It turns out that people who frequently use AI tend to have 'dark personality traits' such as narcissism and psychopathy

The use of AI in education, work, and entertainment has been gaining attention in recent years. A study analyzing data from over 14 million website visits found that frequent users of AI are more likely to have the '

https://www.liebertpub.com/doi/10.1177/21522715251379987

Most people rarely use AI, and dark personality traits predict who uses it more
https://www.psypost.org/most-people-rarely-use-ai-and-dark-personality-traits-predict-who-uses-it-more/

While many studies have asked participants about their AI usage and views, self-reported usage is often unreliable, especially when it comes to new technologies, and misjudgments and memory errors are common. Therefore, Emily McKinley , a doctoral student at the University of California, Davis, and her team conducted a study to objectively measure AI usage using people's web browsing history.

'Despite widespread concerns and excitement about tools like ChatGPT, there was little fundamental understanding of actual usage patterns,' said McKinley. 'We wanted to measure what's actually happening, not just by looking at frequency of AI use, but also the psychological profile of people who have embraced AI and how it's integrated into their broader digital behavior.'

The study used data collected from 499 undergraduate students at two universities and 455 members of the general public. In both studies, participants shared up to 90 days of web browsing history, but because exporting the necessary data required Google Chrome, participants were limited to Chrome users. Participants also completed a questionnaire measuring personality traits, attitudes toward AI, and demographic characteristics.

The researchers used lists of well-known AI websites, such as ChatGPT and Microsoft Copilot, to identify which browser visits were AI-related. They also used a content classification system using large-scale language models to identify other websites and analyzed the percentage of visits to AI-related websites compared to the total number of visits, the types of websites visited before and after using AI, and psychological attributions associated with AI use.

The analysis revealed that AI-related websites accounted for only 1% of the total website visits among student subjects. Furthermore, most subjects did not visit any AI-related websites, with a small number of subjects accounting for the majority of traffic. The most frequently visited AI-related website was ChatGPT, accounting for 85% of AI-related visits. 'We were surprised at how infrequently AI was used, even by students, who are often early adopters of new technologies,' said McKinley.

Students who used AI more frequently also had slightly higher scores on personality traits related to narcissism and psychopathy, and tended to have more positive attitudes toward AI in general. Demographic characteristics such as income and gender were weakly associated with AI use, but age and ethnicity were not significantly associated.

The general public sample showed even lower rates of AI use, with AI-related websites making up just 0.44% of total website visits. While no significant correlation was found between personality traits and AI use in this group, there was a moderate relationship between Machiavellianism and AI browsing.

The research team also focused their analysis on users whose website browsing involved AI at least 4% of their total time. Among student participants, these users scored significantly higher than other students on measures of Machiavellianism, narcissism, and psychopathy. These patterns were less clear among the general public, likely due to lower overall AI use, reducing the power to detect significant differences.

'Interestingly, people who use AI more tend to score higher on undesirable personality traits, particularly Machiavellianism, narcissism, and psychopathy, and these patterns are even stronger among students,' McKinley said.

In addition, the researchers looked at the types of websites participants visited before and after using AI. They found that before using AI, participants visited search engines and login pages, but after using AI, they visited more websites related to education, computers, and professional work. This suggests that AI tools are being used more frequently as part of their workflow, especially in academic or work-related contexts.

Another important finding was the discrepancy between self-reported and actual usage of AI. The general public were asked not only about their actual website browsing history, but also about how often they thought they used AI. Analysis of this result and actual usage data showed only a moderate correlation between the two, suggesting that self-reporting alone does not accurately capture AI usage.

The current study is limited to web-based behavior and does not include AI use through mobile apps. It also focuses only on personality traits and basic demographic characteristics, without examining factors such as emotional well-being, motivation, or social environment.

'This study is one of the first to objectively quantify AI use in natural environments using months of real-world browsing data,' McKinley said. 'While AI now dominates much of our cultural conversation, the data suggests that most people still use it sparingly.'