Researchers in Cambridge, Massachusetts have uncovered a growing trend in academic writing that suggests an increase in the use of AI language models like ChatGPT. The study, led by Dmitry Kobak and a team of colleagues, analyzed over 14 million biomedical abstracts from 2010 to 2024, revealing a distinct change in vocabulary usage attributed to the influence of advanced language models.
One striking finding from the research is the noticeable presence of excessive and flowery vocabulary in academic texts that have been assisted by ChatGPT. The analysis identified 774 unique words that deviate from the norm, suggesting a shift towards more embellished and dramatic language choices associated with AI language models.
The surge in excessive vocabulary usage coincides with the rise of ChatGPT and similar AI tools, indicating a potential impact on the writing style of academic papers. While words related to major events like the SARS-CoV-2 pandemic and Ebola outbreak showed natural increases in frequency, the sudden spikes in flowery language suggest a direct correlation with AI assistance in writing.
Kobak and his team have made the analysis code available for those interested in exploring the phenomenon further. Additionally, the main author addressed concerns about the influence of ChatGPT on writing styles, leaving the question open as to whether individuals are more inclined to adopt AI-inspired vocabulary or consciously avoid mimicking AI language models in their writing.
This study sheds light on the evolving landscape of academic writing, where AI language models are playing an increasingly significant role in shaping language usage. As technology continues to advance, the implications of AI assistance in academic writing raise important questions about authenticity, creativity, and the future of human-machine collaboration in scholarly communication.