Research
Agenda.

Our methodology bridges the gap between quantitative finance and sociological inquiry.

01 — The Thesis

Price is a construct.
Value is a graph.

Traditional hedonic regression fails in illiquid, prestige-driven markets. We operationalize the Noll Framework to decode the sociological and institutional drivers of asset appreciation.

Active Price Management

Quantifying the "marketing premium" extracted by auction houses through catalog placement and sentiment analysis.

Institutional Gatekeeping

Tracking "Canonization" via museum exhibitions and critical reviews using NLP and causal discovery.

The Winner's Curse

Modeling bidder behavior and liquidity dry-ups to signal "Do Not Bid" when market dynamics overheat.

Provenance Networks

Detecting wash trading and "prestige transfer" through Heterogeneous Graph Neural Networks.

02 — The Engine

Deep Tech for Alternative Assets

We don't just predict what happens; we understand why. By moving from correlation to causality, we answer the counterfactuals that define alpha.

Causal Inference
Moving up Pearl's Ladder

Our Structural Causal Models (SCM) and Augmented Time Series architectures allow us to answer counterfactuals: "What would the price be if this artist hadn't had a MoMA retrospective?" This moves us from correlation (Seeing) to intervention (Doing).

MATMCD
LLM-Augmented Discovery

Our Multi-Agent Tool-augmented Multi-Modal Causal Discovery system deploys LLM agents to scrape unstructured text (catalogs, archives) and refine our causal graphs, creating a proprietary "Living SCM" of the art market.

Simulation
Agent-Based Modeling

We simulate auction rooms with heterogeneous agents (Collectors vs. Speculators) using Reinforcement Learning to identify liquidity risks and optimal bidding strategies in low-data environments.

Network Topology
Heterogeneous GNNs

Value is path-dependent. We map the "invisible" connections of provenance using Graph Neural Networks to detect anomalies, wash trading, and prestige transfer across generations.