Education Evidence Check
Goal
Improve paper quality by verifying that important claims are grounded in reliable, recent, and relevant evidence.
Inputs
- Claim, paragraph, draft section, or citation list
- Target standard: course paper, thesis, journal article, systematic review
- Field/context: education level, region, subject area
- Existing references if available
Workflow
- Extract atomic claims from the text.
- Classify each claim:
- background claim
- empirical finding
- causal claim
- theory claim
- method claim
- policy/practice implication
- For each high-risk claim, search supporting and contrasting evidence.
- Check citation quality:
- relevance
- recency
- empirical strength
- citation context
- whether later studies support or contradict it
- Rewrite claims with calibrated strength.
- Produce a citation-gap list.
Tool Calls
Scite
Use for citation context and support/contrast checks.
https://scite.ai/
Recommended checks:
- Is the cited paper supported or contradicted?
- Is the citation used for the same claim?
- Are there later reviews or meta-analyses?
Elicit
Use for evidence reports and structured extraction.
https://elicit.com/
Query pattern:
What evidence supports or challenges the claim that [claim] in [education context]?
Semantic Scholar Citation Lookup
Find paper details:
curl "https://api.semanticscholar.org/graph/v1/paper/DOI:10.xxxx/yyyy?fields=title,year,abstract,citationCount,influentialCitationCount,references,citations"
Find papers on a claim:
curl "https://api.semanticscholar.org/graph/v1/paper/search?query=teacher%20feedback%20student%20writing%20achievement%20meta-analysis&limit=10&fields=title,year,abstract,citationCount,authors,venue"
OpenAlex
curl "https://api.openalex.org/works?search=teacher%20feedback%20student%20writing%20achievement%20meta-analysis&sort=cited_by_count:desc&per-page=10&mailto=YOUR_EMAIL"
Output Format
Return:
| Claim |
Current Citation |
Evidence Status |
Better Evidence Needed |
Suggested Rewrite |
Risk |
Evidence status:
- Strongly supported
- Moderately supported
- Mixed evidence
- Weak evidence
- Contradicted
- Needs citation
- Overclaimed
Quality Rules
- Causal language requires causal evidence.
- Prefer systematic reviews, meta-analyses, and recent empirical studies for broad claims.
- Do not cite a paper merely because it is related; verify the specific claim.
- Mark unsupported claims instead of smoothing them over.
- For AI-assisted writing, preserve a clear audit trail of searched sources and citation decisions.
1---2name: education-evidence-check3description: Use when checking whether claims, theory choices, research questions, variable relationships, or literature review statements in education papers are supported, contradicted, weakly supported, outdated, or missing citations. Uses Scite-style citation context, Elicit evidence reports, Semantic Scholar citation relationships, and manual source verification.4---56# Education Evidence Check78## Goal910Improve paper quality by verifying that important claims are grounded in reliable, recent, and relevant evidence.1112## Inputs1314- Claim, paragraph, draft section, or citation list15- Target standard: course paper, thesis, journal article, systematic review16- Field/context: education level, region, subject area17- Existing references if available1819## Workflow20211. Extract atomic claims from the text.222. Classify each claim:23 - background claim24 - empirical finding25 - causal claim26 - theory claim27 - method claim28 - policy/practice implication293. For each high-risk claim, search supporting and contrasting evidence.304. Check citation quality:31 - relevance32 - recency33 - empirical strength34 - citation context35 - whether later studies support or contradict it365. Rewrite claims with calibrated strength.376. Produce a citation-gap list.3839## Tool Calls4041### Scite4243Use for citation context and support/contrast checks.4445```text46https://scite.ai/47```4849Recommended checks:5051- Is the cited paper supported or contradicted?52- Is the citation used for the same claim?53- Are there later reviews or meta-analyses?5455### Elicit5657Use for evidence reports and structured extraction.5859```text60https://elicit.com/61```6263Query pattern:6465```text66What evidence supports or challenges the claim that [claim] in [education context]?67```6869### Semantic Scholar Citation Lookup7071Find paper details:7273```bash74curl "https://api.semanticscholar.org/graph/v1/paper/DOI:10.xxxx/yyyy?fields=title,year,abstract,citationCount,influentialCitationCount,references,citations"75```7677Find papers on a claim:7879```bash80curl "https://api.semanticscholar.org/graph/v1/paper/search?query=teacher%20feedback%20student%20writing%20achievement%20meta-analysis&limit=10&fields=title,year,abstract,citationCount,authors,venue"81```8283### OpenAlex8485```bash86curl "https://api.openalex.org/works?search=teacher%20feedback%20student%20writing%20achievement%20meta-analysis&sort=cited_by_count:desc&per-page=10&mailto=YOUR_EMAIL"87```8889## Output Format9091Return:9293| Claim | Current Citation | Evidence Status | Better Evidence Needed | Suggested Rewrite | Risk |94|---|---|---|---|---|---|9596Evidence status:9798- Strongly supported99- Moderately supported100- Mixed evidence101- Weak evidence102- Contradicted103- Needs citation104- Overclaimed105106## Quality Rules107108- Causal language requires causal evidence.109- Prefer systematic reviews, meta-analyses, and recent empirical studies for broad claims.110- Do not cite a paper merely because it is related; verify the specific claim.111- Mark unsupported claims instead of smoothing them over.112- For AI-assisted writing, preserve a clear audit trail of searched sources and citation decisions.
ilog3