Chatbots are making our writing more similar but losing the identity and personality behind it
Large language models (LLMs) are making writing styles more similar without changing the overall meaning, according to an analysis of more than 880,000 texts across different writing types published in Nature Human Behaviour. The findings indicate that using LLMs could make it more difficult to identify important clues about aspects of a person's identity, personality and mental health from their…
Large language models (LLMs) are causing written text to become more uniform, according to a study published in Nature Human Behaviour. By examining over 880,000 texts from various genres, researchers found that using LLMs can make it harder to discern aspects of a person's identity, personality, and mental state from their writing.
This is because LLMs tend to favor common language patterns, reducing the diversity in writing styles by 21% to 50%. When LLMs rewrite human-written texts, they maintain the original meaning in 87% of cases, but this uniformity comes at the cost of accuracy in analyzing personal characteristics from the language. The LLMs weaken certain language patterns related to traits like extraversion, friend-related words, loyalty, age, and other markers of identity and personality.
Despite this, some associations remain, such as the connection between negative emotion words and neuroticism, religion-related words and purity, and social words and gender. The study highlights the potential impact of LLM-assisted writing on fields like psychology, mental health care, recruitment, and personalized services, calling for further research to understand why some language markers are preserved while others are weakened.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
