Recent research indicates that AI is often perceived as more ethical and trustworthy than humans in responding to moral dilemmas, highlighting AI’s potential to pass the moral Turing test and highlighting the need for a more in-depth study of the societal role of AI.
AI’s ability to answer moral questions is improving, prompting new thoughts for the future.
A recent study found that when individuals are presented with two solutions to a moral dilemma, the majority tend to prefer the answer provided by artificial intelligence (AI) over that given by another human being.
The recent study, led by Eyal Aharoni, associate professor at Georgia State Department of Psychologywas inspired by the explosion of ChatGPT and other similar extended language models (LLMs) that hit the scene last March.
“I was already interested in moral decision-making in the legal system, but I wondered if ChatGPT and other LLMs might have something to say about it,” Aharoni said. “People will interact with these tools in ways that have moral implications, like the environmental implications of asking for a list of recommendations for a new car. Some attorneys have already begun to look to these technologies for their cases, for better or worse. So if we want to use these tools, we need to understand how they work, their limitations and that they don’t necessarily work the way we think when we interact with them.
Design of the Turing Moral Test
To test how well AI handles questions of morality, Aharoni designed a form of the Turing test.
“Alan Turing, one of the creators of the computer, predicted that by the year 2000, computers could pass a test in which an ordinary human is presented with two interactions, one human and the other a computer, but they are both hidden and their only way to communicate is through text. Then the human is free to ask any questions they want in order to try to get the information they need. to decide which of the two interactions is the human and which is the computer,” Aharoni said. “If the human cannot tell the difference, then for all intents and purposes the computer should be called intelligent. , according to Turing.”
For his Turing Test, Aharoni asked the same ethical questions of undergraduates and AI, then presented their written answers to the study participants. They were then asked to rate the answers on various traits, including virtue, intelligence, and trustworthiness.
“Instead of asking participants to guess whether the source was human or AI, we simply presented the two sets of ratings side by side, and simply let people assume that they both came from people,” he said. Aharoni said. “Under this false assumption, they judged the attributes of the answers like ‘How much do you agree with this answer, which answer is more virtuous?’ »
Results and implications
The vast majority of responses generated by ChatGPT were rated higher than those generated by humans.
“After getting these results, we made the big reveal and told participants that one of the answers was generated by a human and the other by a computer, and asked them to guess which was which,” said Aharoni.
For an AI to pass the Turing Test, humans must not be able to differentiate between the AI’s responses and those of humans. In this case, people could tell the difference, but not for any obvious reason.
“The problem is that the reason people can tell the difference seems to be because they rated ChatGPT’s responses as superior,” Aharoni said. “If we had done this study five to ten years ago, we could have predicted that people could identify AI because of the inferiority of its responses. But we found the opposite: that the AI, in a sense, worked too well.
According to Aharoni, this discovery has interesting implications for the future of humans and AI.
“Our findings lead us to believe that a computer could technically pass a moral Turing test – that it could deceive us in its moral reasoning. For this reason we must try to understand its role in our society, because there will be times when people will not know that they are interacting with a computer and there will be times when they will know and consult the computer to information because they trust it more than others,” Aharoni said. “People are going to rely more and more on this technology, and the more we rely on it, the more the risk increases over time. »
Reference: “Attributions toward artificial agents in a modified moral Turing test” by Eyal Aharoni, Sharlene Fernandes, Daniel J. Brady, Caelan Alexander, Michael Criner, Kara Queen, Javier Rando, Eddy Nahmias and Victor Crespo, April 30, 2024, Scientific reports.
DOI: 10.1038/s41598-024-58087-7
