Where machine learning meets financial markets.

Formalizes access distortion: a failure mode where a world model's aggregate accuracy looks acceptable while decision quality is systematically worse for agents with weaker information access. Two access tiers can be indistinguishable on accuracy while differing roughly twofold in realized decision utility.

Calibration-aware ensemble forecasting framework for macroeconomic indicators with explicit uncertainty quantification and regime-conditional allocation outputs.

Introduces Participation Debt: the gap between formally granted opportunities and opportunities a person can realistically reach. Shows standard fairness metrics can be fully satisfied while Participation Debt grows, particularly in healthcare and public policy contexts.

Empirical analysis of whether recreational cannabis legalization reduces racial disparities in marijuana arrest rates using staggered difference-in-differences estimation across states with varying legalization timelines.

Proves that evaluation frameworks are bounded by construction: entire classes of failure are invisible by design when they fall outside an evaluation space. The formal basis for Inferify's regime validity signal.

Formalizes the Justification Deficit: the measurable gap between predictive success and epistemic warrant. A model can be accurate for the wrong reasons; standard metrics cannot detect this gap.

ETR achieves sensitivity 0.81 for root-cause step identification in multi-step LLM agent trajectories vs. 0.23 for final-outcome evaluation.

Controlled evaluation framework grounded in k-space physics showing that corruption structure determines the qualitative character of failure independently of severity. Structured line dropout produces catastrophic collapse (normalized SSIM drop 0.877) with high cross-subject variance, a failure pattern aggregate PSNR evaluation cannot distinguish.

Discovering candidate structural-functional dissociation patterns in Alzheimer's MRI using multimodal unsupervised methods. Presented Denver, June 2026.

Computational identification of candidate resilience phenotypes under AβO-induced toxicity in MC65 cells using multimodal biological imaging data.