The manuscript claims
“Intervention X reduced outcome Y in population Z.” [12]
Making scientific judgment reusable.
OS-ERIN helps researchers move from a claim in a manuscript to the relevant portions of the sources it cites…without reconstructing the whole path from scratch every time.
Our goal is to build open infrastructure for inspectable claim–source assessment—supporting collective scientific intelligence and a shift from document-centred reporting toward argument-centred deliberation.
One practical problem
What did the cited source actually say?
During peer review, checking an important citation often means finding the source, getting access, searching through it, reconstructing which passages matter and interpreting the relationship. Most of that work then disappears into a review report or private notes.
OS-ERIN makes the cited work available and surfaces candidate passages worth inspecting.
After inspection
“Supports the direction of the claim, but only for population A and a 12-week follow-up.”
The software does not decide whether the source “proves” the claim. It reduces the search and reconstruction burden so a human reader can make a more inspectable judgment.
How it works
Machines help find. Humans judge.
The core interaction is deliberately narrow: one reviewer inspecting one occurrence of a claim in relation to one version of a cited source.
-
Step 01
Open the relation
Start from an important claim and the source the author cites to support it.
-
Step 02
Surface candidate passages
Once the cited source is available, OS-ERIN ranks passages that may be relevant.
-
Step 03
Inspect the source
The reviewer searches, reads the surrounding context, checks qualifications, refines the claim where necessary and can nominate material the ranking missed.
-
Step 04
Retain the reasoning
Relevant passages, the reviewer’s interpretation, scope, provenance and later revisions can remain inspectable.
Why this is different
Not another black-box citation score.
Many citation tools make the document, citation network or model classification the main object. OS-ERIN also makes a reviewer’s inspection of a particular claim–source relationship an inspectable object.
Many citation tools
- The system produces a classification or aggregate signal.
- The reviewer consumes the result.
- Human reasoning is usually outside the record.
- Useful inspection work is often reconstructed again later.
OS-ERIN
- The system nominates where the reviewer may want to look.
- The reviewer remains responsible for the interpretation.
- The relevant passages and reasoning can remain inspectable.
- Later readers may be able to build on earlier work where reuse is legitimate.
Why retention matters
Retained judgments form an inspectable scientific conversation
The immediate value is a better-supported assessment. Retaining these judgments creates a trace of how claims have been interpreted over time, helps direct attention to disagreement and outliers, and gives future readers an informed starting point for their own evaluation.
- Help one reviewer now
Make an important source check easier to complete and inspect during peer review.
- Preserve useful work
Keep the passages, qualifications and reasons that would normally disappear into private notes.
- Build shared scholarly memory
Where reuse is appropriate, later reviewers and researchers may start from an inspectable trace rather than from zero.
Current state
A research demonstrator
OS-ERIN is currently an invited research demonstrator built around the first working part of this vision: helping readers inspect citation-backed claims against their sources and retain their assessments. Diamond Open Access peer review is its first proving ground.
What exists now
A constrained reviewer workbench for moving from citation-backed claims to candidate passages in cited sources, supporting human inspection and retaining an attributable trace of the assessment.
- Testing value risk
- Does this save enough time and improve enough reasoning to become part of a real workflow?
- Testing usability risk
- Can assistance reduce effort without removing the reconstruction through which judgment develops?
- Testing feasibility risk
- Can relevant full text be accessed and candidate passages surfaced reliably enough across realistic literature?
Where we are starting
Start with peer review. Extend through research practice.
The first proving ground is Diamond Open Access peer review. Literature review is a closely adjacent use case because it involves many of the same tasks: identifying claims, tracing evidence, comparing sources and recording a reasoned assessment.
Diamond OA peer review
Help reviewers inspect citation-backed claims and preserve the evidence and reasoning behind their assessments.
Literature reviews
Help researchers trace claims across the literature, compare supporting and conflicting evidence, and retain the reasoning behind their conclusions.
Teaching and risk intelligence
Explore whether the same structured records can support evidence-evaluation learning and, later, responsible research and risk analysis.
Shared scholarly infrastructure
At scale, authorized assessments can form a reusable, inspectable layer of scientific deliberation that other people, tools and services can build on.
Research framing
The simple interaction raises much harder questions.
The product proposition begins with a practical peer-review task. The research lens asks what happens if relation-level inspection becomes recurrent, retained and shared: how prior judgment changes later judgment; how technical systems shape attention; when provenance becomes surveillance; and who controls a scholarly resource once institutions depend on it.
This deeper framing is not a promise about what the current demonstrator already achieves. It is the research programme required to understand what responsible scale would mean.
Scholarly memory
When does preserving judgment save reconstruction, and when does it anchor the next reader?
Technical mediation of attention
Can machines nominate where to look without quietly becoming the authority?
Correction and revision
How should concern, disagreement and changing source status propagate without becoming verdicts?
Identity, power and participation
Can attribution support accountability without making criticism unsafe or hardening competence into rank?
Stewardship and public interest
If accumulated trace becomes valuable enough to depend on, who controls it and who captures the value?