Understanding Risk Preference and Risk Perception When Adopting High-Risk and Low-Risk AI Technologies

Recent advances in AI have significantly changed people’s lives, yet sometimes their inherent risks deter adoption. Risk preference and perception in AI remain understudied. We surveyed 406 participants to explore how risk preferences, risk perceptions, and socioeconomic variables influence AI adoption in high-risk (autonomous vehicles) and low-risk (recommendation algorithms) contexts. Socioeconomic groups overall show different levels of risk aversion and seeking across scenarios. For high-risk autonomous driving, the risk aspects tend to be centralized. In contrast, the risk aspects of recommendation algorithms are more dispersed. These findings indicate a prevailing inclination among individuals toward caution regarding risks, highlighting the need for government policies that distinguish high- and low-risk AI. Regulations for autonomous vehicles should be strengthened to ensure safety and clarify liability, while those for recommendation algorithms should be expanded to enhance public risk awareness. This study aims to support policymakers toward more targeted AI risk management.