How AI Could Affect the Election This November

I'm optimistic about how AI might change voting.

Roughly a third of Americans identify as Democrats and a third as Republicans. The remaining third — the independents who decide elections — have the most to gain from AI, because they're the ones party labels serve worst.

Google search let you type "November election" and find websites arguing for or against candidates and propositions. Useful, but it amounted to locating advocacy that already existed, written months earlier for someone other than you. More arranged marriage than courtship.

An AI conversation works differently. You can say: "I think wealthier people should pay more in taxes, but I'm unsure about the mechanism in Proposition 40.[California Proposition 40, One-Time Wealth Tax for State-Funded Healthcare, Education, and Food Assistance Programs Initiative, more popularly known as the billionaire tax.]

"What are the pros and cons?" Or: "Are there studies on how a billionaire tax might affect startup activity in California?" You bring your own bundle of values and half-formed theories about how the economy works, and the answers get built around them rather than around a position someone was already selling.

You might come away more committed to the billionaire tax than any group campaigning for it. You might find it violates something you care about and decide you can't support it. Either way, the conclusion is more likely to be yours. (And yes, one is vulnerable to manipulation from AI. The odds of that are never zero but my own experience of exploring AI's ability to parse arguments suggest they are at least as objective as policy wonks and professors - perhaps more so. And their responses - unlike found content on the internet - is a direct response to your specific inquiry. And one is hardly limited to a single AI for such queries.)

The shift is from finding a prerecorded message to having a conversation — one where you're heard before you listen.

If this works, political advertising should lose some of its grip. A voter who can say "this matters more to me than that" is harder to move with thirty seconds of television, and money in politics buys less.

Here's a way to test it: the share of ballots that include votes on propositions.

Down-ballot measures consistently draw fewer votes than races for president, senator, or governor. Propositions are confusing and time-consuming to understand, and voters who don't know how one aligns with their values often skip it rather than guess.

If people find a process that gives them confidence about an obscure measure, that drop-off should shrink. Rising proposition turnout wouldn't prove AI caused it, but it would be the first visible sign that voters have found a new way to decide what they actually want.