Comprehensive Scoping Review on Discrepancy in Accuracy of ChatGPT in Dental Health Practice

Abstract

ChatGPT (generative pretrained transformer) is a unique kind of AI model designed for conversational applications, thereby mimicking human conversation by recognizing human speech/text/language/intent and responding in a way imitating human behavior. However, the frail understanding of how ChatGPT is reshaping dental practice is questionable. The aim of this study is to map the existing literature regarding the discrepancies in the accuracy of ChatGPT in dental practice The objective of the study is to identify the knowledge gaps and key areas of findings on specific parameters that contributed to discrepancies and the potential risk of bias regarding ChatGPT application in dentistry. The review was conducted over a 12‐week time frame. The research question is “why is there discrepancy in accuracy of ChatGPT in dental practice?” We applied Arksey and O′Malley′s 2005 methodological framework. The search strategy was initiated using PRESS in databases such as PubMed/Medline, Embase, and Scopus and was conducted with language restriction and time restriction. Publications included in the review spanned original studies and review articles in the domain of dental practice, excluding studies on dental education, academics, and research areas. Data charting was done in two stages: study identifier stage and study characteristic stage involving multiple author pairs (reviewers and librarian). The data was finally mapped in the form of graphics involving tables and representative charts for better understanding of the coverage and synthesis of the topic. The review synthesized a total of 98 publications using search terms “Chat GPT AND Dentistry,” “Chat GPT AND Dental Practices,” “Chat GPT OR Dentistry,” “Chat GPT OR Dental Practices.” After removing duplicate papers, grey literatures, only abstracts, and articles in a language other than English, 56 papers were totally extracted. As per the inclusion criteria, a total of nine papers were synthesized. We applied a specific coding system for the included studies as SC/01–SC/09 and a response rating system to summarize and report the synthesized data. Collation of the included studies reported four studies with “positive response” and five studies with a “negative response.” All the studies, however, showed high concerns of data bias, data breaching, and lack of validity with potential risk of bias regarding the accuracy of ChatGPT in dental practice. This review reported diverse results on the accuracy of ChatGPT applications in dentistry. The current scoping review highlights the immediate need for considerations and implementation of ethical‐based guidelines/frameworks/legal regulations/licensing in the application of ChatGPT in dentistry, thereby laying a foundational platform for drafting specific guidelines/frameworks/legal regulations/licensing by policymakers and researchers for its future role.