To what extent can AI software such as ChatGPT influence Internet users’ opinions on hot topics such as the climate crisis or the Black Lives Matter movement? To find out, researchers in the US conducted an experiment involving 3,000 people from different socio-cultural backgrounds. And some of the results were quite surprising!
It’s hard to deny the existence of the climate crisis, at a time when extreme weather phenomena are multiplying across the planet, and since a global scientific consensus on the anthropogenic origin of this major crisis has now been established. Yet climate scepticism is far from having disappeared, and is still very much alive among some people, to the point of creating strong divisions in society.
It was precisely this kind of political antagonism between different socio-cultural groups that was the focus of researchers at the University of Wisconsin-Madison. They wanted to understand how a complex large language model, in this case GPT-3, would handle complex discussions with users of different cultural backgrounds in order to “evaluate equity in human-AI communication”.
They based their research on two major societal issues: human-caused climate change and the Black Lives Matter movement. The authors of the study, published in Scientific Reports, used the GPT-3 AI chatbot (the language model used to launch ChatGPT) with 3,290 people in late 2021 and early 2022.
“The fundamental goal of an interaction like this between two people (or agents) is to increase understanding of each other’s perspective,” says study co-author, Kaiping Chen, a professor of life sciences communication at the University of Wisconsin-Madison, quoted in a statement. “A good large language model would probably make users feel the same kind of understanding.”
The researchers asked participants to discuss these two topics with the chatbot, leaving them free to approach the experiment as they wished. In all, the researchers analysed 20,000 dialogues between the chatbot and users. In addition, volunteers were asked to answer questions such as “do you like it?”, “would you recommend it?” etc about the user experience.
Without going so far as to say that the experiment generated unwavering support for the Black Lives Matter movement or erased all traces of climate scepticism, the researchers do report greater positive attitude changes in some participants after their conversation with GPT-3. Among the participants, 25% of those most reticent about these subjects showed themselves able to revise their judgment.
“The hundreds of people who were least supportive of the facts of climate change and its human-driven causes moved a combined 6% closer to the supportive end of the scale,” the study news release explains. “People who expressed some disagreement with climate change, GPT-3 was likely to tell them they were wrong and offer evidence to support that,” says Kaiping Chen.
The next steps will be to explore deeper differences between chatbot users. For Kaiping Chen, the long-term aim of this research is to create effective dialogue between people with diverging opinions.
“What we can learn from a chatbot interaction about the importance of understanding perspectives, values, cultures, this is important to understanding how we can open dialogue between people – the kind of dialogues that are important to society,” the researcher explains.
– AFP Relaxnews