← Back

Research

Experiments at the intersection of AI ethics, causal inference, and decision intelligence.

Open science (Zenodo DOI)

EthicaAI

Beyond Homo Economicus

Formalizing Amartya Sen's Meta-Ranking theory — preferences over preferences — as a dynamic mechanism in Multi-Agent Reinforcement Learning. Demonstrating that Situational Commitment is the only Evolutionarily Stable Strategy.

560+ Experiments
60 Figures
4 Environments
1,000 Max Agents
JAX MAPPO Multi-Agent RL Game Theory Sen's Meta-Ranking
Open science (Zenodo DOI)

WhyLab

Don't just predict the future. Cause it.

Causal Decision Intelligence Engine powered by Multi-Agent Debate. 22-cell pipeline bridging Science (causal inference) and Art (business decisions). Benchmark-validated on IHDP / ACIC / Jobs.

22 Pipeline Cells
3 AI Agents
58+ Tests
1.164 IHDP √PEHE
Python EconML DoWhy FastAPI Next.js LLM Debate
Active — v0.4

Neo Genesis Ontology

An operating graph for a one-person AI company

What happens when you treat a company as a queryable graph instead of scattered documents? 529 nodes (meta 310 + business 219) with 100% provenance — every node and edge carries its source. OAG governed execution, Foundry-style Object Sets, GraphRAG (HippoRAG + LightRAG). Provenance integrity gated by daily cron — Cyc taught us "completeness" is anti-goal; reliability is the real target.

529 Nodes
100% Provenance
20+12 Competency Q PASS
3 Store backends
PROV-O OntoClean DuckDB Neo4j AuraDB NetworkX HippoRAG LightRAG FastMCP