MarxSchmarx
29th June 2007, 08:08
Okay, this is going to be rather technical; I don't know if this is even the right forum for it. I'll also try to be concise. I wouldn't be traumatized if this thread dies, but I'm kinda curious what other people who've dabbled in this stuff think.
There is a vast literature out there that tries to describe fields that deal with epistemology like statistics and artificial learning using the framework of decision making. Frequently the problems are posed using a flagrantly capitalist vocabulary like "managing" uncertainty or "optimizing" validity in inferences. In fact, the formal mathematical and computational tools used to "optimize" the essentially epistemological problems posed by statistics are often the same as those that capitalists use to "optimize" and gain "efficiency" in their engineering commodities or productions systems. A good example that comes to my mind is control theory, which was developed in large part to minimize production costs on things like airplanes, but has been applied to machine learning with some success. Another example to consider might be the blurred line between actuarial work and Bayesian statistics. And actually more than a few (analytic) philosophers have tried to formalize epistemological questions using these formalized tools.
So, here are some questions:
1. How much does this focus on "optimality", "minimax", "managing" etc... in fields like statistics and machine learning derive from capitalism, especially its assumptions about allocating scarce resources, versus what may seem a natural sensibility about how we want to obtain knowledge?
Certainly the frequently invoked analogy between the goals of a capitalist and the goals of a person trying to draw inferences smack of capitalist presuppositions, but are they inevitably linked? Or are they largely didactic tools?
2. IF there is a capitalist undercurrent to how these fields are formalized, what could/should/would a radical/non-capitalist approach look like?
3. If, on the other hand, the capitalist phraseology and analogies are more didactic than serious, how could/should/would we use a new, "ideologically correct" jargon to frame the questions of formal epistemology?
There is a vast literature out there that tries to describe fields that deal with epistemology like statistics and artificial learning using the framework of decision making. Frequently the problems are posed using a flagrantly capitalist vocabulary like "managing" uncertainty or "optimizing" validity in inferences. In fact, the formal mathematical and computational tools used to "optimize" the essentially epistemological problems posed by statistics are often the same as those that capitalists use to "optimize" and gain "efficiency" in their engineering commodities or productions systems. A good example that comes to my mind is control theory, which was developed in large part to minimize production costs on things like airplanes, but has been applied to machine learning with some success. Another example to consider might be the blurred line between actuarial work and Bayesian statistics. And actually more than a few (analytic) philosophers have tried to formalize epistemological questions using these formalized tools.
So, here are some questions:
1. How much does this focus on "optimality", "minimax", "managing" etc... in fields like statistics and machine learning derive from capitalism, especially its assumptions about allocating scarce resources, versus what may seem a natural sensibility about how we want to obtain knowledge?
Certainly the frequently invoked analogy between the goals of a capitalist and the goals of a person trying to draw inferences smack of capitalist presuppositions, but are they inevitably linked? Or are they largely didactic tools?
2. IF there is a capitalist undercurrent to how these fields are formalized, what could/should/would a radical/non-capitalist approach look like?
3. If, on the other hand, the capitalist phraseology and analogies are more didactic than serious, how could/should/would we use a new, "ideologically correct" jargon to frame the questions of formal epistemology?