LLM may be standardizing human expression – and subtly influencing how we think (dornsife.usc.edu) AI

A USC Dornsife study argues that widespread use of large language model chatbots could narrow human cognitive and linguistic diversity by standardizing how people write, reason, and form credible judgments. The authors say LLMs often mirror dominant cultural values in their training data and encourage more uniform, linear reasoning patterns, which can reduce individual agency and group creativity. They call on AI developers to deliberately build in real-world global diversity in training—so chatbots better support collective intelligence rather than homogenizing it.