Large-Scale Decision Making
Sent to Danielle Fong, the co-founder of the energy-storage startup LightSail, sketching the long-term software project Alyssa hoped one day to build: a tool for finding the smallest interventions that would change collective outcomes at the scale of nations.
The US interstate highway system was a big project to provide a global benefit, in spite of localized harms. It (mostly) succeeded.
Reducing the amount of CO2 emissions is a big project to provide a global benefit, in spite of localized harms. It has (mostly, so far) failed.
Why?
If you studied history and looked at the details, you might discover an answer like “the US government used to be more authoritarian”. And the reason for that might be “the US government used to have more control over the media”, and so on, going back through the historical record. These answers might be accurate, but they aren’t very useful from a planning perspective. Even if you knew for sure that (say) a more centralized government would help, there’s no knob anyone can turn to increase centralization from 3 to 6. In business-speak, increasing centralization isn’t “actionable”.
Ultimately, there is no one who can take collective actions; people can only take individual actions. To really understand — understand in enough detail that you could reliably change the outcome — why the highways worked and Kyoto didn’t, you would need a way of understanding how individual actions added up to produce the collective outcome. Then you could say that if Bob did X instead of Y, that would help fix things, and then maybe Bob might really do Y.
Luckily, at least for recent times, we do have explanations. If you called up a bunch of government agencies, and paid some fees, you’d get an exact explanation of how various people’s actions ultimately produced success in the highway case and failure in the other. Unfortunately, the explanation would come in the form of a freight train filled with boxes of records, which isn’t very useful either. (I speak from experience.)
If you had that freight train (and lots and lots of time), it’d be relatively straightforward to create some counterfactual under which the highways failed and Kyoto succeeded. If the President had different ideas, and the Cabinet had different ideas, and the clerks had different ideas, and the millions of Americans who wrote their Congresscritters had different ideas, and so on, there would have been a different outcome. But that wouldn’t really work as a plan; convincing millions of people, one at a time, would take way too long.
What you really want is some way of finding, within that big freight train, the smallest successful intervention . Some place where, if two or five or ten people did the right things in the right order, that would cause fifty people to do something and that would cause five hundred people to do something and on and on, until things changed on a global scale. What you need is some kind of software, to read in that whole freight train, and help a human find where the interventions are.
In theory, small interventions might not even exist, but I think in practice they usually will. And you can point to accidents that changed the world, or Great Man theories of history, and so on. But I think there’s also an even more convincing argument: the only way anyone knows about anything (and therefore makes decisions about anything), except for the very small subset that she and her friends can see for themselves, is through a tiny narrow pipe traditionally called “the media”. This is a bit less true with the Internet, but even Wikipedia and similar sites are ultimately based on “reliable secondary sources”. “Everyone knows” that cholesterol is bad for you, in the end, because they (or their friends or parents or teachers) heard about it on TV.
When people talk about “media-influenced social change”, it’s most often about political machines which fill the airwaves with hamfisted propaganda about the glory of Arstotzka. This usually doesn’t accomplish their goals, and creates lots of other problems besides; I have no idea what one would do as an effective small intervention, and the ideas people get off the top of their heads tend to be pretty bad. But I think we can know that there does exist some path — some twenty-year-long sequence of wheels and pistons and Congressional hearings — that starts with a tiny group of people with a biology paper on one end, and has Grandma not eating eggs on the other. The only question is finding it.
I want to try writing that software (over ten or fifteen years, with many intermediate steps). Maybe it’s crazy. But I think it’s worth trying.