ChatGPT Standardizes Human Thought, According to a Study

The Impact of Chatbots on Cognitive Diversity

A recent study highlights the potentially homogenizing consequences of large language models (LLMs) like ChatGPT on our way of thinking. According to an opinion piece co-authored by scientists and psychologists, the use of these tools could diminish the uniqueness of human thoughts, leading individuals to adopt similar modes of communication and reasoning. Zhivar Sourati, a computer scientist at the University of Southern California and the lead author of the article, asserts that individual differences in writing, reasoning, and worldview are diminished through interaction with the same LLMs. This results in a homogenization of expressions and thoughts among users.

The article, published Wednesday in the journal Trends in Cognitive Sciences, explores the impact of the widespread use of a few chatbots by hundreds of millions of people around the world on our individuality. The authors emphasize that LLMs generate writings that vary less than those produced by individuals themselves.

A Massive Adoption of AI

Statistics reveal a growing adoption of ChatGPT. According to Pew Research, one-third of Americans used ChatGPT last year, double the number from 2023. Usage is particularly prevalent among teenagers, with two-thirds reporting that they use chatbots, and nearly one-third using them daily. This trend is also extending to businesses, with a Stanford study indicating that 78% of organizations reported using AI in 2024, up from 55% in 2023.

The Dangers of Homogenized Thinking

The authors of the article warn about the consequences of this cognitive homogenization. Part of the reason LLMs might promote homogenized thinking lies in the data used for their training. This data often over-represents dominant languages and ideologies, leading to outcomes that reflect a narrow and biased view of human experience. Sourati explains that this homogenization threatens pluralism, an essential value for sound judgment and collective intelligence. The authors emphasize that if this trend is not controlled, it could flatten the cognitive landscapes that fuel our adaptability.

The concern is not limited to how LLMs influence individuals' writing or speech. Sourati warns that these tools subtly redefine what is considered credible discourse, a correct perspective, or even good reasoning. This influence extends even to those who do not use chatbots. If many people around an individual think and express themselves in a certain way, that person may feel pressure to align with these modes of thought, as they appear more credible or socially acceptable.