Scholarlink

<p>The relationship between interdisciplinarity and scientific disruption remains controversial, and the potential of artificial intelligence for science (AI4S) to advance science has yet to be effectively demonstrated. In this paper, we propose an LLM-assisted approach to identify interdisciplinary research (IDR) and AI use papers, and build a new dataset to study how IDR, interdisciplinary collaboration, and AI4S relate to disruptive science. Using Scopus records from 2014 to 2025, we analyze 1,154,30 papers published in six generalist journals with strong interdisciplinary coverage. We estimate the associations using fixed-effects models and quantile regressions. The validation results show that the LLM-based approach reliably identifies IDR and AI use papers. Empirically, IDR is positively associated with scientific disruption, whereas interdisciplinary collaboration is negatively associated on average. However, this relationship differs across the disruption distribution, with interdisciplinary collaboration negatively associated with disruption for low-disruption papers and positively associated for highly disruptive papers. AI4S shows a stronger positive association with disruption than IDR, but this association weakens at the upper tail of disruption.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

On the disruptiveness of interdisciplinary research and artificial intelligence for science

  • 期刊论文

  • 发表日期：

    2026年5月1日

查看全文
  • Kun Tang,
  • Xinyuan Zhang,
  • Baiyang Li

摘要

The relationship between interdisciplinarity and scientific disruption remains controversial, and the potential of artificial intelligence for science (AI4S) to advance science has yet to be effectively demonstrated. In this paper, we propose an LLM-assisted approach to identify interdisciplinary research (IDR) and AI use papers, and build a new dataset to study how IDR, interdisciplinary collaboration, and AI4S relate to disruptive science. Using Scopus records from 2014 to 2025, we analyze 1,154,30 papers published in six generalist journals with strong interdisciplinary coverage. We estimate the associations using fixed-effects models and quantile regressions. The validation results show that the LLM-based approach reliably identifies IDR and AI use papers. Empirically, IDR is positively associated with scientific disruption, whereas interdisciplinary collaboration is negatively associated on average. However, this relationship differs across the disruption distribution, with interdisciplinary collaboration negatively associated with disruption for low-disruption papers and positively associated for highly disruptive papers. AI4S shows a stronger positive association with disruption than IDR, but this association weakens at the upper tail of disruption.