A recent study suggests that engaging with an AI chatbot has the potential to sway voters’ decisions, impacting future elections. The research, involving 1,530 Canadians, revealed that the chatbots were more successful in influencing Canadian voters to change their preferences compared to American participants.
Gordon Pennycook, an associate professor at Cornell University and a Canadian, led the study to explore the persuasive power of generative AI in political contexts. According to Pennycook, the study indicated that AI chatbots were remarkably effective in convincing individuals to alter their voting choices, surpassing traditional methods like political advertisements.
Published in the journal Nature, the study disclosed that during the fall of 2024, one in 21 U.S. respondents switched their vote to Kamala Harris after interacting with an AI chatbot, while one in 35 shifted support to Donald Trump. In the Canadian segment conducted during the final week of the federal election in April, participants were surveyed on 17 policy issues influencing their voting decisions.
Results showed that the chatbot interactions prompted changes in voting intentions among participants. Notably, the study highlighted the adaptability of AI chatbots in tailoring arguments for each respondent, which significantly enhanced their persuasive impact, particularly when supported by factual information.
Pennycook emphasized that the prolonged engagement with AI chatbots, typically lasting six to eight minutes, proved more influential than brief ad exposures. The study also pointed out the difference in impact between the U.S. and Canadian contexts, attributing it to the saturation of election-related content in American media.
While acknowledging the potential efficacy of AI-based persuasion in future elections, the study cautioned about the need for regulation to address emerging threats posed by AI technologies in electoral processes. Chief Electoral Officer Stéphane Perrault recommended transparency measures for AI-generated electoral communications and chatbot functionalities to ensure the dissemination of accurate information.
Despite concerns about potential misuse of AI technologies by political parties, the study underscored the significant role AI could play in shaping voters’ perceptions and emphasized the importance of privacy safeguards and government oversight to mitigate adverse impacts of AI in political advertising.
Fenwick McKelvey, an expert in communication studies at Concordia University, praised the study for highlighting the persuasive influence of generative AI in electoral contexts, pointing out its growing adoption by political parties worldwide. McKelvey raised concerns about the ethical implications of combining AI chatbot technology with voter databases and urged stronger privacy regulations to safeguard against unauthorized data usage in political campaigns.