if you have a thousand organisations each pushing in a different cardinal direction in some high-dimensional space, getting backing and making progress based on how important it is to varying numbers of people, that looks a lot like some sort of gradient descent. Maybe this sort of single-issue focus isn’t as inefficient as it might appear?
There are plenty of ways this analogy can break down, and also plenty of ways it can go wrong even within the analogy. A major victory in one direction can easily “overshoot” into a highly sub-optimal state (e.g. revolution), or various factors can consolidate a lot of update power into just two opposed directions (e.g. polarized two-party states).
Plus of course gradient descent is generally based on some error function that can be evaluated precisely and doesn’t change while you’re trying to optimize, neither of which is true in politics, so the analogy is far from perfect.
It’s a reasonable model. One problem with this as a predictive model, however, is that log-rolling happens across issues; a politician might give up on their budget-cutting to kill an anti-business provision, or give up an environmental rule to increase healthcare spending. So the gradients aren’t actually single valued, there’s a complex correlation / tradeoff matrix between them.
It seems like large organizations achieve structure through a combination of legislation and value-setting. They use policies and rules to legislate nuance, but rely on a single value to steer daily decision-making. This whole analysis really needs to be understood as being about the daily decision-making piece of the puzzle.
I think this ignores how decisions actually get made, but I think we’re operating at too high a level of abstraction to actually disagree productively.
if you have a thousand organisations each pushing in a different cardinal direction in some high-dimensional space, getting backing and making progress based on how important it is to varying numbers of people, that looks a lot like some sort of gradient descent. Maybe this sort of single-issue focus isn’t as inefficient as it might appear?
There are plenty of ways this analogy can break down, and also plenty of ways it can go wrong even within the analogy. A major victory in one direction can easily “overshoot” into a highly sub-optimal state (e.g. revolution), or various factors can consolidate a lot of update power into just two opposed directions (e.g. polarized two-party states).
Plus of course gradient descent is generally based on some error function that can be evaluated precisely and doesn’t change while you’re trying to optimize, neither of which is true in politics, so the analogy is far from perfect.
This is going to be a fun idea to think about! Thanks.
It’s a reasonable model. One problem with this as a predictive model, however, is that log-rolling happens across issues; a politician might give up on their budget-cutting to kill an anti-business provision, or give up an environmental rule to increase healthcare spending. So the gradients aren’t actually single valued, there’s a complex correlation / tradeoff matrix between them.
It seems like large organizations achieve structure through a combination of legislation and value-setting. They use policies and rules to legislate nuance, but rely on a single value to steer daily decision-making. This whole analysis really needs to be understood as being about the daily decision-making piece of the puzzle.
I think this ignores how decisions actually get made, but I think we’re operating at too high a level of abstraction to actually disagree productively.