AI Chatbots Give Inaccurate Election Voting Advice

AI chatbots offer inaccurate, inconsistent, and unreliable guidance to voters seeking electoral advice, according to a study conducted during Hungary's parliamentary elections. The research, published by the civil liberties group Liberties, found that these general-purpose AI systems frequently recommended the wrong political parties, failed to mention the correct ones, or listed parties that were not even participating in the election.

The findings raise significant concerns about the dependability of AI systems when used in electoral contexts. Identical prompts given to the chatbots resulted in highly volatile and contradictory answers, underscoring their unsuitability for providing voters with factual and trustworthy information. The study highlights a critical gap in the reliability of AI for sensitive applications like political decision-making.

Liberties' research specifically focused on the Hungarian parliamentary elections, where the AI's performance was evaluated based on its ability to provide accurate party recommendations. The inconsistencies observed suggest a fundamental issue with how these AI models process and present political information, potentially misleading voters and undermining the democratic process. The study's conclusions point to a need for greater scrutiny and regulation of AI applications in political spheres.