Harnessing AI for Politics: A Look Back at the Warren Plan Bot

Author here. I think LLMs make for a really interesting opportunity for getting voters details of candidates' platforms that are relevant to them specifically.

Right now, campaigns strike me as very much a push model (especially at the federal level), where candidates and news media are putting out information they think will be relevant and resonate, and voters need to drink from something of a firehose if they have more specific questions, or care about issues that are less hot-button.

Of course, there are in-person events where voters can ask questions important to them, but that only covers a small fraction of voters, and doesn't generally support detailed question & answer. Volunteers are another key touchpoint, but most of them will only be so informed about detailed plans.

With LLMs, I think it's very imaginable for a campaign to allow digital question / answer so engaged voters can get more specific questions answered, while sticking to desired messaging. Using an LLM to provide _human volunteers_ better information when they speak to voters is another interesting opportunity for more effective messaging.