In conversations, I am often the one guilty of ramping up the abstraction level. This makes people uncomfortable for any number of reasons, and a common retort is at the assumptions I need to make to do this. The discomfort is justified - abstraction engaged for no reason is a recipe for rapid deadlock. However, I think the assumptions are often rejected wthout sufficient consideration. I think things would get a lot further if people tried to understand before disagreeing.

What does it mean to understand an assumption? For me, it is to understand what the completeness in the direction of the assumption consists of. A simple example: I can assume that "my mailbox is red". Well, the complementary statement to that would be "my mailbox is not red". It's a basic exercise in logic - the two together make the sum of possbilities, but in assuming, I am choosing just one.

There are always implicit assumptions that do not get stated. One assumes that the audience knows your language - that the words used have the same relationship to each other and to the emprical world in both speaker and listener. It is when you come up against the borders of shared language that things become icky. To be ready for the possiblity of being constrained by this border, one should know how to resolve ambiguity when necessary - by searching for terms which are truly common, and redefining as appropriate.

I enjoy those excursions, but I have to be careful. There are goals for speech, and algorithms for trial and error, and I have to be wary of timing out.

Link

 Adjunct professor Arnold Kling offered a terser précis of the GMU way. “My simple way of describing it is that at Chicago they say, ‘Markets work; let’s use markets.’ At Harvard and MIT they say, ‘Markets fail; let’s use government.’ And at George Mason, we say, ‘Markets fail; let’s use markets.’” This seeming paradox means, that GMU sees plenty of deviations from the “perfect neoclassical paradigm,” which requires “perfect information, perfect competition,” but that unlike Harvard or MIT, they do not automatically “ring a bell and say, ‘We need more government.’ Markets come up with solutions to problems of information.”

After a few days of dreaming that the Kelly Criterion got rid of axiomatic risk adversity, it's back. Tried some simple, small cases with Eric Zimanyi today, and couldn't convert optimality of the expectation value into anything resembling risk adversity for individual time slices. ARGH.