Overview
- An opinion paper in Trends in Cognitive Sciences argues that large language models are homogenizing writing, reasoning and perspective, which could weaken creativity and group problem solving.
- The authors cite evidence that chatbot outputs are less varied than human writing and tend to mirror WEIRD language, values and reasoning styles.
- A preprint analysis of Reddit posts, news and preprints suggests stylistic diversity declined after ChatGPT’s launch, while another preprint identifies writer groups who retain distinctively human styles.
- Researchers note that common AI workflows favor linear chain-of-thought reasoning, potentially displacing intuitive or abstract approaches and shifting agency from users to model suggestions.
- The team urges developers to build in greater real-world linguistic, cultural and reasoning diversity, and a separate preprint reports frequent chatbot users are far better at spotting AI-generated text than infrequent users.