they’re looking for impossible certainty in an obviously context specific and highly variable situation
We’re looking for a decision procedure. “It’s context-specific; it depends” is a good start, but a useful proposal needs to say more about what it depends on.
A simple example of a decision procedure might be “direct democracy.” People vote on what to do, and whichever proposal has more votes is implemented. This procedure provides a specific way to proceed when people don’t agree on what to do: they vote!
In both the OP and your response to me, you tell a story about people successfully talking out their differences, but a robust institution needs to be able to function even when people can’t talk it out—and the game theory of “What happens if we can’t talk it out” probably ends up shaping people’s behavior while talking it out.
For example, suspects of a police investigation might be very cooperative with the “good cop” who speaks with a friendly demeanor and offers the suspect a cup of coffee: if you look at the radically transparent video of the interview, you’ll just see two people having a perfectly friendly conversation about where the suspect was at 8:20 p.m. on the night of the seventeenth and whether they have witnesses to support this alibi. But the reason that conversation is so friendly is because the suspect can predict that the good cop’s partner might not be so friendly.
I feel like having a trusted leader is a pretty clear tiebreaking decision procedure, no? However, the important parts of this model and the organizations I’ve been a part of is all the OTHER parts that come before that last resort, where people have a clear sense of values, buy into them, and recognize themselves or as a group when they’re not following them. But in the end, if all of those important bits failed, these organizations still have a hierarchy.
ETA: The decision procedure IS the values. The values are hard to pin down because values are hard to pin down, they’re taught through examples and rituals and anecdotes and example and the weights on the neural nets in people’s heads get to learn what following them and breaking them look like. Ultimately theres leaders who can help make tough calls and fix adversarial examples and ambiguous options and the like, but the important part of these organizations is mostly how they’re set up to train that neural net.
The decision procedure IS the values [...] taught through examples and rituals and anecdotes and example and the weights on the neural nets in people’s heads
That makes sense; I agree that culture (which is very complicated and hard to pin down) is a very important determinant of outcomes in organizations. One thing that’s probably important to study (that I wish I understood better) is how subcultures develop over time: as people leave and exit the organization over time, the values initially trained into the neural net may drift substantially.
Very well. I will endeavor to be more direct.
The fourth virtue is evenness! If you first write at the bottom of a sheet of paper, “And therefore, the aforementioned organization would work well!”, it doesn’t matter what arguments you write above it afterward—the evidential entanglement between your position and whatever features-of-the-world actually determine organizational success, was fixed the moment you determined your conclusion. After-the-fact steelmanning that selectively searches for arguments supporting that conclusion can’t help you design better organizations unless they have the power to change the conclusion. Yes requires the possibility of no.
We’re looking for a decision procedure. “It’s context-specific; it depends” is a good start, but a useful proposal needs to say more about what it depends on.
A simple example of a decision procedure might be “direct democracy.” People vote on what to do, and whichever proposal has more votes is implemented. This procedure provides a specific way to proceed when people don’t agree on what to do: they vote!
In both the OP and your response to me, you tell a story about people successfully talking out their differences, but a robust institution needs to be able to function even when people can’t talk it out—and the game theory of “What happens if we can’t talk it out” probably ends up shaping people’s behavior while talking it out.
For example, suspects of a police investigation might be very cooperative with the “good cop” who speaks with a friendly demeanor and offers the suspect a cup of coffee: if you look at the radically transparent video of the interview, you’ll just see two people having a perfectly friendly conversation about where the suspect was at 8:20 p.m. on the night of the seventeenth and whether they have witnesses to support this alibi. But the reason that conversation is so friendly is because the suspect can predict that the good cop’s partner might not be so friendly.
I feel like having a trusted leader is a pretty clear tiebreaking decision procedure, no? However, the important parts of this model and the organizations I’ve been a part of is all the OTHER parts that come before that last resort, where people have a clear sense of values, buy into them, and recognize themselves or as a group when they’re not following them. But in the end, if all of those important bits failed, these organizations still have a hierarchy.
ETA: The decision procedure IS the values. The values are hard to pin down because values are hard to pin down, they’re taught through examples and rituals and anecdotes and example and the weights on the neural nets in people’s heads get to learn what following them and breaking them look like. Ultimately theres leaders who can help make tough calls and fix adversarial examples and ambiguous options and the like, but the important part of these organizations is mostly how they’re set up to train that neural net.
That makes sense; I agree that culture (which is very complicated and hard to pin down) is a very important determinant of outcomes in organizations. One thing that’s probably important to study (that I wish I understood better) is how subcultures develop over time: as people leave and exit the organization over time, the values initially trained into the neural net may drift substantially.