Chatbot Voting Advice Could Help and Harm Voters. Let’s Regulate Accordingly. | TechPolicy.Press

An article published by TechPolicy.Press argues that chatbot voting advice could both help and harm voters, and calls for regulation. The article cites a UK survey indicating that one in seven voters were open to using AI chatbots to decide how to vote, with nearly 5% having already done so. The piece outlines several risks associated with chatbot-based voting advice, including advertiser influence, misclassification of voter intentions, hallucinated facts, and bias. The article is presented as an opinion piece arguing for proactive regulatory measures before chatbot use in voting becomes more prevalent.

The author contends that these risks are inherent to current large language models and could undermine democratic processes if left unaddressed. According to the article, advertiser influence could lead to chatbot recommendations being swayed by paid promotions, while misclassification could result in votes being miscounted or misinterpreted. Hallucinated facts could cause voters to act on false information, and bias could systematically favor certain candidates or parties. The article does not specify the exact survey methodology or sample size but presents the figures as evidence of growing reliance on AI for political decisions. It also notes that chatbot advice could potentially help voters by providing accessible information, but argues that the harms currently outweigh the benefits.

The article suggests that regulation is needed to mitigate these risks, though it does not propose specific policies. It positions the UK as a case study in the intersection of AI and electoral integrity, and calls on policymakers to act before chatbot use becomes more widespread. The article concludes that without regulation, the potential harms of chatbot voting advice could significantly impact electoral outcomes.