Forecasting Democracy
Measuring today's politics, and predicting tomorrow's.
Welcome to Aligned Foresight!
This blog is dedicated to exploring the tools and technologies we use to gauge public opinion and predict the future of American politics. From political forecasting and machine learning models, to polling methods and voting systems, this blog will dive into how we try to understand where we’re headed and what we want to do about it. Through analysis, open-source code, and community engagement, Aligned Foresight aims to critique and improve these methods, helping us all stay better aligned in how we tackle the future.
America’s future has never been so unclear.
We probably all have some sense of what our weekend plans will be, what the week’s weather will look like, and what dreaded deadlines are looming on our calendars. Beyond that horizon, there are also some relative certainties: there will be a new president in the U.S. on January 20th, 2025, the country will have its Semiquincentennial less than two years from now, and bar some Kardashev shenanigans, the sun will still be around in 1,000 years.
But these days, when it comes to many of the most important predictions, even our most reliable crystal balls have never been more clouded. At the time this blog is being launched, the presidential race has been a dead heat for months, hovering at around 50-50 odds no matter who you ask. It’s a rare moment of consensus not only among all the political betting markets1, but also among all the leading statistical models2: there is no clear frontrunner.
The murkiness extends beyond politics.
Climate change continues to defy the expectations of esteemed climatologists3, powerful statistical models, and finely tuned million-line-of-code simulations4.
Strangely enough, our reliance on statistical simulations is itself bringing uncertainty to the future, as philosophers, tech luminaries, and Nobel prize winners warn of a Singularity, when AI models become better at improving themselves than the humans overseeing them, creating a runaway dynamic whose outcome is impossible for any of us to predict today5.
Complicating things further, the attention economy6, the rise in authoritarianism7, and the relentless empowerment of the top one percent8 have turned the world into a stage for fickle minds to prove their worth with grand gestures, leaving everyone guessing what the most influential media, tech, and world leaders will do next in increasingly high-stakes battles for dominance.
There’s a universal principle at play here: truth is stranger than fiction. As soon as we deceive ourselves into thinking we’ve grasped it — that we have a decent model of reality — we’re immediately presented with new and harder evidence that we haven’t. The uncertainties we collectively contend with today, almost by definition, are the most intractable we’ve faced yet.
So here’s one safe prediction for the future: even our best forecasting tools will someday fail us.
18 days until Election Day.
We’ve established that the future is as uncertain as ever. Yet in less than three weeks time, over 160 million registered American voters will be counted on to weigh in on it together. Their voice will be recorded for posterity and painted soon thereafter as irrefutable proof that one or the other way is the only way forward.
Meanwhile, faith in how America’s voice gets recorded is at an all-time low. One side will point to the undemocratic Electoral College, the archaic rules for appointing House and Senate members, and the politicized Supreme Court as evidence that sorely unrepresentative conclusions about their views have recently been drawn.
The other will point to a fragile voting process, a left-leaning media establishment, and the rise of big fundraisers and special interests as proof that their own input has actively been silenced.
Perhaps both narratives are reflective of frustrations with yet another universal principle at play: power is sticky. Over time, even the fairest tools for collective decision-making will invariably become biased toward the ruling classes. And many would say our present tools were never even particularly fair to begin with.
Let’s find ways to do better.
Because of the principles outlined here, society’s methods, both for predicting the future and for polling public opinion, are perpetually doomed to sometimes be wrong. As such, there will always be opportunities to improve these techniques. That is where this newsletter hopes to help out.
Aligned Foresight will cover society’s tools and technologies, both proposed and in use, for predicting where we are headed and for measuring what it is we really want to do about it. That will include forecasting techniques, statistical models, and machine learning, as well as polling techniques, voting systems, and political science.
This newsletter will also be exploratory in nature: rather than simply describing these tools, it will seek to implement, use, and critique them. All code, data, and methodology will be made open-source in the hopes of fostering a community of learners, and most importantly, in the hopes of keeping our foresight aligned.
Polymarket has Trump at 59 percent, PredictIt at 56 percent, Manifold at 51 percent, and Metaculus at 50 percent — all essentially coin flips.
538 has Trump at 52 percent, Nate Silver at 52 percent, The Hill at 50 percent, JHK at 49 percent, Race to the White House at 47 percent, The Economist at 46 percent, and Split Ticket at 46 percent.
Unless you claim you can accurately simulate a superintelligence, in which case congratulations, you are a superintelligence!
The attention economy is causing language itself to become unpredictable, per this Nature paper.
This dynamic even extends to tech leaders, who are asserting control over digital spaces to sidestep democratic accountability, as per this Atlantic feature.