Case study · Harvard University (IQSS)
ReviewIQ: Fraud Screening for the Dataverse Repository
Results
97%
Agreement with expert review in pre-production testing, across 200 historical submissions
~10,000
Submissions processed without failure in scale testing
The engagement
Firebrand built ReviewIQ on FireFoundry for Harvard's Institute for Quantitative Social Science, replacing manual screening of fraudulent submissions to Dataverse — the world's largest general-purpose research data repository — with an explainable review workflow. Rather than a black-box spam filter, ReviewIQ evaluates each submission against Dataverse's curation standards and recommends pass, expert review, or block — each recommendation carrying a confidence score, plain-language reasoning, and a rubric assessment, inside the curators' existing workflow.
Key highlights
- Clear-cut cases resolved automatically; only ambiguous submissions routed to human experts
- A single curator can face ~100 submissions after a long weekend — most of a day's work ReviewIQ is designed to absorb
- Human oversight preserved on every decision — curators act without switching tools or surrendering control
- Scalable foundation for the broader Dataverse community