An inside view focuses on internals of the case at hand, while an outside view compares this case to other similar cases. The less you understand about something the harder it is to apply either an inside or an outside view. So the simplest approach would be to just do the best you could with each view and then combine their results in some simple way.
It is not a dichotomy: you should use as much evidence as you can get your hands on. It is wrong to use strictly outside or strictly inside views, if you can do better using information about both (and, really, there is no strong distinction between different levels of description, some of them more "outsidey", some of them more "insidey"). There are cases when a particular representation is misleading misleading for humans, due to the framing bias and its friends, but it doesn't extend to the general case.
Level, that seems to be Eliezer's meta-view. It isn't very useful, though; he argued for it by denying the principle of all induction, as I illustrated in the comments on his post on Surface Analogies.
It is not a dichotomy: you should use as much evidence as you can get your hands on. It is wrong to use strictly outside or strictly inside views, if you can do better using information about both (and, really, there is no strong distinction between different levels of description, some of them more "outsidey", some of them more "insidey"). There are cases when a particular representation is misleading misleading for humans, due to the framing bias and its friends, but it doesn't extend to the general case.
Level, that seems to be Eliezer's meta-view. It isn't very useful, though; he argued for it by denying the principle of all induction, as I illustrated in the comments on his post on Surface Analogies.
Taboo "outside view" and "inside view" please.
So perhaps the only useful outside view is the meta-view that inside views are preferable to outside views in all basic cases?