Scholarlink

<p>Artificial intelligence (AI) is a plausible tool for enhancing and expediting multivariable, large-scale meta-analyses, yet little is known about its effectiveness at the inclusion/exclusion stage. We randomly assigned 1,200 papers from a three-variable mediation meta-analysis to be screened with or without the assistance of Elicit. Elicit reduced the time it took to make inclusion/exclusion decisions by 20.6%. The effect of Elicit condition on decision time remained significant after controlling for five variables that could explain variance in decision time. Coders could evaluate over 93% of low- and medium-inference criteria and over 82% of high-inference criteria based entirely on information provided by Elicit. Our findings suggest Elicit is a viable AI tool for keeping meta-analyses tractable to researchers.</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

Facilitating meta-analysis inclusion decisions with Elicit AI

  • 期刊论文

  • 发表日期：

    2025年7月25日

查看全文
  • Anne-Marie R. Iselin,
  • Jamie DeCoster,
  • Xiaoye Xu

摘要

Artificial intelligence (AI) is a plausible tool for enhancing and expediting multivariable, large-scale meta-analyses, yet little is known about its effectiveness at the inclusion/exclusion stage. We randomly assigned 1,200 papers from a three-variable mediation meta-analysis to be screened with or without the assistance of Elicit. Elicit reduced the time it took to make inclusion/exclusion decisions by 20.6%. The effect of Elicit condition on decision time remained significant after controlling for five variables that could explain variance in decision time. Coders could evaluate over 93% of low- and medium-inference criteria and over 82% of high-inference criteria based entirely on information provided by Elicit. Our findings suggest Elicit is a viable AI tool for keeping meta-analyses tractable to researchers.