Dependency Atlas — AgentSee Research Notebook

←AgentSeeResearch Notebook

Navigate

Notebook home Dependency atlas

Filter by type

Design Axioms

A0: Biological Realism A1: Living Knowledge Base A2: Individual Model Construction A3: Role Separation A4: Capacity as Terminal Objective A5: Caring Governance A6: Anti-Exploitation

Constructs

Agency Autonomy Backfire Regime Capacity Caring Coherence Controllability Empowerment Intrinsic Motivation Plant Model Positive Interdependence Wanton State

Biological Mechanisms

Acute Stress Effects on Executive Function Catecholamine-PFC Dynamics Controllability Circuit DA Source-Level Gain Control Hopelessness and Controllability: Computational Model LC-NE Adaptive Gain Neuromodulation of Thought Stress-Sensitive Controllability Inference

Integrative Mechanisms

F1: Capacity for Self-Directed Action Is State-Dependent F2: State Changes Exceed Unaided Human Perception F3: Controllability Detection Has Specific Neural Circuitry F4: Controllability Inference Is Itself State-Dependent F5: DA Output Is Gain-Controlled at the Source

Architecture

D1: Human-as-Controller, AI-as-Observer/Stabilizer D2: Capacity as Terminal Objective (Meta-Capability Defense)D3: Caring Constraints as Normative Orientation D4: Two-Layer Machine (Understanding + State-Estimation)

Design Principles

Content-to-Capacity Shift Intrinsic Motivation Restoration Productive Incoherence Distinction Recursion Resolution State-Conditioned Gating

Invariants

I1: No Hidden Goal Substitution I2: Volitional Control Preserved I3: Anti-Capture I4: Dependency Minimization I5: Privacy and On-Device Default I6: Bounded Actuation and Escalation Discipline

Protocols

Actuator Categories Exemplar Interaction Traces Regime Table (State-Conditioned Gating)

Assessment

Measurement Stack Time-to-Regulation

Constraints

Cost/Access Constraint Principal-Agent Threat Model Wrong Understanding Failure Mode

Models

Coherence-Agency Bootstrap CRR: Candidate Temporal Formalism for Regime Transitions Three-Target Optimization

Predictions

P1: Backfire Regime Exists P2: Regime Observability P3: Stabilization Without Autonomy Loss P4: Understanding Layer Adds Value P5: Gains Persist After Removal

Experiments

E1: Observability and Calibration E2: Micro-Randomized Stabilization Trial E3: Semantic Layer Ablation E4: Removal Test E5: Negative Tests (Project-Killers)

Kill Conditions

6.1: Plant Model Impossibility 6.2: AI Understanding Plateau or Regression 6.3: Objective Function Philosophically Untenable 6.4: Irresolvable Paternalism 6.5: Competitive Architecture Achieves Equal Outcomes 6.6: Caring Reduces to Paternalism or Persuasion 6.7: Engagement/Compliance System Matches Outcomes 6.8: Understanding Layer Produces Net Harm

Open Problems

Actuator Systematization Agency Envelope Definition Measurement Feasibility Multi-Agent Architecture Plant Model Plant Model Program Value Calibration Over Time

Literature Positioning

Active Inference for HMI (Schoeller et al. 2021)Active Inference Psychotherapy and Dyadic Inference Adaptive Automation / Augmented Cognition (DARPA, NASA)Behavioral Health Control Engineering (Rivera/Hekler/Murphy, 2007-2026)Cybernetics and Its Modern Descendants Exploitative Technology Design Hostile Scaffolding Literature (4E Cognitive Science)Inner Screen Model (Ramstead et al. 2023)IWMT (Safron 2020, 2022, 2026)LLM + Wearable Personal Health Systems (2024-2026)Personal Informatics (Li et al. 2010)Philosophical Foundations of the Objective The Reference Class Problem Rehabilitation Robotics and Assistive Shared Control Self-Determination Theory

