forme / Cognitive system

How experience becomes an agent's long-term cognition.

forme stores events as memory with provenance, time, relationships, and outcomes. Long-term evidence enters UserModel, AgentSelfModel, PartnershipModel, and CognitiveMap, then informs later judgment through a governed update protocol.

EXPERIENCE
EVIDENCEforme
JUDGMENT

Cognitive formation

Memory preserves experience. Cognition interprets it. Thought organizes decisions.

The three chains share evidence and state. In forme, the thought chain means a reviewable decision-formation path. It does not record hidden model reasoning.

01

Memory chain

Preserve traceable experience

Events, facts, relationships, provenance, and time form a memory graph with recall, association, decay, retraction, and replay.

Real eventProvenance and timeMemory graphContext recall
02

Cognition chain

Form revisable understanding

Verified experience forms persistent models of the user, the agent, their partnership, quality, risk, and blind spots.

Outcome evidenceCandidate judgmentConflict checkCognitive model
03

Thought chain

Form reviewable decisions

Cognition enters goals, situations, resources, completion criteria, and autonomy boundaries to form action intent and a verification path.

Goal and situationStrategy and resourcesAction intentVerified outcome

Individual cognition

Four models maintain a long-term understanding of the user, the agent, and their relationship.

The models correct one another while retaining provenance, counterevidence, and retraction paths. Individualization develops through long-term evidence. A single interaction creates candidates but cannot directly alter the stable structure.

UserModel01

User model

Goals, preferences, behavioral patterns, real constraints, and risks.

AgentSelfModel02

Agent self model

Capabilities, reliability, limits, confidence, and capability gaps.

PartnershipModel03

Partnership model

Roles, trust, correction, authority, and handback conditions.

CognitiveMap04

Cognitive map

Quality, causality, risk, blind spots, and judgment frames.

Governed update

Candidate, review, promotion, and retraction form the cognitive update protocol.

New judgments enter a candidate layer first. Provenance, Conflict, Impact, Confidence, and Counterevidence determine promotion, decay, retraction, or rollback.

01Observation and outcome
02Cognitive candidate
03Provenance and conflict check
04Evaluation and promotion
05Stable model or retraction

Co-evolution

Fit the user while retaining independent judgments of competence and risk.

forme learns the user's goals, preferences, behavior, and real context through sustained interaction. AgentSelfModel continually calibrates capability. CognitiveMap retains judgments of quality, risk, and blind spots. Together, these models define the autonomy range, collaboration points, and conditions for handing control back to the user.

01Intent and context understanding
02Agent capability calibration
03Role and authority adjustment
04Quality and risk judgment
05Handback at critical exceptions

forme / cognition

Understanding the user, calibrating the agent, and preserving correction form a long-term relationship.