# Responsible Use Policy ## 1. Purpose and Scope This Responsible Use Policy governs how the data, records, exports, generated pages, and related materials published by Campus Evidence Lab ("the project," "we," "us") may be used. Campus Evidence Lab is a static-first public evidence archive of United States campus-related civil rights records, drawn from public sources and organized into event records, school records, source records, review logs, correction logs, release notes, snapshot manifests, and CSV/JSON exports. This policy applies to anyone who accesses, downloads, redistributes, cites, or builds on this material, including researchers, journalists, students, advocates, institutional staff, and members of the public. This policy is not legal advice and does not create a legal relationship between any user and the project. It describes the norms, expectations, and limits that govern appropriate use of the dataset and site. ## 2. Appropriate Uses The dataset is intended to support: - **Academic and independent research** into patterns, gaps, and documentation practices around campus civil rights records. - **Journalism**, including investigative reporting that cites, verifies, and builds on public-source records contained in the dataset. - **Institutional accountability work**, including efforts by students, faculty, watchdog groups, or the public to understand how a school or institution has responded to documented incidents. - **Source review and verification**, including checking, corroborating, or challenging the public sources underlying a given record. - **Public-interest analysis** that examines documentation trends, reporting practices, or systemic issues without producing rankings or comparative scores. - **Methodology critique**, including scrutiny of how the project collects, tiers, labels, or presents records. - **Correction work**, including submitting corrections, duplicate reports, or source updates through the project's designated channels. Use consistent with these purposes should still respect the limitations and interpretive rules set out below. ## 3. Inappropriate Uses The dataset must not be used for: - **Harassment** of any individual, including students, staff, complainants, respondents, witnesses, or institutional employees. - **Doxxing** or any attempt to identify, locate, or expose private information about individuals named or implicated in records. - **School or institution rankings** of any kind, including "best," "worst," "safest," or "most dangerous" school lists derived from dataset content. - **Safety rankings or scores** purporting to measure how safe or unsafe a campus or institution is. - **Prevalence claims**, meaning any statement asserting or implying the actual rate, frequency, or incidence of civil rights issues at a school or in a population, since the dataset is not a census and does not support incidence estimates. - **Severity scoring**, meaning any numeric or categorical rating of how serious an incident or pattern of incidents is. - **Legal-liability claims**, meaning any assertion that a school, institution, or individual is legally liable, at fault, or in violation of law based on dataset content. - **Employment, admissions, or disciplinary decisions**, meaning using dataset records as a basis, in whole or in part, to hire, decline to hire, admit, decline to admit, discipline, or otherwise take adverse action against any individual or institution. - **Absence-implies-absence claims**, meaning any assertion that a school's absence from the dataset, or a low number of associated records, means that no incidents occurred there. Absence in this dataset reflects gaps in public-source documentation, not the absence of underlying events. Any use falling into these categories is outside the scope of what this dataset is designed or validated to support, regardless of the user's intent. ## 4. Source Review Expectations Every record in the dataset is expected to trace to a public source. Users engaging in source review, republication, or citation should: - Identify the specific public source or sources underlying a record before relying on it. - Treat a source's own limitations, corrections, or retractions as applying to any dataset record built on it. - Distinguish between a source reporting an allegation and a source reporting an adjudicated or confirmed outcome, and preserve that distinction in any downstream use. - Refrain from supplementing dataset records with private testimony, private screenshots, direct messages, or other non-public material; the dataset is public-source-only, and users should not attempt to fill gaps with unsupported or non-public information. ## 5. Limitations of Public-Source Datasets Users should understand the structural limitations inherent in a public-source archive of this kind: - The dataset reflects what has been reported and made public, not the full universe of incidents that may have occurred. - Reporting rates vary by school, region, time period, and subject matter, meaning the dataset's coverage is uneven and should not be read as proportional to actual incidence. - Public sources vary in reliability, specificity, and completeness, and the dataset inherits these variations. - The dataset does not constitute a legal finding, a ranking, a school safety score, a prevalence estimate, a severity score, an endorsement, or a complete census of campus civil-rights incidents. - Because records may be added, corrected, or reclassified over time, any given snapshot represents the state of the dataset at a specific point in time, not a final or authoritative account. ## 6. Review-Tier Interpretation Records are published under one of the following review tiers, and users must interpret and represent each tier accurately: - **Imported public source**: a record has been added from a public source but has not yet undergone further verification. These records carry the least certainty and should be described as unverified imports, not as confirmed facts. - **Source-family checked**: a record has been cross-referenced against related sources reporting on the same underlying event, providing modest additional confidence. - **Internally certified**: a record has undergone internal project review against the applicable evidence rules and methodology. - **Externally reviewed**: a record has undergone review by a party outside the project, providing the highest level of confidence available in this dataset. Lower-tier records may be published, but only with clear indication of their tier and corresponding limits. Any use of a lower-tier record should carry the same qualification