Does ChatGPT Increase Language Homogenization?
摘要
Since the rise of ChatGPT in late 2022, the use of certain words, such as delve, in published journal articles has increased significantly. ChatGPT uses this word more frequently than most Western English speakers. Research shows that the use of large language models in published articles grew rapidly in 2023, especially in fields like computer science and medicine. The increased reliance on ChatGPT by scholars and students raises questions about its impact on language acquisition and homogenization. Word-processing tools like Microsoft Word or Grammarly have influenced language standardization by changing certain grammatical constructs, but ChatGPT’s ability to generate text independently may have a stronger impact. This could lead to a decline in local dialects and unique linguistic features, and users might adopt the model’s biases, whether political, linguistic, or cultural. The extent of this impact depends on how much a person relies on ChatGPT and their ability to evaluate its output. Students might use ChatGPT for idea generation, literature research and writing, while scholars might use it for translation or proofreading. ChatGPT’s linguistic peculiarities are found more frequently in writing, both in student-written, as well as scholarly articles. As ChatGPT overuses certain words and sentence structures, which have become indicators of AI-written text. Future versions could increase linguistic homogeneity if trained on AI-generated content, but companies are trying to exclude such data. Unique models tailored to an author’s voice could address this issue, but students need to learn to write independently first, which becomes more difficult as ChatGPT (co-) authored content becomes more common in journalistic, as well as scholarly literature. Schools need to incentivize learning without these systems to ensure critical thinking skills and a unique writing voice can be developed before relying on ChatGPT.