This may not be the best place for this question, but it’s something I’ve been wondering for a while: how does causal decision theory fail us humans in the real world, here and now?
This may not be the best place for this question, but it’s something I’ve been wondering for a while: how does causal decision theory fail us humans in the real world, here and now?
Us humans almost never use Causal Decision Theory in the real world, here and now. As such it fails us very little. What humans actually tend to use is about as similar to TDT as it is to CDT (ie. actual practice diverges from each of those ideals in different ways and couldn’t be said to be doing either.)
All right, but how would CDT fail us, if we used it perfectly?
If we used it perfectly it would fail us very little. The ‘used perfectly’ part would prompt us to create and optimize institutions to allow the limitations of CDT to be worked around. It would result in a slightly less efficient system with some extra overheads and some wasted opportunities but it would still be rather good. Specifically it would require more structures in place for creating and enforcing precomittments and certain kind of cooperation would be unavailable.
CDT does, however, tell you to precommit to cooperate in a prisoner’s dilemma against someone who also precommits to cooperate with you, if this is an option.
This may not be the best place for this question, but it’s something I’ve been wondering for a while: how does causal decision theory fail us humans in the real world, here and now?
Us humans almost never use Causal Decision Theory in the real world, here and now. As such it fails us very little. What humans actually tend to use is about as similar to TDT as it is to CDT (ie. actual practice diverges from each of those ideals in different ways and couldn’t be said to be doing either.)
All right, but how would CDT fail us, if we used it perfectly?
If we used it perfectly it would fail us very little. The ‘used perfectly’ part would prompt us to create and optimize institutions to allow the limitations of CDT to be worked around. It would result in a slightly less efficient system with some extra overheads and some wasted opportunities but it would still be rather good. Specifically it would require more structures in place for creating and enforcing precomittments and certain kind of cooperation would be unavailable.
Ah, that’s right. CDT tells you to defect in a prisoner’s dilemma against someone identical to you; TDT tells you to cooperate. So TDT wins here.
CDT does, however, tell you to precommit to cooperate in a prisoner’s dilemma against someone who also precommits to cooperate with you, if this is an option.