Comments on: Looking for that Northwest(ern) passage

By: Peter McB.

Peter McB. — Tue, 23 Aug 2005 17:54:09 +0000

Steve, agreed. One of the hardest things for Computer Scientists to accept — used as we are to designing and buildng systems ourselves — is that agents in multi-agents systems will be diverse, and that the diffusion of beliefs and expectations in such systems generates much of the dynamism which these systems exhibit.

By: steve

steve — Mon, 22 Aug 2005 16:39:35 +0000

The ratex stuff in macro (to the extent one accepts the assumption) works because there is no supernormal profit in equilibrium. In other words, we can all have an accurate model of the economy, base our actions on that model, and collectively make that model true (i.e. find a fixed point in the beliefs/actions correspondence)–but the equilibrium cannot allow anyone to get rich from knowing the true model. So a trend forecasting system that everyone knows about that makes trends come true will not be one that earns a profit.

Personally, I’m skeptical of the assumptions in most ratex models that a homogeneous representative agent is a good way to think about expectations. It seems to me that heterogeneous beliefs are pretty important in understanding the economy.

By: Peter McB.

Peter McB. — Fri, 19 Aug 2005 17:17:04 +0000

As usual, a fascinating post, Grant.

It is not at all obvious to me that widespread awareness by trend-watchers of trends (and even of trends in trend-watching) should make it harder to track trends. After all, the theory of rational expectations in neo-classical economics tackles a very similar problem. This is the theory which attempts to account for the fact that the success or otherwise of a Government’s macro-economic policies can depend upon the expectations of knowledgeable and rational (in the strictly limited sense of economics) participants in the economy about those same policies and their intended effects on the participants.

In other words, the Government might lower our taxes to encourage us to spend more, but if we know that is why the Government is lowering taxes, we may decide not to spend more (or we may over-spend).

I’m working on similar stuff in Computer Science — designing computational systems comprising intelligent software agents aware of the system states (perhaps only imperfectly), using this awareness to predict future states, and also discussing their predictions with one another — intelligent, reflective, anticipatory complex adaptive systems.

By: fouroboros

fouroboros — Fri, 19 Aug 2005 16:41:27 +0000

Wow, I’d subscribe to Steve’s idea, and pay for it!

Very nice post, Mr Popcorn.

(Wait, that popcorn thing was meant as praise, not insult. Very nice post, Mr. McCracken, I mean.)

By: steve

steve — Fri, 19 Aug 2005 16:15:44 +0000

I’d really like to know how the Kellogg marketing folk reacted to your presentation. It can be hard for academics to refocus their attention from the scholarly literature to the outside world. If they were able to do it, it’s praiseworthy.

Your point about how widespread knowledge about a theory of trends would destroy the patterns the theory tries to explain suggests that such theories may not be constructible. Or perhaps that we may find some local or mesocopic rules, but not a grand, overarching Theory that can predict trends before they happen.

Actually, it might be interesting for you to outline your “fantasy” theory of trends: What data would it take as input, what general kinds of procedures would it apply to those data, and what kind of output would it produce? Outlining the input/output ideal “fantasy” might enable you to rule out a lot of avenues and rule in some others.