Lattuce tutorial · Notes & Library
Library screening (dual-review)
Screen research references with dual-review views, inclusion decisions, and project-ready evidence in Library.
Advanced households can run PRISMA-style dual review on a frozen candidate set under Library → Screening (/library?kind=screening).
This is not live screening of an entire publisher index. You freeze Discovery hits, library sources, mission candidates, or a bibliography export into screening_records, then vote independently.
Use screening when the inclusion decision itself must be traceable. It is an advanced review queue, not a faster catalog search: prepare a clean candidate set, define the decision rule, and make sure each reviewer understands the same scope before voting.
Views
| View | Purpose |
|---|---|
| Queue | Title/abstract records awaiting your sealed vote |
| Conflicts | Disagreements for arbiter resolution |
Hub section navigation owns Queue | Conflicts — there is no second in-page tab row.
Review modes
| Mode | Who votes |
|---|---|
| Human + human | Two distinct household reviewers |
| Human + AI | You plus an AI agent (inclusionist-v1 / similar) |
| AI + AI | Two distinct AI personas (no shared output) |
Solo Advanced users cannot create human+human (needs a peer). Use human+AI or AI+AI instead.
Peer votes stay sealed until both slots submit. Conflicts open an arbiter resolution — not a third blind vote. Write the inclusion rule before opening Queue, vote from the available evidence, and avoid discussing a sealed vote with the other reviewer until the conflict stage.
Seed a project
From the Screening hub Seed panel:
- Explore — paste a literature query; Writer freezes OpenAlex / arXiv / Semantic Scholar hits server-side
- Mission — enter a completed
literature_discoverymission ID - Export — upload RIS, CSV, or BibTeX (5 MB max)
- Library sources / collections — attach owned items already in your catalog
From Explore → Literature search, use Add to screening on the current result page (Advanced). Free plans see an upgrade prompt instead of a silent missing control.
Entitlements
library_screening— Advanced (team) plan- AI votes also need
ai_missions
Screening checklist
- Freeze the candidate set from Explore, a mission, an export, or owned Library sources.
- Confirm the review mode and reviewer slots.
- Review Queue and leave disagreements sealed until both votes are complete.
- Resolve Conflicts with the arbiter and record the final decision.
- Return to Library or Writer with the accepted evidence.
If a control is missing, verify the plan entitlement and that the source set is frozen. Do not re-seed the project just because the Queue is empty; an empty Queue can mean every record is resolved.
Related
- Catalog and graph: Library catalog
- Manual literature search: Literature review with real references
- Research handoff: The Writer paper lifecycle
Browse more in Notes & Library or return to all tutorials.