Over 30% shift ethics stance after AI counterargument, Japan study finds

The study by Kouhei Masumoto's team at Kobe University addresses a timely question: how susceptible are humans to AI persuasion on moral questions? By using variations of the trolley problem—a foundational thought experiment in ethics—the researchers created a controlled setting where participants' baseline moral reasoning could be tested against AI-generated opposing arguments. The finding that over 30% shifted their stance is significant because it demonstrates AI's reach beyond factual disputes into the domain of ethical judgment, where personal values and reasoning typically play a larger role. The observation that older adults were more swayed adds a critical layer: as populations age, especially in Japan where demographic decline is acute, the concentration of persuasive AI systems may create asymmetric vulnerability. The study does not report absolute participant numbers or the distribution of shifts across age bands, limiting the precision of the risk estimate, but the directional evidence points to a gap between AI's persuasive capability and current safeguards around its deployment in ethically sensitive contexts.