Research Standards & Integrity

Our research standards.

This page describes the standards visible in HSI’s current published work and the limits of the Institute’s present review system. HSI does not claim that all publications are externally peer reviewed.

On this page
  1. Authorship and responsibility
  2. Sources and evidence
  3. Review
  4. Uncertainty, disagreement, and the Verification Test
  5. Conflicts and funding
  6. Corrections, versions, data, and code
  7. Generative AI
  8. Questions and corrections

Authorship and responsibility

Substantive publications identify their human author or authors and, where relevant, the editor responsible for the series. A byline denotes responsibility for the argument, the treatment of evidence, and the final text. Contributors should be credited for material research or editorial work; institutional affiliation is not treated as authorship.

Sources and evidence

Current HSI publications link or cite the evidence on which material claims depend. Horizon Scans may include a bibliography and a source-cutoff statement; short-form Searchlight issues provide source links in the text. Primary regulatory, legal, scientific, or institutional sources are preferred where the question permits. Commentary and industry material may be used when clearly attributable and relevant.

HSI distinguishes a reported fact from its own interpretation. Quantitative claims should be traceable to a named source, and quotations should be attributed to the speaker or publication from which they were taken.

Review

Current practice includes author review of sources and editorial review before publication. Factual review should test names, dates, figures, links, quotations, and the description of cited institutions. Editorial review tests clarity, internal consistency, attribution, and whether the strength of a conclusion matches the evidence presented.

External review may be commissioned where specialist judgment would materially improve a publication. When external review is used, the publication should say so. HSI does not describe ordinary editorial review as peer review and does not imply that outside review has occurred when it has not.

Uncertainty, disagreement, and the Verification Test

Analysis should state what is observed, what is inferred, and what remains uncertain. Where a conclusion depends on a forward claim, HSI may use the Verification Test: a concise statement of the claim, evidence that would support it, and evidence that would weaken or falsify it. The test is one integrity mechanism, not a substitute for sourcing, review, or corrections.

Examples appear in Searchlight 021, Searchlight 022, and Searchlight 023.

Conflicts and funding

Authors, editors, directors, officers, and key employees are subject to HSI’s published Conflict of Interest Policy. Project or institutional funding that is relevant to a publication should be disclosed with the work. Funders and partners do not receive authority to direct, approve, delay, or suppress conclusions. See Funding & Transparency.

Corrections, versions, data, and code

Substantive corrections or revisions should be dated and described on the publication page; historical conclusions are not silently rewritten. Horizon Scan 001 provides the current model: a source-cutoff statement, edition history, and correction contact.

When a publication depends on a dataset, codebase, scoring rubric, or research instrument that can be shared lawfully and responsibly, HSI should link it or explain why it is unavailable. The Institute does not claim that every publication has associated data or code.

Generative AI

Generative AI is not credited as an author. Where it supports research or editorial work, the named human author and editor remain responsible for verifying sources, protecting confidential material, assessing originality, and approving the final published text. HSI’s current public source package does not document a more detailed organization-wide generative-AI procedure; adopting one is listed for governance review.

Questions and corrections

Evidence, methodological questions, and correction requests may be sent to corrections@horizonsearch.org. General research inquiries may be sent to research@horizonsearch.org.