to be reevaluated before a receiving system acts.
Program structure
A visible sequence across the research
The sequence keeps distinct the persistent relationship, human advancement, durable capacity, authority, provenance, and the methods needed to test them.
Status: HCP, CDWP, individuation, and OD-ID-01 (Inverse Dyad) identify developing parts of the program rather than settled publication records. RDOS (relationship-derived operational state) is the subject of When Collaboration Becomes State (SSRN 7404838), a distributed conceptual paper whose testable conditions have not been validated. OD-ID-01 (Inverse Dyad) is a distinct research branch from OD-INV-01 (Relationship-State Inversion) and does not yet have a canonical public page.
Three programs
How the thirteen active records group
Thirteen active SSRN papers and preprints: eleven distributed, two submitted for SSRN review. Arrows show program structure, not proven causal dependence.
Persistent human–AI state and authority
Operant Dyad → Persistent Primitive → Authority Does Not Travel by Default → Governing Relationship State → Human Agency Under Consequential State (research paper; the policy paper is a separate record) → When Correction Cannot Catch the System → empirical correction / authority testing
Human development and flourishing
Human-Development Conversion Constraint → Beyond Performance → longitudinal human-capacity conversion evidence
Methods, causal structure, and validation
Emergent Novelty → LiO → The Water We Swim In → replication, corpus validation, and external raters
Research connections
Connections across the research areas
These papers examine connected questions about persistent interaction, authority, human capacity, and economic consequence. Their relationships help organize the inquiry; they do not make every publication part of a single theory.
Layer 01
Relationship and accumulated state
The Operant Dyad defines the configured working relationship as an object of study. The Persistent Primitive identifies the state that accumulates through that relationship and the requirements for keeping continuity under human governance. Emergent Novelty asks whether some valuable output may arise from the interaction dynamics of a particular pairing; its evidence remains preliminary. When Collaboration Becomes State asks when a persistent configuration shaped by repeated interaction should be attributed to a specific human participant.
Layer 02
Authority over continuing state
Authority Does Not Travel by Default isolates the problem created when useful state crosses a technical or institutional boundary. Governing Relationship State develops the broader account of lawful continuing reliance and the duties of actors able to make that state consequential. Human Agency Under Consequential State (research paper) and its policy paper follow the corrected state into the consequences that depend on it, and When Correction Cannot Catch the System asks whether an accepted correction still reaches the reliance paths that remain.
Layer 03
Evidence of human conversion
The Human-Development Conversion Constraint establishes the proposed evidentiary boundary: increased artificial capability, adoption, or output does not establish human advancement. Beyond Performance turns that boundary into a diagnostic method for testing a specific claim.
Layer 04
Human-centric economic measurement
Human-Centric Economic Impact brings the earlier layers into the economic account. When AI produces a measurable gain, the framework asks where the capability resides, what remains when the configuration changes, who controls the accumulated state, and how the resulting value and dependency are distributed.
Current status: This final layer is a proposed synthesis. The earlier papers support its conceptual structure; they do not yet establish a broad empirical economic effect.
Layer 05
Environmental partition
The Water We Swim In proposes a candidate provenance-and-control-surface partition between inherited human conditions and engineered synthetic conditions in persistent human-AI interaction, restricting Ecotone to observable, activated boundary processes between them.
Current status: Candidate architecture, not validated. The paper carries its own falsifier and an adversarial pre-pilot gate (OD-ENV-01), held at the provenance gate.
Inference control
LiO adds a narrow computational layer.
LiO v0.1 asks what downstream assertion permissions remain after every indispensable inferential bridge in a declared derivation has been considered.
What assertion classes and distinctions belong in the universe?
What does each indispensable bridge permit, exclude, or leave unresolved?
Given those inputs, what authorization survives every indispensable bridge?
Scope: LiO v0.1 addresses composition. It does not claim to solve representation or adjudication, and it does not establish domain validity or universal correctness.
Analytical signature
The corpus separates layers that are often collapsed.
The recurring move is diagnostic: the visible persistence or performance of one layer does not settle the status of another.
- Current outputis not the same ashuman capacity
- Technical persistenceis not the same aspresent validity
- Possessionis not the same asauthority
- Portabilityis not the same asauthorization
- Historical authorityis not the same ascurrent authority
- Correctionis not the same asdownstream correction
- Record retentionis not the same ascontinuing reliance
Status: This is a recurring analytical pattern across the corpus, not a named or validated general theory.
Continuity sequence
How the first inquiry has moved
Each paper changes the object under examination. The sequence begins with a bounded working relationship, then follows the state produced by that relationship into persistence, transfer, lawful reliance, and possible emergence.
The working relationship
Operant Dyad proposes a task-bound relationship formed through repeated correction between one human and one functionally bounded adaptive system.
Human-governed continuity
The Persistent Primitive distinguishes retained state from legitimate use and derives requirements for correction, revocation, purpose, and temporal validity.
Legal provenance
Authority Does Not Travel by Default asks what changes when relationship state crosses a provider, application, role, or jurisdiction.
Continuing reliance
Governing Relationship State asks which actors must reevaluate authority when persistent state remains available for consequential use.
Emergent novelty
Emergent Novelty proposes a way to study output attributed to the interaction dynamics of a particular pairing. Its evidence remains preliminary.
Current empirical priority
Testing whether a correction reaches what depends on it
The immediate empirical program tests correction reach, also called temporal adequacy. Its core question: can a computational system accept an authoritative correction to consequential personal state while materially consequential reliance paths continue to operate on superseded state?
Status: The first experiment is still in design. No empirical result has been produced from it. This is separate from the two public-record pilots reported in the Human Agency research paper.
Frozen queue
- Correction Reach / Temporal Adequacy experiment
- Relationship State causal test
- Human Development conversion study
- Independent LiO replication
- Water boundary validation
- Governance primitive ablation
The program is now prioritizing tests capable of producing negative as well as positive evidence. A null result can narrow or defeat a stronger claim.
Open questions
What the papers do not yet settle
The frameworks require outside testing, contrary cases, and further legal and empirical development. Several questions now guide that work.
Validity over time
How should a system determine when retained state has become stale, context-bound, or unsafe to reuse?
Authority after transfer
Which provenance fields allow a receiving system to reevaluate whether state remains operable in its new context?
Correction in practice
What must a person be able to inspect, contest, supersede, or revoke for continuity to remain meaningfully governed?
Inference boundaries
How can systems preserve useful context without allowing prediction to become an uncorrectable identity record?
Measurement
Which observable failures distinguish retrieval quality from continuity that remains legitimate and under human judgment?
Replication
Can proposed measures of dyadic novelty survive independent testing across people, models, tasks, and time?
From Research to Policy
Perfinitive Research is developing a policy-translation layer that converts stable research findings and bounded conceptual work into decision-facing materials without changing their evidentiary status.
Research → Policy Interface → Policy Brief → Proposal Module → Institutional Proposal
Proposals are derivatives of the research program, not drivers of it.
The policy layer distinguishes what research establishes, what remains conceptual, who holds decision authority, available policy levers, measurable outcomes, and the conditions under which an action should be reversed or halted.
Consistent with that separation, the research program is being mapped to international governance and human-AI collaboration questions; the underlying research remains upstream of any specific proposal.
Next stage
From framework to evaluation
The first public design draft converts authority transfer into staged tasks and draft outcome categories. A second draft defines an economic measurement layer that separates assisted output from durable human capability, governed extension, dependency, and value capture. Neither instrument reports results.