Key terms in Verifiable Cognitive Intelligence, AI authorship accountability, and cognitive traceability. Each term links to the RedInk Paper where it was introduced.
The capacity of a system to generate, preserve, and make available for third-party review a contemporaneous record of the cognitive process by which a human author produced work with AI assistance. VCI is not a claim about the quality of AI assistance, or about whether AI use is appropriate. It is a claim about traceability.
The structured evolution of human reasoning, supported and made explicit through AI systems. This definition departs from prevailing AI paradigms by prioritizing process over output, transparency over opacity, and collaboration over substitution.
A four-layer model for understanding where professional accountability is located in AI-augmented work. The layers are: Human Intent (goals and constraints), Cognitive Process (reasoning steps and assumptions), AI Interaction (prompts and responses), and Output (the final artifact). Current AI systems capture Layer 4 and partially expose Layer 3. Layer 2 remains invisible.
VCI Criterion 1. The requirement that a system must distinguish and record which reasoning elements originated with the human author versus the AI assistant, at the point of generation, not reconstructed afterward. Each discrete event in the session record must carry an explicit attribution: human-originated, AI-originated, or human revision of AI output.
VCI Criterion 2. The requirement that the audit record must be generated contemporaneously with the work. Timestamps must be server-side. Each event must include a cryptographic hash of the preceding event, making any alteration detectable. The system must prevent retroactive record creation from finished documents.
VCI Criterion 3. The requirement that the system must capture not just what was produced but the iterative reasoning states that produced it, including prompt evolution, discarded AI outputs, rejected directions, and document deltas.
VCI Criterion 4. The requirement that the audit record must be structured so authorship claims can be verified by a third party without reliance on self-attestation alone. The record must include a unique verifiable identifier, a session integrity hash, and sufficient metadata to withstand institutional scrutiny independent of platform access.
Embedded cognitive structures within AI guidance modules that constrain AI behavior from the inside. Rather than filtering outputs after generation (guardrails), integrity kernels shape the reasoning process itself. They ensure the AI asks questions rather than provides conclusions, diagnoses rather than resolves, and returns decisions to the author.
InkTrail is RedInkAI's AI audit trail system for professional writing. It records every reasoning step, documents every decision point, preserves revisions and the alternatives they replaced, and captures the evolution of ideas from initial intent to final output. Every event is hash-chained using SHA-256, source-labeled (human, AI, or human edit of AI), and timestamped server-side. Where traditional version history records what changed, InkTrail records why. InkTrail is the evidentiary foundation for verifiable AI writing and AI output verification.
An approach to AI assistance that structures and supports human reasoning rather than generating content. A cognitive scaffolding system asks diagnostic questions, surfaces assumptions, stress-tests arguments, and returns decisions to the author. The author does the cognitive work; the system makes that work more rigorous and traceable.
The point at which AI contribution no longer supports human reasoning, but displaces it as the primary source of intellectual authorship. Identified as a normative question that requires domain-specific engagement and cannot be resolved by infrastructure alone. VCI provides the precondition for threshold-setting by generating cognitive records with sufficient resolution to distinguish between human reasoning and AI contribution.
VCI Criterion 5. The requirement that the system must define and log, for each session, what AI assistance was permitted and what constraints were in force, including active guidance configurations, domain restrictions, and operational parameters. The scope must be captured as it was at the time, not as currently configured.
VCI Criterion 6. The requirement that where a system includes cognitive scaffolding modules, each invocation must be logged with sufficient context to reconstruct the reasoning arc: which module was used, what the author prompted, what it returned, and whether the author adopted, revised, or disregarded the guidance.
VCI Criterion 7. The requirement that the audit record must be exportable in a format interpretable by parties outside the originating platform, including regulatory bodies, legal proceedings, and institutional review. The export must include structured, machine-readable fields as well as human-readable narrative, and must be self-explanatory to a reviewer who has never used the system.
The condition in which AI systems are evaluated, adopted, and governed primarily on the basis of output quality, while the integrity of the cognitive processes that produce those outputs is systematically neglected. Output quality is measurable and visible; cognitive process integrity requires infrastructure that does not currently exist to make it visible.
A tamper-evident, chronological record of every interaction between a human author and an AI system during the creation of a work product. Unlike simple chat logs or version history, an AI audit trail uses cryptographic hash chaining to ensure immutability, source labeling to distinguish human from AI contributions, and adoption tracking to record what the author did with AI outputs. RedInkAI's InkTrail is an implementation of this concept.
AI-assisted writing that can be independently verified by a third party. This requires more than disclosure ('I used AI'). Verifiable AI writing provides structural evidence: hash-chained event logs, source separation, decision records, and exportable attestation reports that allow reviewers to reconstruct the authorship process without relying on the author's self-report.