According to a new study published in the Journal of the Royal Society Interface, artificial intelligence can be programmed to feel guilt for committing "betrayals." Scientists simulated a game in which participants (AI or humans) had to make a moral choice between dishonest behavior for personal gain and honesty.

During the experiment, researchers identified two types of programmed guilt in AI. "Social" guilt arises when the AI observes its game partner also showing remorse; in this case, the AI changes its behavior and begins to cooperate. "Asocial" guilt manifests as an internal "compass," regardless of the partner's actions. In such a situation, the AI can become a victim of manipulation if the other participant remains cynical.

According to the scientists, programming AI for guilt is not always constructive. It raises an important philosophical question about whether honest robotic behavior should extend to immoral humans. Ultimately, such behavior could inadvertently help people commit immoral acts, which is potentially harmful to society.

ORIENT