Cognitive abundance
Machine cognition sharply lowers the cost of producing candidate arguments, code, hypotheses and other intellectual artefacts.
Research provenance
Preserve the path without inventing another provenance island.
Agalmic Research originally built a local discovery graph before adequately surveying mature provenance standards. That graph remains as historical evidence. New work uses W3C PROV concepts, RO-Crate packaging and CRediT wherever they already solve the problem.
Correction
The local graph is an experiment, not a new standard.
Provenance, contributorship and graph-based scholarly communication already have substantial prior art.
W3C PROV models entities, activities, agents, attribution, derivation and responsibility. RO-Crate packages research objects in machine-readable JSON-LD. CRediT standardizes scholarly contribution roles. MIRA demonstrates graph-based questions, claims and evidence with granular attribution.
Use established provenance vocabulary wherever it fits. Extend only where the research problem genuinely requires information it cannot express.
Standards stack
Reuse first. Extend second.
The layers answer different questions and should not be collapsed into a single project-specific schema.
W3C PROVWho or what generated, used, influenced, derived or was associated with a research entity or activity?
CRediTWhich standardized scholarly contribution roles did a human contributor perform?
RO-CrateHow is the research object, its files and contextual metadata packaged for exchange and preservation?
Agalmic extensionWhat is the claim-level authority state, branch state, handoff need, search boundary, reversal condition or possibility-portfolio status?
W3C PROV's term SoftwareAgent is a provenance category. Using it for an AI system does not imply that the system is a moral person, legal author, epistemic knower or scientific discoverer.
Machine-readable artefacts
The migration is now being exercised session by session.
The repository contains an experimental RO-Crate 1.3 metadata document, a PROV-O-aligned JSON-LD projection for the earlier lineage review, and a separate PROV-O-aligned session bundle for the possibility-portfolio / Active Frontier work.
“PROV-O-aligned JSON-LD” means the documents use PROV ontology terms in JSON-LD. It does not claim conformance to every serialization profile or certification by W3C.
Chain currently under scrutiny
Cognitive abundance → authorial capacity gap → defensible stewardship → separable contribution and authority → epistemic handoff → epistemic uptake scarcity → institutions for routing knowledge.
The novelty review found antecedents for every major link. The chain is retained as a synthesis and research genealogy, not as evidence that Agalmic Research discovered seven new principles.
Possibility layer
Provenance should preserve branches without turning them into obligations.
The Memory Palace records the topology of the search. The Possibility Portfolio records what deserves another move.
Public-safe possibilities now have explicit states such as seed, option, active, handoff, realized, parked and retired. The state is an operating decision, not an epistemic truth. A parked or retired branch keeps its provenance, rationale and potential reversal condition.
Two records
Contribution provenance is not epistemic authority.
Standards can tell us a great deal about contribution and derivation without proving a claim true.
CRediT and PROV should record contribution and process. A separate experimental authority layer records which claims, if any, a person or institution can presently defend and the evidence supporting that status.
ContributionUse CRediT for human scholarly roles where it fits and preserve finer process descriptors only when materially useful.
AuthorityRecord claim-sensitive support such as expert review, empirical evidence, formal verification, reproduction or demonstrated defence. Do not infer it from contribution.
Origin deserves credit. Authority requires warrant.
Session protocol
Capture discovery while it is happening.
The Memory Palace Protocol records the shape of a research search, not hidden model reasoning.
It preserves visible prompts and outputs, sources, artefacts, decisions, branch states, roads not taken, search boundaries, contribution roles and authority changes. The companion Possibility Portfolio prompt captures materially new ideas as options and constrains the Active Frontier.
Historical local graph
Do not rewrite the first map after learning cartography.
The v0.1 local graph predates the standards-first correction. It is preserved because that history is itself provenance. Future projections can map it into PROV rather than pretending the project began with the later design.
exploratory
Authorial Capacity Gap
Artefact sophistication can increase faster than the human curator's ability to understand and defend it.
Principle of Defensible Stewardship
Epistemic authority should not exceed defensibility.
Principle of Separable Epistemic Roles
Epistemic contribution and epistemic authority should be separable.
Epistemic handoff
A structured transfer of candidate knowledge from an initiator toward actors who can frame, validate, verify, develop or realize it while preserving originator credit.
Epistemic Uptake Constraint
As candidate knowledge becomes abundant, the capacity to notice, route, evaluate, validate, integrate and realize valuable contributions becomes scarce.
Institutions for routing knowledge
Knowledge systems need mechanisms for routing low-authority candidate ideas toward scarce expert attention, verification and realization while preventing appropriation, credential capture, flooding and filter monopoly.
Human + machine process
Record events without prematurely settling agency.
Distributed-discovery scholarship makes a flat “human author + AI assistance” statement insufficient. A session may distribute problem framing, candidate generation, significance recognition, validation, integration and realization across different actors.
Human curator: has initiated core questions, directed search, selected or rejected branches, recognized promising connections and made publication decisions.
OpenAI ChatGPT: has materially contributed literature search, candidate generation, terminology, synthesis, criticism, drafting and repository transformations.
PROV can record those activities and associations. It should not be used to smuggle in the philosophical conclusion that either participant alone “discovered” the resulting knowledge.
Research objects
Papers and proposals retain their genealogy.
The local graph still records early contribution reconstruction. New research objects additionally carry standards-oriented provenance, lineage status and, where relevant, possibility-portfolio state.
The Authorial Capacity Constraint
Working Paper 001 on authorship, defensibility and epistemic stewardship under cognitive abundance.
src/pages/publications/authorial-capacity-constraint.astro
Epistemic Handoff: Contribution, Authority and Uptake Under Cognitive Abundance
Draft paper developing separable contribution and authority, handoff, uptake scarcity and institutional routing.
src/pages/drafts/epistemic-handoff.astro
The Defensibility Layer
A proposed open infrastructure layer for recording human/machine contribution, verification and demonstrated defensibility in AI-assisted research.
Memory-palace direction
Eventually, navigate both discovery and choice.
The long-term interface remains a navigable research memory palace.
A reader should be able to enter at a claim and travel backward through sources and activities, sideways into competing branches and objections, and forward into validation, handoff, implementation and correction. A researcher should also be able to see which branches are active, preserved as options, parked or retired, without confusing portfolio status with truth.