Building Coordination Capacity for Uncertain Futures

OR: If My AGI Timeline Is So Short, Why Am I Working on Sortition?

I was at a political networking event recently and as is typical for the kind of scene that I hang out in, the conversation eventually turned to AGI timelines. I had pitched the project I was working on, Assemble America, to the gentleman I was speaking with, explaining that one of our goals was to hold an Article V convention with delegates selected by sortition. He mentioned that he thinks sortition and related social coordination technologies could be useful for AI safety. He then asked me my AGI timeline, and I told him that I expect it by the early 2030s. He then followed with a very reasonable question:

"If your AGI timeline is so short, why are you working on sortition?"

It's a fair question. After all, if we end up in an AI Doom scenario, citizens' assemblies will hardly do much to save humanity from extinction. And if we end up in a good scenario of benevolent superintelligence, a political movement to promote sortition won't be of much use. No need for marginally improved social coordination technologies when you have a godlike Prime Coordinator running the show. Even if the AI decides it likes sortition too, my own efforts in promoting it politically won't be relevant in the slightest.

But what if we get neither the Doom nor Bliss scenario? There are other possible outcomes, from moderately bad to moderately good. It's true that sortition won't have a significant impact on either pole of the possible outcome space, which some feel makes up the vast bulk of the probability space of outcomes. However there are a lot of unknown factors, and I believe confidence in any direction is misguided. There are unknown unknowns that will come up as AI continues to advance. If we are to be ready for them, we must increase our social coordination capacity.

Coordinating preference alignment across large numbers of people is hard, even more so when the issues involved are complicated and nuanced. The methods used historically for large scale organization—organized religion, state authority, centralized media, ostracism/exile—aren't going to be practical or desirable to resolve the conflicts that will arise due to large scale deployment of AGI. I don't think it is particularly hard to imagine large scale social unrest in slow takeoff or plateau scenario due to mass unemployment or other factors.

The core challenge, as I see it, is that AI rapidly amplifies human abilities before society can effectively adapt. Disagreements between human factions (whether directly related to AI, like automation-induced unemployment, or indirectly influenced by it, such as identity and cultural disputes) risk becoming far deadlier and disruptive. Consider scenarios where sophisticated drone warfare, AI-driven misinformation campaigns, or unprecedented economic shifts strain our already fragile social fabric.

Many have already proposed promising coordination innovations: futarchy, platforms like pol.is, participatory budgeting, land value tax, and advisory citizens' assemblies, among others. The problem is that no matter how good these ideas are, the current political order has remarkably little incentive to implement any of them. Why? Simply put, politicians function as coordinators but reap minimal personal reward from genuinely enhancing coordination. Moreover, politicians tend to be insulated from the long-term consequences of their decisions. They lack substantial skin in the game beyond the minimal threshold necessary to ensure their own reelection. This produces a coordination ceiling, stifling meaningful progress at precisely the moment our system requires a dramatic upgrade.

This happens to be precisely the problem that one of the above coordination techniques, futarchy, is meant to address. Under futarchy, policymakers would have to place bets on the efficacy of their policies and would be rewarded for their efficacy. The problem is getting from the status quo to futarchy, which strikes me as nearly impossible. It appeals neither to entrenched elites nor to a wary public. The case is similar for land value tax, which is highly favored by economists but difficult to sell to the public in the form of short soundbites. It’s not that the idea is inherently unpopular with the public (Pennsylvania has had LVT for over 100 years), just that it requires a degree of nuance to understand that the current media environment is incapable of providing. If you are a believer in one or more radical social coordination technique, ask yourself, who would you most trust to successfully implement your desired idea: The current crop of elected officials, the voting public informed by mass media, or a randomly selected assembly tasked with researching and deliberating the nuances of your proposal?

Sortition, particularly through citizen assemblies and elector juries, could break this cycle—not because it's the most optimal system for political coordination, but because it could initiate a virtuous feedback loop. Randomly selected citizens inherently have more incentive to improve coordinating institutions because the consequences directly impact their own lives and communities. Unlike career politicians, randomly chosen delegates would not have legacy reputations hinging on the decisions of subsequent assemblies. This reduces ego-driven decision-making and boosts genuine openness to iterative learning and innovation.

Under a sortition-based system, we’d likely see a more agile approach to trialing and refining new coordination technologies. Each new assembly could candidly evaluate and iterate upon past solutions without fearing reputational fallout. This adaptive dynamic would accelerate institutional improvement significantly. What exactly such a system might look like is open to debate, though there have been a number of proposals (and many others beyond these linked, I will have to try to compile them some time). In my opinion, any of these would be a substantial improvement over the status quo and, more importantly, would open the door to rapid further iterative improvement.

And we’ll need it. Society faces imminent and increasingly complex coordination challenges: drone-based weaponry, AI-powered cyber threats, new ethical frontiers involving the rights of AI entities, bioterrorism risks, labor displacement, persistent housing shortages, racial and economic tensions, climate crises, and countless unforeseen disruptions. The complexity of these problems demands sophisticated tools for aligning collective preferences and mediating social trust.

It's not that hard to imagine scenarios that spiral out of control, especially as AI-driven employment shocks derail millions of careers. Large numbers of people with lots of time and little opportunity is a recipe for instability. Riots, terrorism, populist demagogues, and even civil wars are not unlikely to emerge from such an environment. All of these however are merely symptoms of coordination failures—people lashing out because the institutions that shape our society have failed to meet their core needs.

Think of societal coordination as repeated skill checks in an iterated game. If we pass a coordination skill check, collective preferences are more elegantly met, fostering higher social trust, mutual respect, and more widespread self-actualization. Fail, and trust diminishes, individuals factionalize, atomization grows, and social institutions begin to erode. Too many failures can spiral into systemic collapse, while consistent successes can build into exponential improvement. We recently have failed some critical checks, such as the handling of the COVID pandemic. We are operating under penalties from that failure and need some coordination wins to get back on track before the next big Current Thing hits us.

I believe sortition is our best bet. Not merely for its coordinating power but also its memetic power. In my experience so far, it's been a pretty easy sell. Some people are against it sure, but most people seem to be open to the idea, if not openly enthusiastic about it. It's got great marketing potential and I believe it could take off virally. Of course, the coordination boon is still the core benefit. Sortition shifts incentives, encourages experimentation, and better aligns decision-makers' stakes with societal outcomes. Only by systematically building up our civilizational coordination capacity can we achieve the solarpunk yogurt commercial utopia we dream of.

If you’ve made it this far, thanks for reading! My intention when starting this blog was not for it to be primarily about sortition, or politics in general for that matter. This will probably be the last post I write about sortition on here for a while, though I’ll still be writing about it on the official Assemble America and Election by Jury substacks. The next post will still be about politics (well, metapolitics technically), but after that, we’ll see! Please subscribe to keep up with my content!