Literature Gaps

What Kind of Contribution This Might Be What Doesn't Exist and Why

Navigate

Notebook home Dependency atlas

Filter by type

Design Axioms

A0: Biological Realism A1: Living Knowledge Base A2: Individual Model Construction A3: Role Separation A4: Capacity as Terminal Objective A5: Caring Governance A6: Anti-Exploitation

Constructs

Agency Autonomy Backfire Regime Capacity Caring Coherence Controllability Empowerment Intrinsic Motivation Plant Model Positive Interdependence Wanton State

Biological Mechanisms

Acute Stress Effects on Executive Function Catecholamine-PFC Dynamics Controllability Circuit DA Source-Level Gain Control Hopelessness and Controllability: Computational Model LC-NE Adaptive Gain Neuromodulation of Thought Stress-Sensitive Controllability Inference

Integrative Mechanisms

F1: Capacity for Self-Directed Action Is State-Dependent F2: State Changes Exceed Unaided Human Perception F3: Controllability Detection Has Specific Neural Circuitry F4: Controllability Inference Is Itself State-Dependent F5: DA Output Is Gain-Controlled at the Source

Architecture

D1: Human-as-Controller, AI-as-Observer/Stabilizer D2: Capacity as Terminal Objective (Meta-Capability Defense)D3: Caring Constraints as Normative Orientation D4: Two-Layer Machine (Understanding + State-Estimation)

Design Principles

Content-to-Capacity Shift Intrinsic Motivation Restoration Productive Incoherence Distinction Recursion Resolution State-Conditioned Gating

Invariants

I1: No Hidden Goal Substitution I2: Volitional Control Preserved I3: Anti-Capture I4: Dependency Minimization I5: Privacy and On-Device Default I6: Bounded Actuation and Escalation Discipline

Protocols

Actuator Categories Exemplar Interaction Traces Regime Table (State-Conditioned Gating)

Assessment

Measurement Stack Time-to-Regulation

Constraints

Cost/Access Constraint Principal-Agent Threat Model Wrong Understanding Failure Mode

Models

Coherence-Agency Bootstrap CRR: Candidate Temporal Formalism for Regime Transitions Three-Target Optimization

Predictions

P1: Backfire Regime Exists P2: Regime Observability P3: Stabilization Without Autonomy Loss P4: Understanding Layer Adds Value P5: Gains Persist After Removal

Experiments

E1: Observability and Calibration E2: Micro-Randomized Stabilization Trial E3: Semantic Layer Ablation E4: Removal Test E5: Negative Tests (Project-Killers)

Kill Conditions

6.1: Plant Model Impossibility 6.2: AI Understanding Plateau or Regression 6.3: Objective Function Philosophically Untenable 6.4: Irresolvable Paternalism 6.5: Competitive Architecture Achieves Equal Outcomes 6.6: Caring Reduces to Paternalism or Persuasion 6.7: Engagement/Compliance System Matches Outcomes 6.8: Understanding Layer Produces Net Harm

Open Problems

Actuator Systematization Agency Envelope Definition Measurement Feasibility Multi-Agent Architecture Plant Model Plant Model Program Value Calibration Over Time

Literature Positioning

Active Inference for HMI (Schoeller et al. 2021)Active Inference Psychotherapy and Dyadic Inference Adaptive Automation / Augmented Cognition (DARPA, NASA)Behavioral Health Control Engineering (Rivera/Hekler/Murphy, 2007-2026)Cybernetics and Its Modern Descendants Exploitative Technology Design Hostile Scaffolding Literature (4E Cognitive Science)Inner Screen Model (Ramstead et al. 2023)IWMT (Safron 2020, 2022, 2026)LLM + Wearable Personal Health Systems (2024-2026)Personal Informatics (Li et al. 2010)Philosophical Foundations of the Objective The Reference Class Problem Rehabilitation Robotics and Assistive Shared Control Self-Determination Theory

Literature Gaps

What Kind of Contribution This Might Be What Doesn't Exist and Why

↑↓ navigate↵ openesc close