Sorry, I didn’t mean that “LW is outside standard left vs right.” I meant that “post-political” politics is categorically impossible when you can’t exhaustively evaluate Solomonoff Induction. You cannot reduce an entire politics to “I rationally evaluated the evidence and updated my hypotheses”, because the relevant set of propositions doesn’t fit the necessary axioms. Instead, I think we have to address politics as a heuristic, limited-information, online-learning utility-maximization inference problem, one that also includes the constraint of trying to make sure malign, naively selfish, ignorant, and idiotic agents can’t mess up the strategy we’re trying to play while knowing that other agents view us as belonging to all those listed categories of Bad People.
So it’s not just an inference problem with very limited data, it’s an inference about inference problem with very limited data. You can’t reduce it to some computationally simpler problem of updating a posterior distribution, you can only gather data, induce improved heuristics, and hope to God you’re not in a local maximum.
I think rationality in the LW sense can to be said to be about heuristics, limited-information and an online-learning utility-maximization inference problem.
If you say that on LW. People are generally going to agree and maybe add a few qualifiers. If someone on Huffington post would say: “We should think about politics as being heuristics, limited-information and an online-learning utility-maximization inference problem.”, the audience wouldn’t know what you are talking about.
the strategy we’re trying to play while knowing that other agents view us as belonging to all those listed categories of Bad People
That assumes that the best way to act is in a way where other agents get a sense that you are playing or what you are playing.
There no reason to believe that a problem being visible makes it important. Under the Obama adminiratrion the EPA managed to raise standards on mercury pollution by being able to calculate that the IQ points of American children are worth more than the money it costs to reduce polution.
The issue didn’t become major headlines because nobody really cared about making it a controversial issue.
Nobody had the stomach to hold a speech about how the EPA should value the IQ of American kids less.
At the same time the EPA didn’t get anything done on the topic of global warming that was in the news.
Naomi Klein description about how white men in Africa kept economic equality when the gave blacks “equal rights” is a good example of how knowledge allows acting in a way that makes it irrelevant that the whites where seen as Bad People.
Sorry, I didn’t mean that “LW is outside standard left vs right.” I meant that “post-political” politics is categorically impossible when you can’t exhaustively evaluate Solomonoff Induction. You cannot reduce an entire politics to “I rationally evaluated the evidence and updated my hypotheses”, because the relevant set of propositions doesn’t fit the necessary axioms. Instead, I think we have to address politics as a heuristic, limited-information, online-learning utility-maximization inference problem, one that also includes the constraint of trying to make sure malign, naively selfish, ignorant, and idiotic agents can’t mess up the strategy we’re trying to play while knowing that other agents view us as belonging to all those listed categories of Bad People.
So it’s not just an inference problem with very limited data, it’s an inference about inference problem with very limited data. You can’t reduce it to some computationally simpler problem of updating a posterior distribution, you can only gather data, induce improved heuristics, and hope to God you’re not in a local maximum.
I think rationality in the LW sense can to be said to be about heuristics, limited-information and an online-learning utility-maximization inference problem.
If you say that on LW. People are generally going to agree and maybe add a few qualifiers. If someone on Huffington post would say: “We should think about politics as being heuristics, limited-information and an online-learning utility-maximization inference problem.”, the audience wouldn’t know what you are talking about.
That assumes that the best way to act is in a way where other agents get a sense that you are playing or what you are playing.
There no reason to believe that a problem being visible makes it important. Under the Obama adminiratrion the EPA managed to raise standards on mercury pollution by being able to calculate that the IQ points of American children are worth more than the money it costs to reduce polution. The issue didn’t become major headlines because nobody really cared about making it a controversial issue. Nobody had the stomach to hold a speech about how the EPA should value the IQ of American kids less.
At the same time the EPA didn’t get anything done on the topic of global warming that was in the news.
Naomi Klein description about how white men in Africa kept economic equality when the gave blacks “equal rights” is a good example of how knowledge allows acting in a way that makes it irrelevant that the whites where seen as Bad People.
Playing 1 or 2 levels higher can do a lot.