Researchers had an AI argue with people about a moral dilemma. Over 30% just changed their mind.
The Japan Times reported on a Kobe University study led by psychology professor Kouhei Masumoto, testing how AI-generated counterarguments affect moral judgment.
The Setup
56 people ages 18-30 and 74 people over 65 were given classic trolley-problem dilemmas (the switch version and the footbridge version), then shown ChatGPT-crafted counterarguments against whichever choice they made.
The Results
32.3% of younger participants reversed their decision on the switch dilemma, and 36.9% did so on the footbridge dilemma, from a single AI counterargument.
The Age Gap
Older adults showed a greater tendency to be swayed than younger adults, raising concerns about persuasive AI in a rapidly aging society like Japan's.
The Researcher's Own Reaction
"Existing research suggests that people adopt only about 20% to 30% of advice they receive from others," Masumoto said. "I was struck by the fact that more than 30% of participants completely reversed their judgments based on the AI's opinion."
The Takeaway
Masumoto notes this isn't necessarily bad -- counterarguments can help people consider new perspectives. But the study is a clear signal: AI's ability to shift moral judgment, particularly for older adults, is a real and measurable effect, not a hypothetical one.