Akademik araştırmada yenilik: insan ve yapay zeka perspektifleri

Özet (abstract)

Purpose Novelty is commonly regarded as one component of high-quality scholarship. Nevertheless, little is known about how novelty is claimed and what is considered novel across academia, including any field differences. In response, this article uses a sequential design to investigate novelty claims and field differences in academic research. Design/methodology/approach First, a simple classification scheme was created for theory, methods, data, integrative synthesis and contextual transfer novelties. Next, the Large Language Models (LLMs) Gemma 3 27B, GPT-OSS-120B and GPT-OSS-20B were used to identify original contributions in articles from the full texts of eight large journals representing contrasting general fields. Finally, a random sample of 50 articles per journal was classified by the authors with the same scheme and the results compared. Findings The three LLMs tested achieved moderate or substantial agreement with academic coders on the occurrence of data, theory and methods novelty within full text journal articles, but there was little agreement within and between humans and LLMs on integrative synthesis and contextual transfer novelties, suggesting that they may not be well defined concepts. There were also substantial field differences in the prevalence of data, theory and methods novelty. Practical implications Authors and reviewers should be careful to identify valid research contributions in interdisciplinary research or fields other than their home fields. Originality/value This is the first attempt to automatically classify academic research theory types.

Yazarların özeti; kaynağından alınmıştır. Journal of Documentation, 2026 · DOI ↗