Sull’affidabilità delle bibliografie generate dai chatbot. Alcune considerazioni

Abstract

The aim of this paper is to examine the use of chatbots in academic settings, focusing on certain aspects of the reliability of bibliographies generated by artificial intelligence (AI). This study uses data obtained from querying ChatGPT 3.5 and Bard to produce a bibliography on a specific topic (Open Access Books). The analysis of the query results, along with comparisons to similar studies, has highlighted the low level of accuracy in the bibliographic citations generated by chatbots and the prevalence of phenomena such as hallucinations and confabulations in AI, which can affect the accuracy of the information. The chatbots use real author and editor names and existing titles that are consistent with the prompt terms, but they frequently use them incorrectly.