Better decisionsunderuncertainty.

Cumulant Research

Research model

“Cumulant combines human judgment, AI-assisted investigation, and deterministic analysis to study difficult questions in finance, allocation, and research reliability. The aim is not automated judgment. It is greater investigative reach without losing traceability or accountability.

Evidence fieldMultiple analytical paths remain connected to the question that produced them.

Research that leaves a record.

Cumulant studies financial inference, capital allocation, humanitarian finance, replication, and the reliability of AI-assisted research.

The question determines the method. Literature, data, code, criticism, and revision remain part of one inspectable record.

AI extends the investigation. Frozen artifacts, reproducibility checks, and human review keep responsibility with the researcher.

A painted violet boundary holds competing marks inside a deliberate field. — Boundary settingQuestions, constraints, and release rules are declared before results are interpreted.

Research methodA question moves through design, investigation, review, and preservation.

From question to inspectable result.

Cumulant does not force questions through one fixed pipeline. The question determines the method.

Across projects, a small set of disciplines recurs: trace the evidence, test alternatives, reproduce the result, and require human review before release.

Ask

01Find the puzzle

Identify the tension, anomaly, or decision that merits investigation.

02Study the field

Review prior work, competing explanations, and unresolved gaps.

Design

03State the question

Define a specific question and the boundary of the claim.

04Choose the method

Select a design that fits the question instead of forcing a standard recipe.

05Build the evidence

Acquire, validate, and preserve data and sources with provenance.

Investigate

06Analyze

Run deterministic code and save machine-readable outputs.

07Challenge

Use critical review to search for leakage, weak identification, alternative explanations, and unsupported claims.

08Stress-test

Test alternate samples, variables, models, windows, costs, assumptions, and out-of-sample behavior.

Preserve

09Verify

Trace numerical claims and citations to saved evidence.

10Reproduce

Rebuild results from preserved inputs, code, seeds, and configurations.

11Human review

Require human evaluation of the design, interpretation, limitations, and release decision.

12Release

Publish a versioned output with methods, code, data status, limitations, and correction history.

Different strengths. One accountable process.

AI extends the investigation. It does not inherit responsibility.

Human judgmentSets the question, boundaries, interpretations, and release decision.

AI-assisted investigationSearches, compares, drafts, and surfaces alternatives for review.

Deterministic recordComputes, tests, reproduces, and preserves the inspectable artifact.

Current research.

Cumulant works across five connected programs concerned with uncertainty, allocation, inference, and research reliability.

Financial inference & model risk

Statistical inference and model reliability in financial data, where limited decision counts and flexible analysis can make findings fragile.

FOCUS

Backtest inference

Decision timing

Researcher degrees of freedom

Model selection

False discovery

Robustness

Event studies

Statistical fragility

Many painted model paths diverge from one observed trace, with fragile paths fading at the edges. — Financial inference & model riskConceptual visual for Backtest inference, Decision timing, Researcher degrees of freedom. It is not an empirical result.

Selected work.

Current and completed investigations, presented with their actual status, limitations, and evidence record.

Program

Status