Education

AI outperforms humans in personalised debates, raising election integrity concerns

Tuesday, 20 May 2025 12:31AM UTC

New research reveals that AI, particularly ChatGPT-4, can persuade debate participants more effectively than humans when using personal information, spotlighting risks of political manipulation and the urgent need for regulatory frameworks in electoral processes.

Recent research has illuminated the remarkable capabilities of artificial intelligence (AI) in the realm of persuasion, particularly within the context of debates. A study published in Nature Human Behaviour indicates that AI can match or even outperform human participants when it comes to swaying opinions during discussions, raising serious implications for election integrity and the broader landscape of public discourse.

This study, led by Francesco Salvi from the Swiss Federal Institute of Technology in Lausanne, involved online experiments with 600 participants who were either matched against human opponents or ChatGPT-4, a sophisticated large language model (LLM). The debates tackled various propositions ranging from the mundane—such as dress codes in schools—to heated topics like abortion rights. While AI was found to be equally persuasive as human opponents in general scenarios, its effectiveness notably increased when it had access to personal information about its debate partner, such as demographics or political affiliation. In fact, the data revealed that AI changed participants' views more effectively than human defendants 64% of the time when such personal data was used.

Salvi expressed concern over the potential for persuasive AI to manipulate undecided voters on a large scale, creating "armies of bots microtargeting" with nuanced narratives that could elude regulation and accountability. He warned that it would not be surprising if malevolent entities had already started exploiting these capabilities to disseminate misinformation and propaganda. However, he also pointed out that these technologies could be harnessed for positive outcomes, including the reduction of conspiracy theories and political polarisation.

Prof Sander van der Linden, a social psychologist at the University of Cambridge who commented on the study, echoed the alarm bells, stating that this research revives pertinent discussions around the mass manipulation of public opinion through personalised LLM interactions. He noted that while the use of analytical reasoning by AI enhances its persuasive power, other studies have contested the influence of personal information on artificial entities.

The potential misuse of AI technologies extends beyond mere persuasion; it raises ethical questions about the information landscape in electoral contexts. Recent legislative efforts in the United States, such as the Candidate Voice Fraud Prohibition Act and the Securing Elections From AI Deception Act, are taking shape to address these issues. These initiatives aim to prevent the use of AI-generated content that could mislead voters, demonstrating the urgency of governing AI in electoral processes.

Moreover, research published in the Proceedings of the National Academy of Sciences highlights that personalised political advertising, tailored to resonate with individuals based on their personality traits, is far more effective than generic campaign messages. This further complicates the ethical landscape, as the ability to craft targeted political messages poses risks that call for vigilant scrutiny and policy solutions to safeguard electoral integrity.

As generative AI technology advances, its capacity to facilitate disinformation campaigns is set to become a pressing concern. Experts warn that individuals or groups could leverage these tools for large-scale misinformation efforts with limited resources, effectively democratizing information warfare. Notably, while there are potential benefits—like improved election administration through better translation tools for diverse communities—the risks demand proactive measures to protect the integrity of electoral processes.

As we look toward upcoming elections, the interplay between AI capabilities and public sentiment presents a battleground where ethical considerations and regulatory frameworks must evolve to keep pace with technological advancements. The discourse around the persuasive power of AI is set to shape not only the political landscape but also the very fabric of democratic society.

Reference Map

Core focus on AI persuasion and debate from the lead article.
Summary of the study results and implications for election integrity.
Legislative efforts regarding AI in elections.
Study on personalized political ads and ethical considerations.
Discussion on the potential for disinformation campaigns using generative AI.

