Artificial intelligence (AI) could significantly enhance human cognition, potentially driving advancements in various fields. However, new research suggests that AI might also dull our cognitive abilities.
The large language models (LLMs) powering AI applications often provide information that is average and normative for the population as a whole, simplifying complex ideas into bland, generalized concepts.
Columbia Business School professor Sandra Matz led a study analyzing over 110,000 real-world decisions made by 1,000 individuals. The researchers compared these to choices made by both generic and personalized AI agents, using data from Facebook’s myPersonality project, which conducted personality tests on users who shared their profiles for research.
According to Matz, computational social scientist with a background in psychology and computer science, relying on AI for decisions such as vacation destinations or shoe purchases leads individuals toward common choices rather than distinctive ones. This results in narrowed exploration across topics and psychological affinities.
AI systems are designed to minimize risk to keep users engaged, showing them content they already like rather than exploring the outskirts of their interests. Matz suggests tech developers incorporate an “exploration mode” for users who desire more unexpected recommendations, preventing individuals from becoming boring and ensuring cultural diversity is maintained.