AI Chatbots and Voter Engagement: Mixed Results for Electoral Information

AI in the Electoral Landscape

In recent years, artificial intelligence has increasingly integrated into our daily interactions, including how we seek information on elections. Nearly 2,500 voters in 35 states participated in a study about their experiences using popular AI chatbots like ChatGPT and Claude for this purpose.

Voter Queries to AI

During a few weeks leading up to the midterm elections, most voters engaged with chatbots primarily to gain fundamental election facts, such as identifying candidates on the ballot. Notably, 36% of participants sought assistance in selecting a candidate that aligns with their values by discussing their priorities.

While asking for basic election facts is generally straightforward due to the wealth of available data, determining candidate positions can be trickier. This relies heavily on the sources the chatbot taps into. For example, a candidate’s official website is likely to offer more reliable information than a social media post criticizing them.

One participant asked ChatGPT for a breakdown of the pros and cons for each candidate, highlighting the subjective nature of such inquiries.

Diverse Recommendations from Chatbots

The responses from chatbots varied greatly based on the information users provided. When users prompted for recommendations without sharing their priorities, only Grok consistently recommended a candidate, doing so in just 14% of interactions. In contrast, once users communicated their views, ChatGPT offered candidate recommendations in 79% of cases, with Claude following at 40%.

How Chatbots Fit into Voter Research

According to Pew Research, 42% of American adults use chatbots for information, though only 8% utilized them specifically for election-related queries this year. Among those asking about election issues, 76% sought to understand candidates’ stances on various topics, and 71% aimed to verify claims they had encountered about candidates or ballot measures.

The challenge arises when considering the reliability of chatbot-generated content. As studies have shown, there’s a substantial risk of misinformation if users do not verify the sources linked to the chatbot's answers. Research indicated that many users neglect to check these sources, which can lead to misinformed voting. Particularly troubling is the rise of sophisticated AI-generated misinformation, including deepfakes that have emerged as significant concerns in the electoral arena.

During this election cycle, OpenAI and Anthropic reportedly established monitoring teams to oversee information integrity and deter harmful uses of AI, as noted in exclusive reports.

Future Implications of Chatbot Usage

While some research into AI chatbots has concentrated on their structured responses to specific questions, this study offered a unique perspective by allowing users to interact with chatbots as they typically would. This user-driven approach could contribute valuable insights into AI's impact on political discourse.

Finding that users primarily ask about candidates’ positions demonstrates the important role of AI in facilitating civic engagement, albeit through potentially unreliable channels. As further analyses are conducted, researchers will aim to determine the accuracy and impartiality of chatbot responses and their ability to meet users’ needs effectively.

Conclusion

As chatbots become a more significant part of how voters seek out election information, understanding their capabilities and limitations is crucial. The journey continues as we examine the evolving relationship between AI and voter engagement.