Recursive Thinking Lab
Cognitive Assessment & Drift Detection Platform
Overview
The Recursive Thinking Lab measures recursive cognitive abilities through a 45-item RCSI (Recursive Cognitive Systems Index) assessment across six dimensions. It adapts interactive thinking sessions to the user's cognitive profile using Claude AI, while monitoring for "drift" — the tendency to uncritically accept AI outputs. The platform bridges cognitive science and AI safety, serving both individuals seeking cognitive development and enterprises monitoring team-wide cognitive health.
Key Features
45-Item RCSI Assessment
Six dimensions scored 0-100: Recursive Thinking, Metacognitive Awareness, Systems Thinking, Pattern Sensitivity, Identity Multiplexing, AI Awareness
Three-Phase Lab Sessions
Stability (clarify & surface assumptions), Reflection (identify blind spots), Expression (synthesize into briefing) — AI acts as Socratic coach
Drift Detection Engine
Composite score from 6 signals: sycophancy acceptance, identity blurring, reality testing failures, first-output acceptance, edit frequency, pushback rate
Track System
Three developmental tracks — Exploratory, Recursive Operator, Strategic Partner — with auto-promotion based on RCSI improvement and behavioral signals
Enterprise Dashboards
Org-wide RCSI histograms, drift monitoring, session analytics, member management, audit logs, and CSV/JSON exports
AI-Generated Narratives
Claude-powered second-person profile narratives covering cognitive signature, track position, drift stance, and growth opportunities
Architecture
1Assessment Pipeline
45-item Likert responses are scored across 6 dimensions to produce an RCSI composite (0-100), drift risk band, AI interaction level, and predicted emergence phase.
2Session Architecture
Three-phase sessions with AI phase signaling — the model appends [PHASE_TRANSITION] and [SESSION_COMPLETE] markers that the backend parses to manage progression.
3Drift Detection
Per-session signals (0-1 scale) are aggregated into a composite drift score. Admin sees clinical alerts; users see growth-framed nudges. Longitudinal tracking computes metacognitive health.
4Profile Adaptation
RCSI scores become prompt modifiers — low metacognitive awareness triggers extra scaffolding, high pattern sensitivity with low systems thinking pushes toward causality, elevated drift triggers reality-testing checkpoints.