Over 30% Changed Their Stance on Moral Dilemmas After an AI Counterargument, Kobe Study Finds

A study led by Kouhei Masumoto, a professor of psychology at Kobe University, suggests that artificial intelligence can measurably shift people's moral judgements. When participants were shown AI-generated counterarguments to the choices they had made on ethical dilemmas, more than 30% changed their stance. The result points to how strongly automated arguments can influence views on questions that have no easy answers.

To test this, the researchers posed ethical dilemmas that are variations of the classic trolley problem, a thought experiment in which a person must choose between two harmful outcomes with no clean solution. Participants first made their choice. For each of the two options they could have taken, the team then presented an artificial-intelligence-generated counterargument, and measured whether opinions moved afterward.

The study also found that older adults showed a greater tendency to be swayed by the AI-generated arguments than younger participants did. The researchers frame that pattern as a potential danger of persuasive technologies in a rapidly ageing society such as Japan, where a growing share of the population is older and could be more exposed to such influence.

The finding adds to wider concern about the sway of artificial intelligence over contested questions, at a time marked by both misinformation and an overload of information. Rather than testing whether the AI's arguments were correct, the work measured only how readily they moved people's opinions on moral questions where reasonable people disagree.