Have you ever imagined debating with an AI? To debate AI is to effectively persuade through arguments based on personal information, making it a powerful tool for shifting opinions and influencing perspectives.
A study conducted by scientists from the Swiss Federal Institute of Technology (ETH Zurich) in Lausanne, Princeton University, and Fondazione Bruno Kessler in Italy, published in Nature Human Behaviour, reveals that AI debate can surpass human debaters in persuading others. Concerns have risen about the potential impact of AI persuasion on political debates, misinformation, and elections.
In the study, 900 participants were paired in online debates. A third debated human opponents, another third faced debate AI (powered by GPT-4), and the final group debated both AI and human opponents with access to personal information about each participant.
Could this AI debater actually be more convincing than real figures?
Human vs AI Debate
Francesco Salvi, the lead author from the ETH Zurich, highlights the risks of choosing to debate AI in political contexts.
“If persuasive AI can be deployed at scale, you can imagine armies of bots microtargeting undecided voters, subtly nudging them with tailored political narratives that feel authentic,” Salvi says.
He warns this type of influence is difficult to track and disqualify currently, something that may be disastrous for elections, but also finding potential for positive applications of AI persuasion, like political polarization readuction and healthier habits.
“I would be surprised if malicious actors hadn’t already started to use these tools to their advantage to spread misinformation and unfair propaganda,” he adds.
An AI debater performed similarly to humans when no personal data was involved. But, when given personal information, AI was significantly more persuasive, changing participants’ opinions 64% of the time.
“It’s like debating someone who doesn’t just make good points: they make your kind of good points by knowing exactly how to push your buttons,” says Salvi, highlighting the effectiveness of AI persuasion.
The Potential of Generative AI for Personalized Persuasion at Scale
EPFL’s research, also indicates that GPT-4 can become extremely persuasive if loaded with personal details.
“In recent decades, the diffusion of social media and other online platforms has expanded the potential of mass persuasion by enabling personalization or ‘microtargeting’—the tailoring of messages to an individual or a group to enhance their persuasiveness,” researchers wrote.
AI Debate software, like GPT-4, is becoming more sophisticated in shaping opinions. Social psychologist Professor Sander van der Linden of the University of Cambridge warns that generative AI debate can shape public opinion on a large scale.
Against potential dangers, University of Oxford computer science professor and AI researcher Michael Wooldridge, points to favorable applications of AI in health or education, where arguing with AI would be capable of advancing learning. But he highlights the danger of misuse of AI.
“As AI develops, we’re going to see an ever-larger range of possible abuses of the technology,” he cautions.
With the ability to customize arguments based on one’s data, GPT-4 and other debate AI models can potentially transform political campaigns and online debates, because they can adapt based on personal preferences. This new generation debaters have the tendency to be better than humans in debates in an effective way.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.