My version of your synthesis is something like as follows:
This is closer; I’d just add that I don’t think activism is too different from other high-stakes domains, and I discuss it mainly because people seem to take activists more at face value than other entities. For example, I expect that law firms often pessimize their stated values (of e.g. respect for the law) but this surprises people less. More generally, when you experience a lot of internal conflict, every domain is an adversarial domain (against parts of yourself).
I think there’s a connection between the following
storing passwords in plain text|encrypting passwords on a secure part of the disk|salting and hashing passwords
naive reasoning|studying fallacies and biases|learning to recognise a robust world model
utilitarianism|deontology|virtue ethics
I think you lost the italics somewhere. Some comments on these analogies:
The idea that some types of cognition are “fallacies” or “biases” and others aren’t does seem like a pretty deontological way of thinking about the world, insofar as it implicitly claims that you can reason well just by avoiding fallacies and biases.
As the third step in this analogy, instead of “learning to recognize a robust world model”, I’d put “carrying out internal compromises”, i.e. figuring out how to reduce conflict between heuristics and naive reasoning and other internal subagents.
Re the passwords analogy: yes, deontology and virtue ethics are adversarially robust in a way that utilitarianism isn’t. But also, virtue ethics is scalable in a way that deontology isn’t, which seems well-captured by the distinction between storing passwords on secure disks vs salting and hashing them.
This is closer; I’d just add that I don’t think activism is too different from other high-stakes domains, and I discuss it mainly because people seem to take activists more at face value than other entities. For example, I expect that law firms often pessimize their stated values (of e.g. respect for the law) but this surprises people less. More generally, when you experience a lot of internal conflict, every domain is an adversarial domain (against parts of yourself).
I think you lost the italics somewhere. Some comments on these analogies:
The idea that some types of cognition are “fallacies” or “biases” and others aren’t does seem like a pretty deontological way of thinking about the world, insofar as it implicitly claims that you can reason well just by avoiding fallacies and biases.
As the third step in this analogy, instead of “learning to recognize a robust world model”, I’d put “carrying out internal compromises”, i.e. figuring out how to reduce conflict between heuristics and naive reasoning and other internal subagents.
Re the passwords analogy: yes, deontology and virtue ethics are adversarially robust in a way that utilitarianism isn’t. But also, virtue ethics is scalable in a way that deontology isn’t, which seems well-captured by the distinction between storing passwords on secure disks vs salting and hashing them.