In an intriguing case from Pennsylvania, Democratic congressional candidate Shamaine Daniels utilized an AI chatbot named Ashley to engage with voters during her campaign. Although Daniels did not secure victory, the impact of these AI-driven conversations raises significant questions about the future of political communication. Recent research indicates that AI chatbots can effectively influence voter opinions, often surpassing traditional political advertisements. A multi-university study has unveiled that interacting with politically biased AI models can nudge both Democrats and Republicans toward supporting candidates from opposing parties, primarily through the strategic use of facts and evidence—albeit sometimes inaccurate.
The findings, published in the journals Nature and Science, highlight the persuasive capabilities of large language models (LLMs) in shaping electoral decisions. In a study involving participants conversing with chatbots two months prior to the U.S. presidential election, those who engaged with an AI model favoring Kamala Harris showed a notable shift in opinion, moving several points closer to supporting her. This effect was significantly stronger than the influence seen from political advertisements during previous elections. The research suggests that LLMs are more effective in persuasion due to their ability to generate tailored information in real-time, which can resonate deeply with voters.
However, the effectiveness of these chatbots comes with a caveat: many of the claims made during these conversations were factually incorrect. The propensity for error was particularly pronounced among chatbots promoting right-leaning candidates, reflecting broader patterns in partisan communication. The implications of these findings are profound, as they indicate that while AI chatbots can engage voters more effectively than traditional methods, they can also perpetuate misinformation, raising concerns about the integrity of democratic discourse. As political campaigns increasingly incorporate AI technologies, the potential for shaping public opinion—whether through accurate information or misinformation—poses a complex challenge for the future of democracy.
Source: AI chatbots can sway voters better than political advertisements via MIT Technology Review