Predicting a coup looks like a fascinating sociomathematical challenge, halfway between Asimov’s psychohistory and recurrent military dreams of geopolitical precognition. As Gizmodo wrote about, its possibility, even partial, suggests uneasy questions about the implications of that kind of power.
The problem is that this isn’t true.
Sure, there are unstable countries that seem ever-vulnerable, and sometimes surprising coups in places that didn’t seem so, but democratic collapse nowadays isn’t subtle or complex to predict. Donald Trump was literally saying he might not accept a negative election result since long before the election. The January 6 coup attempt wasn’t some sort of black swan; it seemed surprising only because the major political analysis frameworks in the US are in the complex cognitive situation of having one of the major parties engaged in a decades-long subversion of the democratic system but not feeling themselves able to allow themselves to see it. Predicting it, or something like it, wasn’t difficult because of chaos theory or some other unfathomable technical difficulty; it was simply an issue of willful refusal to take evidence at face value.
The US is far from the only democracy facing this sort of overt attack from a major or at least legitimate political party. It’s not the one closest to becoming a Potemkin democracy, but it’s not the one furthest from it either. Across the world, predicting or diagnosing future or ongoing attacks on democracy requires no more algorithmic subtlety than listening to what politicians say and taking them at face value. A politician who argues for an ethnonacionalist country or against an independent judiciary or media is, at the very least, open to the possibility if given enough power and motive.
Much like climate change, and often with the same actors, the main difficulty isn’t algorithmic or scientific, but rather that
- The institutions in a political system cannot cope institutionally with a combination of deliberate subversion and passivity or collusion from the professional political actors.
- Enough of the population is actively on board with de-democratization, and enough is at least open to the possibility of erosion of democratic rules in exchange for other political goals, that there’s no extra-institutional pressure to stop it either.
What can AI and data science do here? We could, and some very smart and dedicated people do, analyze evidence to present it in the most compelling way, but what do you do when the problem isn’t that something is unknown, but rather that too many people in power, and too many people on the street, are ok with it? What’s the algorithm for a politician who happily trades eroding democratic rules in exchange for short-term political benefits? What sort of AI helps when a strong minority, and sometimes a majority, of a country happily votes for them?
The simplest explanatory answer, and as data scientists it’s the one we are professionally bound to, is Nothing. We have a tool, and it’s a powerful and useful one, and it has its role here. But it might not have more of an impact that it already has, and it’s not enough.
The thing to do, I think, is to go back to the domain experts, seek those who have effected change and are trying to do it, and ask them how we can help, if at all. The experts on a given tool are always the first to explore its potential, but a tool only achieves real impact when it’s put at the service of those who are building a larger thing.