The study revealed that texts generated by AI can be just as persuasive as those written by humans. Interestingly, when respondents were informed that the author was AI, they were more likely to agree with the arguments presented, even if those contradicted their own views.
With the emergence of large language models, the question arises as to how AI will influence political communication. What will happen if such texts become the norm in the information space? Researchers suggest that machines can both amplify polarization and promote openness to diverse opinions.
The researchers set two main objectives: to determine how persuasive AI-generated messages are compared to human ones, and to assess whether knowledge that the text was written by AI affects the perception of these messages.
In the experiments, participants analyzed texts related to various political issues, such as smoking bans in public places, gun control, vaccination, and carbon tax. Some messages were written by humans, while others were generated by AI. Some participants were informed that the author was a human, while others were told it was AI. The researchers then examined how this influenced changes in the respondents' opinions.
The results showed that the persuasiveness of AI messages was, on average, comparable to that of human counterparts. Moreover, knowing that the author was AI made participants more receptive to arguments that diverged from their own beliefs. Texts from humans were perceived through the lens of personal experience, while AI-generated materials were evaluated in terms of logic and facts. Participants were also more likely to share their views and seek additional information if they believed the text was created by AI, and they exhibited less aggression towards opposing viewpoints.
On one hand, AI can become a powerful tool for communication in politics, helping to reduce polarization and increase openness to diverse opinions. On the other hand, the willingness to accept arguments coming from AI does not always depend on their validity, which poses a risk of spreading misinformation.
Despite the relatively modest effect of AI texts, their potential to reinforce existing beliefs and scale up can be utilized for both positive and negative political communication. Researchers emphasize that in the lead-up to elections, it is important to recognize the impact of AI on public opinion and to develop strategies for the responsible use of such technologies.
