文献速递牙体牙髓病

ChatGPT作为牙髓局部麻醉科学信息可靠来源的综述

A Review of ChatGPT as a Reliable Source of Scientific Information Regarding Endodontic Local Anesthesia.
PubMed 文献 · J Endod · 2025-02-12

本研究评估ChatGPT对16个牙髓局部麻醉问题的回答是否可靠。与同行评审循证文献相比,ChatGPT仅50%可靠(32个问题中16个一致);51条参考文献中59%引用错误、12%无法检索、18%为虚构。作者认为人工智能需进一步训练才能被信任提供准确的牙髓麻醉信息。

要点

  • ChatGPT回答与循证文献一致率为50%(32题中16题一致)。
  • 51条参考文献中59%引用错误、12%无法检索、18%为虚构。
  • 作者认为人工智能需进一步训练才能被信任提供准确的牙髓麻醉信息。

核心结论

  • 人工智能需进一步训练才能被信任提供牙髓麻醉领域的准确信息。
  • ChatGPT应继续改进以为提供者和患者提供可靠信息。

研究设计与样本

  • 本研究为综述,选取16个牙髓局部麻醉代表性问题,要求ChatGPT回答并提供参考文献,由2名作者独立用5级Likert量表与同行评审循证文献比较。

主要发现

  • 与同行评审循证文献相比,ChatGPT可靠率为50%(32个问题中16个回答一致)。
  • ChatGPT提供的51条参考文献中,59%(30/51)引用错误、12%(6/51)无法检索、18%(9/51)为虚构。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

INTRODUCTION: ChatGPT is an artificial intelligence chatbot, developed by OpenAI, which uses Deep Learning technology for information processing. The chatbot uses natural language processing and machine learning algorithms to respond to users' questions. The purpose of this study was to review ChatGPT responses to determine if they were a reliable source of scientific information regarding local anesthesia for endodontics. METHODS: Sixteen representative questions pertaining to local anesthesia for endodontics were selected. ChatGPT was asked to answer the 16 questions and provide supporting references. Each provided ChatGPT reference was evaluated for accuracy using NLM NIH.GOV (PubMed), Google Scholar, journal citations, and author citations. Peer-reviewed, evidence-based literature citations related to the initial questions were collected by the authors. The 2 authors independently compared the answers of the ChatGPT to the peer-reviewed, evidence-based literature using a 5-answer Likert-type scale. RESULTS: ChatGPT was reliable 50% of the time when compared to the peer-reviewed, evidence-based literature. That is, ChatGPT had the same literature-based response as our peer-reviewed, evidence-based literature in 16 of the 32 questions. Of the 51 total references for Chatbot, 59% (30 of 51) had the wrong reference, 12% (6 of 51) of the references couldn't be retrieved and 18% (9 of 51) of the references were hallucinations (made up references). CONCLUSIONS: Artificial intelligence needs further training to be trusted for accurate information in the field of endodontic anesthesia. ChatGPT should continue to improve to provide reliable information for providers and patients alike.

来源:J Endod,查看原文。版权归原作者所有。

引用

  1. Reader A, Drum M. A Review of ChatGPT as a Reliable Source of Scientific Information Regarding Endodontic Local Anesthesia. J Endod. 2025. PMID: 39952316. doi:10.1016/j.joen.2025.02.002

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。