What Type of Advice does Change Decisions on Moral Dilemma Effectively?

While the development of artificial intelligence is promising and may bring us prosperity, it is not without serious problems. For example, how should an autonomous vehicle system judge moral dilemmas? How can and should people be encouraged to accept the advice of AI in such situations? In order to implement acceptable and explainable AI, it is important to study how people accept the advice of others. In the present study, we focused on everyday moral dilemmas and investigated which advice is more likely to be accepted. The study was conducted in three steps. In Step 1, a list of moral dilemma advice was created, which was evaluated by another group of people in Step 2. In Step 3, the moral dilemmas were presented together with one of the created pieces of advice. There were four types of advice (emotional or rational
$\times$
altruistic or egocentric directions). Participants were instructed to read the advice and make their judgements about the dilemmas. The results showed that the advice that included guidance in an altruistic direction changed the judgments in an altruistic direction. However, the egocentric advice did not change the judgments to egocentric, suggesting that it is more difficult to encourage egocentric judgments than altruistic judgments. Future research should also focus on how AI should provide advice when it determines that all risks can be avoided by prioritizing the driver's own protection.