Map how AI alters the human mind.
TSOT converts empirical HCI discoveries into structured, actionable guidelines for engineers, designers, and founders building AI-powered products.
The Source of Truth (TSOT) is a research registry designed to map how current AI interface structures alter human cognition, memory recall, planning, and belief calibration.
Live at https://sourceoftruth-three.vercel.app/. Unlike typical AI databases focused on model benchmarks or generic developer tooling, TSOT works from the outside in. We track the empirical effects of agentic design systems on the human mind, converting scientific discoveries into structured, actionable guidelines.
Every audit maps to one of four empirical research pillars.
Deals with technologies that erode human prospective memory, structural planning, logical reasoning, and long-term retention. Evaluates the long-term impact of delegating cognitive tasks to machines.
Measures interface interventions that interrupt automated habits. Focuses on productive friction paradigms that calibrate user expectations and restore active verification loops.
Evaluates response timing, system latencies, and output pacing. Researches how real-time streaming and immediate reply structures affect human turn-taking thresholds and anthropomorphism.
Studies how generative search formats, personalized sentiment mapping, and contextual source filtering change how users establish truth, review evidence, and shape beliefs.
Every record cites the source. No AI opinion without empirical grounding.
Our registry queries OpenAlex using specific target keywords. Abstracts are parsed through Gemini Flash via a strict schema designed to convert complex statistical data into accessible, human-centric findings.
To guarantee integrity, our editorial prompts forbid corporate cheerleading or alarmist framing. If a paper abstract fails to report clean, empirical metrics, it is flagged as such. Every registry record links directly to its original scholarly publication so users can cross-validate all findings.
Retrieval-augmented generation grounded in real science.
The MCP audit engine uses in-memory BM25 full-text indexing to query the registry. When you prompt your editor, it retrieves the most relevant research records as grounding context before generating any compliance finding.
It then conducts a strict comparison: analyzing where your architecture presents risks against empirical HCI findings, and supplies explicit, sprint-actionable verdicts with mandatory statutory citation tags like [Art. 50(1)].
Explore the Research Ledger.
Browse 9,200+ indexed HCI papers organized by cognitive pillar and EU AI Act statutory article.