{
  "id": 17542,
  "title": "Scientists using LLMs will ‘do more, less well’, modelling study predicts",
  "url": "https://urgent.news/2026/07/31/scientists-using-llms-will-do-more-less-well-modelling-study-predicts",
  "topic": "ai",
  "section": "AI",
  "published": "2026-07-31T00:00:00.000Z",
  "source": {
    "name": "Nature",
    "slug": "nature",
    "url": "https://www.nature.com/articles/d41586-026-02397-5"
  },
  "original_language": "en",
  "account": "Researchers utilizing large language models (LLMs) to aid their work will likely spend less time refining their projects and more time starting new ones, according to a modelling study1. The authors contend that LLM adoption will lead to scientists \"doing more, less well — rather than the same amount, better\"1. However, they argue that these artificial intelligence tools are not the sole cause of this outcome. The results stem from a flawed incentive system in science that prioritizes quantity over quality, according to study co-author Carl Bergstrom, a biologist at the University of Washington in Seattle1. Bergstrom explains that scientists are incentivized to produce numerous papers, and LLMs assist them in meeting this ever-growing demand1. \"LLMs are rarely the problem themselves,\" states Bergstrom1. \"LLMs hold up a mirror to problems that we already have.\" The study, posted on the arXiv preprint repository on 19 July, has not yet undergone peer review1. To predict how LLMs will impact scientific productivity, the authors divided the research process into distinct stages1. The initial discovery phase involves brainstorming hypotheses and conducting preliminary experiments to assess the viability of a project1. Following this, there is a two-part development phase comprising required work (such as creating figures and drafting papers) and discretionary development (like conducting follow-up experiments and refining writing1). The researchers utilized optimal-foraging theory methods — which examines how an animal maximizes energy gain while conserving resources1 — to analyze how scientists will shift their efforts after incorporating LLMs into their work1. The model assumes LLMs are operating at their best — being inexpensive, swift, and accurate1. The study predicts that LLMs can expedite all phases of the scientific process1, but that this speed will not result in improved papers1. Accelerated discovery and required-work phases mean researchers can publish more papers rapidly than without LLM assistance1, and the pressure to publish provides little incentive to spend additional time polishing these papers1.",
  "summary": "Nature, Published online: 31 July 2026; doi:10.1038/d41586-026-02397-5 Research suggests that the combination of incentives to publish and the use of large language models will lead to more papers, but they will be less refined.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}