Research Integrity Screening

What is Research Integrity Screening?

An MCP server that connects AI agents (Claude, Cursor) to automated academic fraud detection across seven live data sources. It screens researchers, detects paper mill output, flags citation manipulation via Benford’s law analysis, assesses journal quality, and audits NIH grant–publication linkages, returning a composite Integrity Score (0–100) with a clear verdict.

How to use Research Integrity Screening?

Add the server’s URL to your MCP client’s configuration (Claude Desktop, Cursor, Windsurf) and include your Apify token in an Authorization header. Invoke any of six tools—such as screen_researcher_integrity or generate_integrity_report—via standard MCP protocol to screen a researcher or compare institutions.

Key features of Research Integrity Screening

• Benford’s law citation analysis for manipulation detection
• Paper mill template detection from repeated title prefixes
• ORCID verification with identity-unverified risk signals
• Publication velocity monitoring and year‑over‑year spike detection
• Four independent scoring models with a weighted composite score
• Hard override logic that forces HIGH_RISK on critical findings

Use cases of Research Integrity Screening

• Pre‑award grant screening of principal investigators
• Journal submission integrity review for paper mill detection
• Faculty hiring due diligence with full composite reports
• Research institution partnership assessment via side‑by‑side comparison
• Funding portfolio audit to surface high‑risk grants
• Citation manipulation investigation using digit‑by‑digit Benford’s law

FAQ from Research Integrity Screening

What data sources does the server query?

It queries OpenAlex, ORCID, PubMed, Semantic Scholar, Crossref, CORE, and NIH Grants in parallel.

How is the integrity score calculated?

The composite score combines four models (Researcher Integrity 30%, Paper Mill 25%, Journal Quality inverted 25%, Funding Risk 20%) with hard override logic for critical verdicts.

How long does a screening take?

A single tool call completes in under 2 minutes, thanks to parallel data fetching from all seven sources.

Do I need an API token to use it?

Yes. The server requires a Bearer token set via the Authorization header, obtained from your Apify account.

What verdicts can the server return?

The composite score maps to four verdicts: CLEAR, MINOR_CONCERNS, INVESTIGATION_NEEDED, or HIGH_RISK. Each sub‑model also provides a five‑tier label (e.g., CLEAN through CRITICAL for integrity).