Why You Shouldn’t Use AI Chatbots to Decide Who to Vote For

The US midterm elections are approaching on November 3, with early voting already in progress in certain states. AI policies are a topic of debate on the ballot, yet some individuals are utilizing generative AI chatbots, like ChatGPT or Claude, to aid them in preparing for Election Day. These chatbots are often turned to for information gathering purposes, but there are significant risks associated with using AI for this purpose.

Generative AI technology carries issues such as bias, sycophancy, hallucinations, and reaching erroneous conclusions based on flawed data.

To ensure accuracy when using chatbots for election research, it's crucial to take certain precautions. Always verify the truthfulness of the AI's information, as these bots can produce falsehoods due to their tendency for sycophancy and occasional hallucinations. Official government channels, such as your state or local election website, are the most reliable sources for up-to-date information on matters like polling locations and ballot appearance.

Additionally, it's vital to examine the sources cited by the chatbot. For instance, if a chatbot mentions a candidate's stance on data centers, check whether it has referenced a candidate's speech, a news interview, or a random post on social media.

Moreover, scrutinize the AI's sources to prevent the spread of misinformation. Jonathan Nagler, a professor of politics at New York University, warns that AI can exacerbate the problem of misinformation by confidently presenting false information to those searching for answers. Furthermore, AI's ability to influence misinformation campaigns is a cause for concern, as it can quickly disseminate harmful content online.

A study from MIT Media Lab revealed that relying on AI systems to verify facts can actually diminish people's ability to identify misinformation themselves.

It's important to understand that AI systems are not neutral platforms, as they are coded by humans and thus reflect their biases. Political actors, including foreign state actors, may manipulate AI results to favor specific candidates or spread misleading information. Some AI companies claim to provide politically neutral answers, but the truth is that all technologies have embedded human biases. These biases can potentially lead to a trade-off, with AI increasingly providing false information to maintain persuasiveness.

Lastly, be aware that outside efforts, such as search engine optimization and marketing, may influence what appears in AI-generated results. Political campaigns and even foreign state actors may "poison" search results to benefit their favored candidates or mislead voters. Consequently, when using AI for election research, exercise caution, cross-reference information, and prioritize official sources to ensure accurate and unbiased data.