Source: Noah Wire Services

More on this

https://www.theguardian.com/technology/2025/may/19/ai-can-be-more-persuasive-than-humans-in-debates-scientists-find-implications-for-elections - Please view link - unable to able to access data
https://www.theguardian.com/technology/2025/may/19/ai-can-be-more-persuasive-than-humans-in-debates-scientists-find-implications-for-elections - A study published in Nature Human Behaviour reveals that AI can be as persuasive as humans in debates, raising concerns about election integrity. Researchers conducted online experiments matching participants with human opponents and ChatGPT-4, finding that AI performed similarly to humans in persuading others, especially when provided with personal information. The study highlights potential risks of AI in spreading misinformation and influencing elections, while also acknowledging possible benefits in reducing conspiracy beliefs and political polarization.
https://www.ncsl.org/elections-and-campaigns/ai-in-elections-a-look-at-the-federal-and-state-legislative-landscape - This article from the National Conference of State Legislatures discusses recent legislative efforts to address the use of AI in elections. It highlights several bills, including HR 4611, the Candidate Voice Fraud Prohibition Act, which prohibits the distribution of paid political communications containing materially deceptive AI-generated audio. The article also covers HR 6936, the Federal Artificial Intelligence Risk Management Act of 2024, focusing on managing AI risks in electoral processes, and HR 8858, the Securing Elections From AI Deception Act, which prohibits the use of AI to deceive voters.
https://www.pnas.org/doi/full/10.1073/pnas.2403116121 - A study published in the Proceedings of the National Academy of Sciences examines the persuasive influence of large language models (LLMs) in political microtargeting. The research demonstrates that personalized political ads tailored to individuals' personalities are more effective than nonpersonalized ads. The study highlights the potential risks of utilizing AI and microtargeting to craft political messages that resonate with individuals based on their personality traits, emphasizing the need for ethical scrutiny and policy-oriented solutions to safeguard electoral integrity.
https://www.secureelectionsproject.org/ai-election-2024/ - This article from the Secure Elections Project explores the implications of AI in the 2024 U.S. elections. It discusses how generative AI could enable large-scale disinformation campaigns with fewer resources, potentially democratizing information warfare. The article also highlights the potential benefits of AI in election administration, such as enhancing public trust through improved translation tools for diverse communities. It emphasizes the need for vigilance and proactive measures to address the challenges posed by AI in the electoral process.
https://www.ncsl.org/elections-and-campaigns/ai-in-elections-a-look-at-the-federal-and-state-legislative-landscape - This article from the National Conference of State Legislatures discusses recent legislative efforts to address the use of AI in elections. It highlights several bills, including HR 4611, the Candidate Voice Fraud Prohibition Act, which prohibits the distribution of paid political communications containing materially deceptive AI-generated audio. The article also covers HR 6936, the Federal Artificial Intelligence Risk Management Act of 2024, focusing on managing AI risks in electoral processes, and HR 8858, the Securing Elections From AI Deception Act, which prohibits the use of AI to deceive voters.
https://www.pnas.org/doi/full/10.1073/pnas.2403116121 - A study published in the Proceedings of the National Academy of Sciences examines the persuasive influence of large language models (LLMs) in political microtargeting. The research demonstrates that personalized political ads tailored to individuals' personalities are more effective than nonpersonalized ads. The study highlights the potential risks of utilizing AI and microtargeting to craft political messages that resonate with individuals based on their personality traits, emphasizing the need for ethical scrutiny and policy-oriented solutions to safeguard electoral integrity.

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 9

Notes: The narrative references recent research and current legislative efforts, indicating a contemporary context. However, without specific publication dates for referenced studies, there is a slight uncertainty about the exact novelty of the information.

Quotes check

Score: 8

Notes: The narrative includes quotes from Francesco Salvi and Prof Sander van der Linden. While these quotes are likely recent, the exact source or original publication date for these statements is not provided.

Source reliability

Score: 10

Notes: The narrative originates from The Guardian, a well-known and reputable publication.

Plausability check

Score: 9

Notes: The claims about AI's persuasive capabilities align with ongoing advancements in AI technology. The narrative also references relevant legislative efforts, which supports its plausibility.

Overall assessment

Verdict (FAIL, OPEN, PASS): PASS

Confidence (LOW, MEDIUM, HIGH): HIGH

Summary: The narrative is well-supported by reputable sources and recent research, indicating a high level of freshness and plausibility. The quotes appear to be original, enhancing the credibility of the narrative.

Artificial intelligence
Elections
Misinformation
ChatGPT
Political campaigns