the project applies to it; users must not strip or upgrade a record's stated tier when citing or republishing it. ## 7. Confidence Label Interpretation Where the project assigns a confidence label to a record, that label reflects the project's assessment of how well the available public sources support the record's factual claims. A confidence label is not a measure of the severity, importance, or legal significance of the underlying event, and it is not a prediction of how an adjudicative or investigative body would resolve the matter. Users should carry confidence labels forward whenever they cite or summarize a record, rather than presenting the record's content as though it were uniformly certain. ## 8. Affected-Community Label Interpretation Where the project labels a record with an affected community or population, that label describes who the public source material identifies as involved or impacted. It is not a statistical claim about which communities are more or less affected overall, and it must not be aggregated across records to produce prevalence, comparative, or risk claims about any community. Affected-community labels exist to preserve accurate context for individual records, not to support population-level conclusions. ## 9. Institution Response-Depth Interpretation Where the project documents an institution's response to a record, the depth and content of that documentation reflects only what is publicly available about the institution's response, not a comprehensive account of everything the institution did or did not do. A limited or absent response entry means limited or absent public documentation of a response, not confirmation that the institution failed to respond. Users must not characterize response-depth entries as a complete institutional record. ## 10. Citation and Snapshot Expectations Anyone citing, quoting, or republishing dataset content should: - Cite the specific record, its identifier, and its review tier. - Reference the dataset snapshot or release version used, since records may be corrected, updated, or reclassified in later releases. - Avoid presenting a single snapshot as a live or continuously current account; snapshots are point-in-time representations. - Preserve any limitation language, tier labels, confidence labels, or affected-community labels attached to a record when reproducing its substance elsewhere. - Link back to the source record or export where practical, so readers can verify context and any subsequent corrections. ## 11. Correction and Right-of-Reply Expectations Schools and other named institutions have a correction and right-of-reply path. Corrections, duplicate reports, source submissions, and school metadata corrections may be submitted through the project's GitHub issue templates or the site's structured submission page. Users who identify an error, omission, outdated status, or unfair characterization in a record are expected to raise it through these channels rather than through informal or public pressure directed at the project or at named individuals. Institutions or individuals named in a record may submit a reply, correction request, or additional context through the same channels. The project reviews such submissions under its review model and updates review tiers, labels, or record content as warranted. Submitting a correction request does not guarantee removal of a record, since the dataset's purpose is to document public-source reporting, not to adjudicate underlying disputes. ## 12. AI and Automated Analysis Cautions AI tools may assist with extraction, summarization, duplicate detection, and drafting in support of the project's own workflow. AI does not independently publish records and does not substitute for human review at higher review tiers. Users who apply their own AI or automated tools to the dataset should: - Verify AI-generated summaries or extractions against the underlying record and its cited public sources before relying on or republishing them. - Avoid using AI to generate rankings, severity scores, prevalence estimates, or liability conclusions from dataset content, since these outputs fall within the inappropriate uses described in Section 3 regardless of whether a human or an AI system produced them. - Disclose when AI-assisted analysis has been applied to dataset content in any publication or redistribution, so readers can assess the analysis accordingly. ## 13. Examples of Acceptable and Unacceptable Claim Phrasing **Acceptable phrasing:** - "According to a source-family-checked record in Campus Evidence Lab, a public report describes an incident at [School] in [year]; the underlying source is [source]." - "Campus Evidence Lab currently documents an unverified, imported-public-source record concerning [School]; no further verification has occurred." - "Public records reviewed by Campus Evidence Lab show a documented institutional response to [record], based on the sources cited; no further public documentation of the response was found." - "This analysis discusses documentation and reporting patterns across the sources included in Campus Evidence Lab as of [snapshot/release]; it does not estimate how common such incidents are overall." **Unacceptable phrasing:** - "[School] is the most dangerous campus in the dataset." - "The data shows [School] has a higher rate of civil rights incidents than other schools." - "[School] is legally liable for the incidents described in this dataset." - "Because [School] has no records in Campus Evidence Lab, no incidents have occurred there." - "This individual's record in Campus Evidence Lab confirms they engaged in misconduct," where the underlying record reflects an unverified or lower-tier public-source report rather than a confirmed finding. - Any claim, ranking, or score built by aggregating records across schools, communities, or time periods to suggest a comparative or population-level conclusion the dataset was not designed to support. ## 14. Enforcement and Reservation of Rights The project may decline to assist, may note publicly that a use is inconsistent with this policy, or may take other reasonable steps available to it, including corrections to its own published materials, in response to uses that violate this policy. This policy does not limit the rights any person or institution may separately have under the applicable open-source software license or the Creative Commons Attribution 4.0 International license governing the dataset, except that redistribution or use inconsistent with the terms above is discouraged and may be publicly identified as inconsistent with the project's intended use.