A browser-based forensic typing analysis framework to distinguish genuine human cognition from AI-assisted transcription.
ThoughtTrace Pro* is a browser-based forensic typing analysis framework designed to distinguish genuine human cognition from AI-assisted transcription.
Unlike traditional NLP-based AI detectors (which analyze the final text), ThoughtTrace analyzes the behavioral biometrics of the writing process.
Current AI detection tools fail when users manually transcribe AI-generated text or perform heavy "AI-assisted editing." They also struggle with false positives, accusing genuine students of cheating.
ThoughtTrace monitors keystroke dynamics, context switching, and input entropy to create a Session Audit Trail. It doesn't judge the words; it judges how the words were typed.
- Baseline Calibration: Establishes a personalized typing fingerprint (Jitter/CV) before the session begins to prevent device bias.
- Sync Coverage Algorithm: Compares DOM text length with actual keystroke events. Instantly flags massive clipboard injections (Copy-Paste).
- Cognitive Entropy (CV) Analysis: Analyzes the Coefficient of Variation (CV) of keystroke delays. Authentic human thought produces non-linear, rhythmic jitter. Manual transcription of AI text produces mechanical, linear input.
- Focus Loss Tracking: Monitors tab-switching and window blur events to detect "Synthesizer" behavior (reading from an external AI window).
This is a zero-dependency, local-first vanilla JavaScript prototype.
- Clone the repository.
- Open
index.htmlin any modern browser. - Complete the Initial Calibration to unlock the secure buffer.
100% Local Processing. ThoughtTrace Pro relies on localStorage and client-side processing. No keystroke data or written content is ever sent to a remote server.
Generates a downloadable PDF Forensic Audit Report containing a timeline of behavioral shifts (e.g., [CRITICAL_PASTE] followed by [HUMAN_RECOVERY]), providing educators with context rather than just a binary score.