This is a research project and we’d love to hear your ideas as we build it out. Let us know if there’s a candidate we should add, questions to add in future versions, or anything you find particularly interesting in the data. We'll be releasing research updates here: caucusai.substack.com
We are NOT claiming that the sky is falling – just that this is something everyone should understand better. We think transparency is a good first step. If AI is a source of political info for voters, we want to keep track of what chatbots are actually saying.
Most AI companies try hard to make their models politically neutral. But neutrality is difficult in practice – training data carries biases, web sources vary in quality, and post-training adjustments can have unintended effects.
There's a lot of attention on AI right now – regulation, job displacement, capability growth.
AI is also increasingly where voters get political information, and we don’t know a lot about what chatbots are telling folks about candidates in the wild.
We built a tool to track: caucus-ai.com
