Artificial Intelligence and ChatGPT: perspectives and challenges for Bibliographic Classification

Abstract

It presents perspectives and challenges of applying Artificial Intelligence in the field of Library Science. To this end, it focuses on the practice of Bibliographic Classification and conducts a comparative study between the results obtained through classification performed by a human and by ChatGPT, the chatbot tool used in this analysis. The study is based on the Dewey Decimal Classification and the Universal Decimal Classification. The results revealed significant divergences between both methods, exposing mistakes and hallucinations of the generative AI model GPT-3.5, on which the current free version of ChatGPT is based. Challenges and limitations to the effective applicability of ChatGPT in the context of Bibliographic Classification are highlighted, as well as the relevance of the librarian in the practice of classification and thematic analysis, given the importance of mental exercise and critical analysis in interpreting the subject to be classified. It should be noted that the study does not conduct a comprehensive comparative analysis. The need for future, more comprehensive investigations with different perspectives from human experts and AI models is emphasized.