Agentra LabsAgentra Labs DocsPublic Documentation

Get Started

Core Concepts

AgenticCognition is built on five foundational concepts that together form a living model of human consciousness. Each concept operates independently but gains power through int...

AgenticCognition is built on five foundational concepts that together form a living model of human consciousness. Each concept operates independently but gains power through interaction with the others.

Living User Model

A continuously evolving representation of a human mind. Unlike static user profiles or preference databases, the living user model has biological properties:

Lifecycle Stages

The model progresses through six lifecycle stages:

StageDescriptionTrigger
BirthModel is created with minimal dataFirst creation
InfancyRapid belief accumulation, high plasticityInitial interactions
GrowthBelief graph deepens, patterns emergeSustained engagement
MaturityStable belief system with crystallized keystonesExtended use
CrisisContradictions force belief restructuringMajor value conflicts
RebirthModel restructures around new core beliefsResolution of crisis

Heartbeat Mechanism

Each interaction records a heartbeat that advances the model. Heartbeats carry context (the topic of interaction), observations (inferred beliefs), and timestamps. The accumulation of heartbeats drives lifecycle transitions.

Consciousness Map

The model maintains a consciousness map showing which cognitive regions are currently active. Regions include values, identity, relationships, work, creativity, and growth. Activity levels shift based on recent interactions.

Belief Physics

Beliefs are not static strings. They have physical properties that govern how they interact, evolve, and resist change over time.

Confidence

A real number from 0.0 to 1.0 representing the strength of the belief. Confidence increases when evidence supports the belief and decreases when counter-evidence is observed. Confidence is not the same as truth -- it measures the model's certainty that the person holds this belief.

Crystallization

As a belief accumulates supporting evidence over time, it crystallizes. Crystallized beliefs are resistant to change. They form the bedrock of identity. Crystallization is measured on a scale from 0.0 (fluid) to 1.0 (fully crystallized).

High crystallization is neither good nor bad. Core values often crystallize naturally. However, rigid crystallization of inaccurate self-assessments can create blindspots.

Entanglement

Beliefs are connected through entanglement links. Two beliefs can be entangled in several ways:

  • Support: one belief reinforces another
  • Tension: two beliefs create cognitive dissonance
  • Derivation: one belief is derived from another
  • Opposition: two beliefs directly contradict each other

The entanglement network forms the belief graph, a living structure that evolves as new beliefs are added and existing ones shift.

Conviction Gravity

Keystone beliefs exert gravitational pull on surrounding beliefs. When a keystone belief shifts, connected beliefs experience stress. Conviction gravity determines the propagation radius: how far a belief change ripples through the graph.

Collapse

When contradictions between entangled beliefs become irreconcilable, a belief collapse occurs. One belief is abandoned (or significantly weakened) and the belief graph restructures around the surviving belief. Collapse is the mechanism for resolving deep cognitive dissonance.

Shadow Psychology

The shadow is the part of the psyche operating below conscious awareness. AgenticCognition detects and maps shadow structures through behavioral analysis.

Shadow Beliefs

Beliefs held unconsciously that influence behavior without the person's awareness. Shadow beliefs are inferred from patterns: when stated beliefs and observed behavior diverge consistently, a shadow belief is likely operating.

Projections

Attributes that a person ascribes to others but which actually belong to themselves. Projection detection works by identifying recurring attribution patterns that contradict the person's self-model.

Blindspots

Gaps between self-assessment and observed behavior. If a person believes "I am patient" but behavioral evidence shows frequent frustration in specific contexts, the gap is a blindspot.

Defended Regions

Areas of the self-concept that are heavily guarded against change. Defended regions show high crystallization, low openness to counter-evidence, and emotional reactivity when challenged.

Bias Fields

Cognitive bias patterns that distort perception and decision-making. Bias fields are mapped spatially across belief domains, showing where systematic distortions are most active.

Emotional Triggers

Specific topics, patterns, or contexts that activate strong emotional responses disproportionate to the stimulus. Triggers are linked to shadow beliefs and defended regions.

Drift Tracking

Longitudinal monitoring of how beliefs, values, and identity shift over time. Drift tracking operates on three timescales.

Short-term Drift

Changes occurring within days or weeks. Short-term drift often reflects mood, context, or recent experiences rather than genuine value shifts.

Value Tectonics

Slow, deep movement of core value systems over months or years. Like geological tectonic plates, values move gradually but can produce earthquakes when they collide. Value tectonics tracks the direction and velocity of these deep shifts.

Growth Rings

Archaeological layers preserving who the person was at each stage. Like tree rings, growth rings are immutable snapshots of past identity states. They enable temporal comparison: who the person was a year ago versus who they are now.

Drift Alerts

When drift magnitude exceeds a threshold, the system generates an alert. Drift alerts notify the person (or their agent) that a significant belief or value shift is occurring.

Prediction Engine

The preference oracle, decision simulation, and future projection systems use the full belief graph to predict reactions, choices, and identity trajectories.

Preference Oracle

Given a question or option, the preference oracle predicts the person's preference based on their belief system, past decision patterns, and value alignment. Predictions include a confidence score and a reasoning chain.

Decision Simulation

Given a scenario with multiple options, the decision simulator evaluates each option against the person's values, beliefs, and behavioral patterns. It produces per-option scores and a recommended choice with explanation.

Future Projection

Using current drift trajectories and value tectonics, the future projection system estimates where the person's identity is heading. Projections are probabilistic and include confidence intervals.