I’ve always been fascinated by politics and polling. For about 20 years, I’ve absorbed everything I could find during election season, then gone back afterward to see what happened. Where were the polls right? Where were they wrong? What did they miss, and what could we learn?
When Nate Silver started his forecasting website, I was in awe. Here was someone putting numbers and analysis behind the questions I had been asking for years. Other forecasters followed, and every election season I would scour their sites, comparing predictions and trying to understand why they saw things differently.
I had plenty of ideas of my own. Things I wanted to test. Patterns I wanted to investigate. Questions I couldn’t answer just by reading someone else’s forecast.
Now, with AI (yes, NOT SI), I can put those ideas into words and actually work on them. ChatGPT’s Sol model has helped me collect and organize data, build tools and explore analyses that I could not realistically have taken on alone. Some work that once felt impossibly time-consuming can now begin in hours.
That is a wonder to me. It really is amazing. It’s how 21CEE became possible.
But speed does not make an answer correct. Neither does a polished chart or a confident AI response. The evidence still has to hold up. Ideas need to be tested against elections they weren’t trained on. And when something doesn’t improve the forecast, we need to say so.
I have strong political opinions. I’m not going to pretend otherwise. But what I want to happen in an election cannot determine what this project reports. Democrat, Republican or independent, I want to understand what people actually think, how they poll and how they vote. If the evidence challenges my expectations, then my expectations need to change.
That is why transparency matters so much to me. You should be able to see where the data came from, how it was used, what is uncertain and what we got wrong. We have to earn trust through that work, not simply ask for it.
American Mood is part of that curiosity: looking beyond the horse race at how people feel about the economy, their lives and the issues facing the country. Whether those signals can improve election forecasts is something to demonstrate, not assume.
My excitement about AI comes with serious concern. I believe it could bring changes on a scale humanity has rarely experienced, perhaps comparable to fire or the wheel, as the saying goes. A project like 21CEE is only a drop of water in that ocean.
I believe the United States, China and the other AI super powers need to act now, together, on meaningful regulation and safeguards. The possibilities are extraordinary. So are the risks. I can imagine a true golden age for humanity, and I can imagine catastrophic outcomes if we handle this badly.
I also believe AI will play a growing role in this and future elections, and could be a central issue on voters’ minds by 2028. Much will depend on what happens over the next year and a half, including what we learn from these midterms.
So I ask everyone to get involved. Whatever your opinion, make it heard. Contact your elected officials where you can. Ask questions. Take part in the discussion. People around the world deserve a voice in decisions that could fundamentally change their lives.
For my part, 21CEE is a way to keep asking questions, test ideas and share what we learn. I want the facts, even when they aren’t the facts I hoped for.