AI summary by TwinMind
Alarming: Fraud spreads in Science -- and I fear it will become worse
Sabine Hossenfelder
7m 56s
Scientific Publishing Crisis Overview
- Science facing rapidly worsening problem past 2 years
- Fraudulent papers increasing due to AI advances
- Fraudsters becoming increasingly aggressive
Record Retractions in 2023
- Nature magazine data: 10,000+ retracted papers (all-time record)
- Most retractions not honest mistakes but fabricated content:
- Sham data
- AI-generated text
- Repurposed figures/images
- Retraction rate rising faster than total publications
- 2 in 1,000 scientific papers now retracted
- Fraud identification becoming more difficult as AI improves
Geographic Distribution of Retractions
- Saudi Arabia leads retractions
- Followed by: Pakistan, Russia, China, Egypt
- Predominantly Eastern problem
- Most retractions from publisher Hindavi
- Concentrated in special issues
Special Issues Problem
- Original concept: collections on specific topics/recent developments
- Editorial process outsourced to guest editors
- Guest editors invited friends, guaranteed publication
- Became "junkyards of scam papers"
- No accountability for guest editors
- Journals profit from subscriptions regardless of content quality
- Researchers spammed with contribution invitations
Hindavi/Wiley Response
- Hindavi subsidiary of Wiley publisher
- Wiley recognized special issue problem
- Announced major changes
- Stopping Hindavi brand usage
- Uncertain if this solves core problem
Paper Mills Network
- Semi-legal networks producing scam papers
- Guide papers to publication for money
- Offer authorship on specific topics
- Price depends on author list position
- Originated in China (academics paid for papers/promotions required)
- Spread to Russia, India, Iran, Eastern Europe
Public Health Targeting
- Many papers in drug development, psychology
- Drug development papers look similar (swap drug names)
- Dangerous: fake papers cited supporting useless drugs
- Example: controversial drug ivermectin
Tortured Phrases Phenomenon
- Automated systems rewriting technical terms
- Examples:
- "Magnetic resonance" → "attractive reverberations"
- "Fuzzy logic" → "fluffy rationale"
- "Breast cancer" → "bosom peril"
- "Renal failure" → "kidney disappointment"
- "Ant colony" → "subterranean insect province"
Editorial Bribery Trend
- Paper mills bribing journal editors
- Frederick Ying (Retraction Watch) reports alarming trend
- Peer-review manipulations increasing:
- AI-generated review text
- Pretending to review papers
- Affects major publishers:
- Elsevier: editors offered cash weekly
- Taylor & Francis: bribery attempts at journal editors
Systemic Problem
- Scientific system values quantity over quality
- Well-known issue, little action taken
- AI improvement will worsen fraud detection difficulty
- Fraudulent work spreading to more disciplines
