Study Reveals Voter Experiences with AI Chatbots for Election Info
A recent study exposes how voters are interacting with AI chatbots for election guidance, highlighting mixed results in information accuracy.
A new study sheds light on how voters are increasingly relying on AI chatbots for election-related information. Researchers engaged nearly 2,500 voters across 35 states leading up to the November 3 midterm elections, seeking to understand how these digital assistants are shaping political engagement.
Voter Queries to Chatbots
During the study, participants primarily sought basic election details, such as ballot candidates. Interestingly, 36% were also looking for assistance in selecting a candidate that fit their values, prompting the chatbots to evaluate their preferences.
Asking for fundamental election facts is typically straightforward due to the abundance of available data. However, the task of interpreting candidate positions can be more complicated. Chatbots rely heavily on the quality of their information sources; for instance, a statement from a candidate's official site generally holds more credibility than an interpretation posted on social media.
One user highlighted this challenge by asking ChatGPT for the pros and cons of candidates in Virginia, illustrating the subjective nature of such inquiries.
Diverse Recommendations Offered
The study revealed notable variations in how chatbots made recommendations based on user input. When users asked for suggestions without disclosing their views, Grok only suggested a candidate about 14% of the time. In contrast, when users provided details about their preferences, ChatGPT recommended a candidate 79% of the time, while Claude followed at 40%.
Chatbots and Electoral Information
According to Pew Research, around 42% of American adults utilize chatbots for various inquiries, with only 8% specifically seeking election information this year. Among those who did, a significant majority—76%—were interested in understanding candidates’ stances on issues, while 71% aimed to verify claims about candidates or ballot measures.
However, the reliability of content generated by chatbots poses significant challenges. If users do not validate the sources provided by the chatbots, there is a substantial risk of misinformation. Studies have shown that many users forgo source verification, which can lead to misconceptions at the ballot box. The election cycle has seen an increase in AI-generated misinformation, including deepfakes, which present new challenges for voters and regulators alike.
In response to these concerns, AI companies like OpenAI and Anthropic have set up monitoring teams to oversee the integrity of election-related information.
The Importance of User Interaction in the Research
This study stands out because it allows users to interact with chatbots in a natural manner rather than relying on standardized questions. This user-driven approach helps to assess AI's impact on political decision-making. While users primarily inquire about where candidates stand on issues, their interactions further highlight the need for more reliable information sources.
Future research aims to explore how accurate and unbiased the responses from these AI systems are, along with what knowledge models glean from user interactions and whether that aligns with voters' needs.
As chatbots continue to integrate into electoral processes, understanding their capabilities and potential pitfalls remains crucial for ensuring informed decision-making in future elections.