Agent-Based Systemic-Risk Dissertation
This private dissertation coordination project organizes a research program on agentic AI in financial markets. The central concern is how AI-mediated decision systems can create invalid evaluations, correlated behavior, liquidity fragility, and systemic risk.
Core thesis
As agentic AI systems become more common in financial decision-making, risk can emerge from two connected failure modes:
- Invalid historical evaluation: agents use information that was unavailable at the simulated decision time.
- Systemic interaction: many individually plausible agents produce correlated perception, correlated action, shared data-source dependence, shared prompts, and cascade-like stress.
Research structure
The dissertation plan is organized around three paper directions:
- Financial replay and temporal leakage
- Agentic herding and systemic-risk metrics
- Interventions, governance, and mitigation design
Why it matters
Agentic AI should be evaluated as part of a system, not only as an isolated model. This work connects analytics, agent-based modeling, market microstructure, financial risk, and AI governance into one research agenda.
Status
This is a private research-in-progress repository. The portfolio page describes the academic direction without exposing private notes, drafts, or implementation details.