Frameworks / CRF
CRF Methodology
How the Cognitive Risk Framework measures cognitive cost across domains.
Mission
Risk Awareness as a Product
The Cognitive Risk Framework (CRF) is a standardized instrument for measuring the cognitive cost a digital system extracts from its users. It is domain-agnostic: the same five dimensions apply to trading, gambling, gaming, adult content, AI chatbots, and social media. What changes between domains are the weights and calibration parameters.
CRF is not a moral judgment. It is not a clinical assessment. It is a structured measurement instrument that makes invisible behavioral patterns legible and actionable.
Scoring
The formula
totalScore = EI*wEI + RLS*wRLS + ER*wER + ID*wID + AD*wAD
finalScore = clamp(totalScore * AIM, 0, 100)
- Inputs - 0-100 per dimension (self reported or computed from raw data)
- Weights - domain-specific, must sum to 1.0
- AIM - AI Influence Multiplier (1.0-2.0) scales final score based on platform AI personalization depth
- Output - integer 0-100, clamped
Dimensions
What is measured
EI
Exposure Intensity
Measures how frequently and deeply users are exposed to the system's feedback loops. Inputs include time spent per day, financial exposure as a percentage of total resources, frequency of engagement, and leverage involved. Higher EI means more surface area for cognitive impact.
RLS
Reinforcement Loop Strength
Measures the strength of variable reward schedules. Variable and uncertain rewards increase persistent responding. Inputs include reward variability, streak mechanics, notification pressure, and social validation loops.
ER
Escalation Risk
Measures the degree to which intensity is increasing over time. Inputs include session-over-session intensity slope, loss-chasing events, and content escalation patterns.
ID
Identity Drift
Rooted Reality's differentiator. Measures the degree to which the system reshapes the user's sense of self. Inputs include self-worth tied to system outcomes, emotional volatility from system feedback, loss of offline grounding, and narrative absorption in system identity.
AD
Autonomy Degradation
Measures reduction in voluntary disengagement capacity. Inputs include difficulty stopping voluntarily, AI-suggested actions adopted without deliberation, skill atrophy from automation, and dark patterns that make exit difficult.
AIM
AI Influence Multiplier
Scales the total score based on how much AI personalization the platform employs. A platform with no algorithmic personalization uses AIM = 1.0. A platform with deep AI-driven engagement optimization can use AIM up to 2.0.
Domains
Domain adaptability
Each domain uses the same 5-dimension skeleton with different weights, AIM calibration, risk thresholds, and boundary suggestion templates. Adding a domain requires only a config file; no engine changes.
Trading: ER weight = 0.30 (escalation is the primary risk)
Gambling: RLS weight = 0.30 (variable reward is the defining mechanic)
Gaming: RLS weight = 0.30 (loot boxes, daily rewards)
AI Systems: ID weight = 0.25 (anthropomorphization, identity reflection)
Social Media: EI weight = 0.25 (ambient availability, notification pressure)
Guardrails
Ethical boundaries
No diagnosis. CRF flags pattern signals. It does not label users with disorders or clinical conditions.
No moral judgment. Domains including adult content and gambling are described analytically. Output language uses posture and signal wording.
No investment advice. The Trading Copilot enforces user-authored rules. It does not recommend securities, entries, or exits.
No medical claims. Boundary suggestions are self-directed tools, not prescriptive treatment plans.