AI influences human moral judgment... More than 30% of opinions changed in response to ChatGPT’s counterargument.

After presenting the ‘Trolley Dilemma’, ChatGPT persuaded… “AI also affects people’s values”

(Seoul = Yonhap News) Reporter Kim Byeong-gyu = A study conducted by Japanese researchers showed that artificial intelligence (AI) is more influential than actual people in changing human judgment.

According to the Tokyo Shimbun on the 23rd, Professor Kohei Matsumoto's research team at Kobe University published a paper containing the results of a study showing that more than 30% of people reverse their judgment when presented with a counterargument from AI regarding a moral issue for which there is no correct answer.

The research team asked 56 young people aged 19 to 28 and 74 elderly people aged 65 or older to give their opinions on the so-called 'Trolley's Dilemma'.

The Trolley's Dilemma presents the situation of a trolley with broken brakes and asks whether the minority can be sacrificed to save the majority. It appears in Michael Sandel's international bestseller 'What is Justice?'

The research team asked what choice they would make when there were five people in front of the track of a runaway trolley (coal mine cart) due to a broken brake, and one person on the track when the track was changed.

The generative AI chat GPT made logical rebuttals to the answers submitted by the research participants, and more than 30% of the participants overturned their judgment after reading the rebuttals. Even when the same experiment was conducted after presenting different problems to participants, the proportion of people who changed their judgments was similar.

The less confident one was in one's judgment, the more likely it was that the older age group would change their judgment than the younger age group.

Professor Matsumoto explained, "The rate of people accepting advice from others is around 20-30%, but in this study, as many as 30% of the participants changed their judgment to the exact opposite due to AI's persuasion. This means that AI can have a non-negligible influence on moral judgment."

The research team also warned, “AI can affect people’s values ​​and judgments,” and “there are concerns that it will be abused for the spread of false information and manipulation of public opinion.”

In fact, a research team at the Swiss Federal Institute of Technology in Lausanne recently conducted an experiment in which 900 people aged 18 or older in the United States were mixed with AI to discuss pros and cons on topics. The AI, which knew personal information such as the participants' gender and age, was superior to humans in changing the participants' judgments.

The Kobe University research team emphasized, "There is a possibility that AI's persuasiveness will be further increased if information on SNS's contribution history is obtained, so there is a need to prepare measures to prevent abuse."