Where would an independent evaluation help?
The Unjournal commissions and publishes evaluations of research relevant to global welfare. This project helps find candidates, connect them to important questions, and gather evidence for deciding what to evaluate.
This tool is in early-stage use for research discovery and prioritization. A high score suggests a paper may be worth evaluating; it does not certify the paper’s findings or mean an evaluation has been commissioned.
Choose what you want to do
Find research
Search by title, author, or topic. Open a result for its source text, possible decision relevance, scoring explanation, and feedback. Use “More filters” for a particular discovery source or focused search.
Give an independent rating
Read the paper or abstract before judging whether an evaluation would help. This view hides prior scores and analyses until you rate each paper in your browser.
Examine a candidate in depth
A small set of detailed pages gathers the case for evaluation, public attention and use, open questions, and evidence behind the scores. These pages are selected automatically from papers with human feedback.
Find a question that matters
A crux is an uncertainty that could change someone’s view or action. Explore public forum posts, comments, and The Unjournal’s pivotal questions. Check the original source and whether the question is precise enough for research to answer.
Suggest research
Point to a paper and explain which decision an evaluation might inform. Authors can use the author submission form. Suggestions require team review; submission does not guarantee an evaluation.
Read the scores with context
The dashboard’s default 0–100 score combines an AI estimate with human ratings where available. Where there are no human ratings, it remains an AI estimate. You can switch to human-only or AI-only scores and inspect the weights. Small differences should not decide between papers.
The AI shortlist and AI watchlist labels preserve the original screening recommendation. They can differ from the displayed score because that score uses your selected weighting and rating source. A request for further team assessment is a separate workflow status.
What evidence is most useful to add?
Name a decision and who faces it; explain which finding matters; link an existing critique, replication, or documented use; or identify a check that could change the conclusion. Distinguish a possible audience from evidence that an organization actually uses the research. A specific source or correction is often more useful than another unsupported score.
How reliable are the AI judgments?
The prompt is informed by past team judgments, but accuracy on new candidates remains uncertain. The 24-paper, 192-rating pilot found ordinary rerun variation and substantial differences from historical human ratings. Reproducibility and agreement with people are distinct questions; neither establishes the value of an evaluation by itself.
What happens to my feedback?
Ratings contribute to public aggregates under the form’s disclosure. You choose whether discussion is public or restricted to the team. Team members review evidence before making commissioning decisions. See the rating and privacy guide for revisions, attribution, and data handling.
For team members
Use the existing Coda Research Prioritization Hub for team assessments and assignments. “My ratings” on this public site reflects records saved in your current browser; it is not a cross-device account or a complete record of assigned work.
Specialist views and methods
- Topic-focused search catalog: see what was searched and why certain topics are overrepresented.
- Conflict research: assess data availability, feasibility, and useful robustness checks for a replication workshop.
- Legal research: explore candidates for the currently paused legal-scholarship initiative.
- Human-rating statistics: inspect the available feedback and its limited sample sizes.
- Score-stability report: examine the pilot, findings, and limitations.