For Researchers
Behavioral trace data with social-science rigor.
Gorebet is built on the theoretical foundations of collective efficacy, routine activity, and environmental criminology. We offer researchers request-based data access, open methodology, a reproducibility commitment, and honest disclosure of what our data cannot show.
Theoretical grounding
Gorebet's design is explicitly grounded in the criminological and sociological literature. Our collective-efficacy proxy is a behavioral trace derived from Sampson, Raudenbush & Earls (1997) — not a validated psychometric instrument. Our safety scores use Bayesian shrinkage to handle low-count woredas, following hierarchical modeling practice in the small-area estimation literature.
Our approach page documents eight theoretical pillars with primary citations, capability status labels (Operational / Roadmap / Aspirational), and explicit statements of what we have and have not validated.
Read our theoretical frameworkReproducibility commitment
All data shared with researchers includes: the exact aggregation parameters used, the suppression thresholds applied, the temporal window, and a copy of the methodology documentation current at the time of data delivery. We request that publications using Gorebet data cite the methodology documentation version.
We have not run a causal study of Gorebet's impact on crime rates or community outcomes. That requires partner-agency ground-truth data and a properly identified study design (pre-registered, with a control group). Researchers interested in impact evaluation are invited to contact us about a formal research partnership.
Available data products
All data products are subject to minimum-N suppression and coverage-indicator disclosure. Sensitive categories (gender-based violence, missing persons, child-involved incidents) are excluded from all request-basis products and require a governance agreement with IRB or equivalent documentation.
| Data product | Format | Access | Notes |
|---|---|---|---|
| Woreda-level incident aggregates | CSV / JSON | Request | Trailing 90-day windows. Minimum-N suppressed. Coverage indicators included. |
| Collective-efficacy proxy time series | CSV | Request | Reaction rate and co-membership count per neighborhood per month. Not a validated psychometric instrument. |
| Safety score time series | CSV | Request | Bayesian-adjusted, temporally decayed. Full model specification in methodology documentation. |
| Report-desert indicators | CSV / GeoJSON | Request | Population-weighted expected vs. observed reporting rates. Two-SD threshold for flagging. |
| Anonymized incident stream | JSON | Governance agreement | Coarsened to neighborhood level. Excludes sensitive categories. Requires IRB protocol or equivalent. |
What we are looking for in research partners
We prioritize research collaborations that are oriented toward community benefit, that commit to returning findings to the communities studied, and that include explicit bias and fairness analysis in their study design.
We are particularly interested in research on: platform effects on community trust, disparate reporting patterns across demographic groups, validation of behavioral proxies against survey instruments, and quasi-experimental designs that can isolate causal effects of neighborhood information systems.