AI could make people dull, one scientist fears. Here's why.

Artificial intelligence could one day boost human cognition, leading to significant advancements in science, technology and other fields. However, new academic research suggests AI could also make us dull. This is because large language models, the backbone of AI apps, often provide information that is predictable and normative for the population, reducing life's complexity to a bland, watered-down set of ideas.

Sandra Matz, a professor at Columbia Business School and author of the study, explained to CBS News that these models predict the most likely next word in a sentence or event in a sequence, which by definition, is average. Consequently, they homogenize decisions, and all users end up with the same output. To draw their conclusions, Matz and her co-authors analyzed over 110,000 real-world decisions made by 1,000 people and compared these to choices made by generic and personalized AI agents.

They also utilized data from Facebook's myPersonality project, which conducted personality tests on users who agreed to share their Facebook profiles for research purposes. According to Matz, a computational social scientist with a background in psychology and computer science, using AI to shape a decision, such as choosing a vacation destination or walking shoes, steers individuals towards the most common choices and away from more distinctive or quirky behaviors and preferences.

She added that LLMs play it safe within a user's preferences, nudging behavior towards more normative options and narrowing the range of what individuals explore. Matz emphasized that AI inherently avoids risk due to training methods, as it aims to keep users on the platform by showing them what they already like rather than introducing them to new or unconventional choices.

To prevent AI agents from diminishing the richness and diversity of human experience, Matz recommends incorporating an "exploration mode" option in AI apps so users can access more unexpected, less conventional recommendations. This would help ensure that individuals do not become boring and that culture does not collapse into a single set of preferences.