Have continued mulling over what seemed to me @ first a strong (& irreconcilable?) juxtaposition between Mark's claim that "Prediction is the Gold Standard of Understanding"" and Paul's argument that the goal of science is bringing into existence "previously unconceived alternatives." Alan's talk this week got me visualizing a way to take these two bookends, these "two sides of emergence," of expansion and contraction, and "twist" these parts (with a turn like that which produced Alan's moebius strip) into a whole that has a single "side"--a single coherent argument. It does seem to me now that the sort of expansion of alternatives that is Paul's goal is facilitated by the sort of reduction of possibilities that is Mark's. That gives us a single process: a loop--or moebius strip--in which contraction serves the function of creating previously unconceived alternatives.
Date: Mon, March 31, 2008 8:18 pm
From: "Paul Grobstein"
Subject: unintended consequences, unconceived alternatives, and goats
Thanks Mark, Anne, all for rich fodder to chew on. For the resulting output see http://serendip.brynmawr.edu/exchange/node/2268 .
Date: Wed, 26 Mar 2008 13:12:53 -0400
From: Mark Kuperberg
Subject: Prediction and Understanding
In my effort to "get Paul's goat", I appear to have gotten my own instead, so I would like to clarify my view of the relationship between prediction, understanding, meaning and explanation.
Understanding is a rich concept that includes explanation and prediction and possibly meaning, so the correct saying should be "Understanding is the Gold Standard of Understanding", but that does not get us very far. Understanding requires a model of the phenomena that is more than simply a machine that can grind out successful predictions. The model must contain a causal mechanism which explains why the model is grinding out the predictions that it is grinding out. A model that is just a black box does not rise to a level such that it passes the Understanding Test.
Still, it is too easy to generate explanations for past phenomena. There must be a mechanism to prune down these explanations and that mechanism is prediction. So for example, if you take a series of numbers, there are many, many (really infinitely many) arithmetic functions that can generate that series of numbers. To choose among them, you must try to predict what the next number will be. Only a subset of those arithmetic functions will pass the prediction test. There will still, however, be an infinite number of functions remaining, but as the process goes on and on, some of these functions will get pretty esoteric (which is where Occum's Razor comes in). With real world phenomena, it will become harder and harder to even think up alternative explanations.
So, this is why Paul and I are bookends: Paul is always interested in expanding the universe of possible explanations (generativity) and I am always looking for ways to narrow down the number of candidate explanations.
Date: Thu, 27 Mar 2008 18:40:04 -0400
or, rather than bookends: > a loop between expansion/contraction.