PSJ Special Segment: The First AI-Assisted Election | The Pirate Street Journal

Podcast: Play in new window | Download (Duration: 23:42 — 16.3MB) | Embed

Subscribe: Apple Podcasts | Spotify | Pandora | RSS | More

The midterm elections wrapped up yesterday, and plenty of people are still sorting through the results. But the hosts of the Pirate Street Journal came away with a different kind of takeaway. This may have been the first AI-assisted election at any real scale, and each of them used the technology in a very different way. One built a voting framework with Claude, one cross-checked three chatbots on California propositions, and one skipped AI entirely and tested what campaign marketing does to a voter’s head. What they found was useful, a little uncomfortable, and in one case very funny.

You’re listening to Christopher Lochhead: Follow Your Different. We are the real dialogue podcast for people with a different mind. So get your mind in a different place, and hey ho, let’s go.

Building a Voting Plan With a Chatbot

Eddie went into this cycle with a problem most of us share. Beyond the two or three names at the top of the ballot, he had no idea who half the candidates were. So he gave Claude his address, asked it to pull up his ballot, and had it interview him about what he cares about. The result was a personal framework: a sensible economy, a reflective stance on AI, leaders who look forward, and measurable outcomes over speeches. He voted ahead of time, so the booth held no pressure.

The tool had gaps. It missed a few races, and judges and forest preserve candidates were beyond its reach because no information existed. Eddie filled the holes with a loose belief in checks and balances, voting against incumbents in a heavily Democratic Illinois and tossing a few random votes to Republicans for the sake of competition. When he had nothing at all, he confessed he voted for Asian candidates, a bit of what he called positive racism borrowed from a Ronny Chang comedy routine.

Cross-Checking Props and Local Races

Christopher faced a different headache in California. Many ballot propositions are worded so confusingly that even when you know what you want, you can’t tell whether to vote yes or no. He ran the props through ChatGPT, Perplexity, and Grok, and argued with each of them, since he figures a single chatbot is too eager to please. Gemini, he said, shut down any political conversation almost immediately. About two thirds of his props turned out to involve raising taxes or making it easier to raise them.

He was also surprised by how much the models knew about hyper-local races. ChatGPT had details on a city council contest in Capitola, a town of roughly 10,000 people, and even tracked the vote count drops. That raises an opening for entrepreneurs: a nonpartisan voting agent that gives someone in rural Alaska the same depth of information as someone in a California beach town.

Bias, Marketing Noise, and a Peaceful Handoff

Christopher raised the touchy question of who builds these models. He pointed to a stat that roughly 80% of engineers at the major labs lean left, and argued that bias deserves a public conversation, ideally solved by builders and not by new regulation. He also noted that models are designed to agree with you, so you have to tell them to give you facts and not flattery. Bri ran the opposite experiment. She skipped AI, judged the election by the flyers, texts, and TV ads she saw, then discovered the “vote no on number two” slogans had little to do with what was printed on her ballot.

Christopher closed on a note that stuck. He recalled George H.W. Bush saying that watching power change hands without a drop of blood shed is the real marvel. Whatever side you’re on this morning, the peaceful transfer is worth appreciating.

To hear more about the topics in this week’s The Pirate Street Journal, download and listen to this episode. You can also read more Pirate Street Journal entries in the Category Pirates newsletter.

We hope you enjoyed this episode of Christopher Lochhead: Follow Your Different™! Christopher loves hearing from his listeners. Feel free to email him, connect on Facebook, X, LinkedIn, and subscribe on Apple Podcast / Spotify!