Irrational behavior doesn’t equal random behavior and is typically easier to model than actually rational behavior. Think of a drunk thug, only capable of following very few rules-of-thumb which do not even accomplish his actual aims (‘irrational’).
That’s a good point, maybe instead something “sufficiently removed from what I’d do in any remotely similar situation that my theory of mind breaks down in trying to predict the details of how such behavior would play out.”
Irrational behavior doesn’t equal random behavior and is typically easier to model than actually rational behavior.
Not necessarily. Think about a game-theory scenario, e.g. a negotiation. A fully rational counterparty is relatively easy to model and game. A somewhat insane counterparty is much harder to deal with. That’s why signaling irrationality is a common negotiating tactic.
I’d disagree with “a fully rational counterparty is relatively easy to model and game” on the grounds that you’re basically saying “you can easily win against fully rational counterparties”, which isn’t the accepted usage for “rational” on LW.
Typically I encounter the “irrational-and-predictable” variant more often than the “irrational-and-unpredictable” kind. It’s the actual rational actors that have the oomph to wrap their desires in an enigma, if it serves their purposes (as you say).
Hence the “relatively” :-) But yes, I agree that I should have said “a rational counterparty that doesn’t play games” or, maybe, a “naively rational” counterparty.
Irrational behavior doesn’t equal random behavior and is typically easier to model than actually rational behavior. Think of a drunk thug, only capable of following very few rules-of-thumb which do not even accomplish his actual aims (‘irrational’).
Compare HPMOR!Quirrell and a redneck neighbor.
That’s a good point, maybe instead something “sufficiently removed from what I’d do in any remotely similar situation that my theory of mind breaks down in trying to predict the details of how such behavior would play out.”
Not necessarily. Think about a game-theory scenario, e.g. a negotiation. A fully rational counterparty is relatively easy to model and game. A somewhat insane counterparty is much harder to deal with. That’s why signaling irrationality is a common negotiating tactic.
Hence the ‘typically’.
I’d disagree with “a fully rational counterparty is relatively easy to model and game” on the grounds that you’re basically saying “you can easily win against fully rational counterparties”, which isn’t the accepted usage for “rational” on LW.
Typically I encounter the “irrational-and-predictable” variant more often than the “irrational-and-unpredictable” kind. It’s the actual rational actors that have the oomph to wrap their desires in an enigma, if it serves their purposes (as you say).
Hence the “relatively” :-) But yes, I agree that I should have said “a rational counterparty that doesn’t play games” or, maybe, a “naively rational” counterparty.