In an era characterized by a vast volume of scientific publications, traditional author-generated abstracts and keywords can be susceptible to authorial bias and inherent subjectivity, potentially undermining research discoverability. This study explores the application of generative artificial intelligence by employing large language models—ChatGPT-4o and Claude 3.7 Sonnet—to generate clearer and more descriptive abstracts and keywords directly from full-text articles, thereby offering a potential solution to mitigate authorial subjectivity. A data set comprising seven articles from a special MDPI issue on journalism and media in the age of AI was used, with each article processed using a structured prompt that specified elements of background, methods, results, and conclusions. The generated outputs were then evaluated using the Gemini 2.0 Flash Thinking model on a five-point Likert scale, with the results indicating that both AI systems achieved higher scores than the traditional, human-generated versions. Furthermore, Claude 3.7 Sonnet showcased a minor advantage in clarity and precision. The findings suggest a potential for integrating LLMs into the scientific publication workflow to improve the quality of research summaries and mitigate authorial subjectivity. This integration could, in turn, potentially lead to more effective indexing and information retrieval in academic databases.
Kamińska et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: