How our research works
ARIAA publishes two kinds of research: methodology whitepapers (how we approach a decision class, what engines we compose, where the uncertainty lives) and calibration reports (Brier scores, coverage tests, and reliability diagrams against real historical outcomes). What we do not publish: source code, training corpora, domain-pack YAML, or solver internals. That is the moat.
Every paper is released under a short email gate — name, work email, institution, and two lines on the decision you're working on. We use the email to send the paper, to send future companion papers in the same series if you opt in, and for nothing else.
Available titles
Computational Feasibility as a Service
The framing paper: why feasibility under live constraints is the missing primitive, how ARIAA composes solvers to answer it, and where generic LLMs and GRC tools fall short.
Request →Electoral feasibility — 2022 calibration study
Brier scores, coverage tests, and reliability diagrams from ARIAA's reference-domain electoral forecasts. The 2022 Brazilian cycle is the anchor case — dense polling, hard public outcomes, six months of daily re-forecasts. Method is disclosed; the signal feed, weights, and corrections are not.
Request →Cross-domain contagion pathways
The framework behind ARIAA's cross-domain engine: pathway discovery, transfer-entropy estimation, and impulse-response simulation across signals that cross traditional silos.
Request →Calibration as a product discipline
How we operate Brier-score-driven calibration as a standing discipline — what we measure, how we decide when a forecast is decaying, and why we publish calibration reports at all.
Request →Request access
Fill the form below. We respond within one business day with a PDF link and, if the title is in active distribution, a one-page executive summary within the hour.