Artificial intelligence and the ballot box

Fourteen minutes after the hour, more Americans are turning to artificial intelligence for answers, including answers about politics and elections. But how do these apps make sure the information they're providing is accurate and unbiased? Joining us is Joe Mirnick, a Ph.D. candidate at Stanford, studying how voters use A.I. tools and how using these tools shapes their attitudes. Joe, thanks for joining us this morning. Morning, Andre. Thanks so much for having me. All right. So, Joe, millions of Americans are heading to the polls and half the country, they now use AI. Why should a voter or should a voter ask an AI chatbot about their ballot instead of opening that traditional voter guide that we all get in the mail? Yeah, it's a it's a new technology that a lot of people are turning to. And for good reason, I think there's the ability for these tools to cut through a lot of the framing and other information that candidates might put out and just give a voter the facts that they're looking for. At the same time, we don't know a lot. right now about what kind of information they're going to be providing. So far, a lot of the models don't seem to be providing recommendations and are instead just comparing the candidates on the ballot. But it's important, just like with any information that people are consuming, that people are paying attention to the sources that they're looking at and making sure that they believe what the chatbot is drawing on. And so going forward, I think I think it's better than nothing to use an AI voter guide, but it's still probably not the first choice that I would recommend to most voters out there, unless there is no other option. Right. Well, you're currently testing a purpose-built civic tool like VoteMate against giants like ChatGPT, Claude Grok. When a voter asks who to vote for, do general purpose bots really handle political bias safely or are they steering people without knowing it? That's something that we're studying right now. It's a little hard to know because these models are changing constantly. And right now they seem to be. Being pretty careful about which information they're providing, and they're drawing mostly on sources that we tend to think of as pretty high quality, sources like Ballotpedia or official government websites. But at this time, it's still an open question what kinds of bias might be sneaking in, and that's something that we're looking at right now. So your research shows AI voter tools actually increase someone's intention to turn out and vote. Is AI truly making voters smarter, or is it just giving them false confidence at the ballot box? We're not sure yet. I think more people voting is probably for the best, but of course we want to make sure that those voters are coming out with good, complete information. So as long as people are really engaging with the sources that these tools are citing and making sure that they trust what information they're getting instead of just sort of blindly reading what's on the screen and believing it as fact, I think it could be sending voters to the polls with more knowledge. So we know like, you know, big campaigns, you know, district campaigns for local Senate, maybe, you know, Congress nationally. We know those are out there, but we also have school boards and city measures. How is that information being surfaced through on these AI chatbots? Are you noticing that they're providing good information on these smaller school board members, you know, city measures that people have to vote on as well? I can't speak to whether the information is good right now because we haven't done a thorough review, but I can say that it is a much easier way to sort through a lot of the local news information that might cover these smaller candidates because they typically don't have as much funding to put out a website or a lot of good information. And so if they've been interviewed, for example, by a local news organization, it's likely that the chatbot I thought we'll be able to find that and maybe. Find it more easily than a voter would if they're just searching on something like Google. What do you think about the neutral civic AI tools? Can they actually help? Is that something that you're a part of your studies that you're looking into in your studies? I think we have pictures of those that go ahead, Joe. I think there is a lot of potential when we know that the tools are being cited or the tools are citing information that is pretty accurate. Places like Ballotpedia or USA Facts typically have fact-checked nonpartisan information. And so a tool like VoteMate will prefer those kinds of sources. And we're not sure if the frontier models like ChatGPT will do that yet. Still early in the game with all these new AI tools. So we'll see what happens. Joe Miernick, thank you so much for joining us and providing your insight into this this morning. Appreciate it. Thanks for having me. Have a good one.