People are asking ChatGPT to help them decide how to vote in the midterms
Adam Johnson of Morgantown, West Virginia, spent hours asking ChatGPT about candidates on his midterm ballot. He said the chatbot is already part of his daily routine.
Johnson asked the tool to list his ballot’s contests, then explored candidates in light of his own principles. ChatGPT said it could compare evidence and priorities, but should not direct his vote.
Other voters told NPR they use chatbots to research races with limited coverage, compare candidates, check viral claims and build ballot lookup tools. Pew Research Center data cited in the article show widespread chatbot use among U.S. adults.
Researchers warn that chatbots can influence users even when they appear to be offering neutral help. Rafael Batista of Johns Hopkins University said responses may reinforce a person’s existing leanings and leave them more certain without improving their understanding.
The article says it is difficult to assess potential bias because AI companies do not disclose how commercial models are trained or how they select sources. Voters using chatbots therefore have limited visibility into what may shape the answers they receive.
Key points
- Voters are turning to chatbots for candidate research, especially in down-ballot contests.
- Researchers caution that chatbot replies may reinforce existing political preferences.
- Limited disclosure about training and sources makes possible biases hard to evaluate.