I simply mean that, yes, all predictions require some form of "theory", even if it's solely the unconcious (or programmed in) ontology used to look at, think about, filter the world/data. I.e. any form of inference is subject to the organizing effect of the machine doing the inferring ... premature registration biased by one's own perspective. But it's too strong to assert that all types of inductive inference will be biased or mis-organized by that a priori ontology/perspective.
And making that argument against the induction tools (especially considering the more hybrid inference you get in typical machine learning, where one does a little induction, a little deduction, and a little abduction in order to arrive at a useful solution) could be the "you do it too" fallacy. If all the accuser's reasoning _does_ require the a priori organization, accusing any given set of machine learning methods of doing it too is, effectively, "You do it too!" It's not an adequate defense of doing it. It might be reasonable to assert that induction is the only (or closest to pure) form of bias-free inference available to us. For example, one could brute-force evaluate all the theorems in a simple formal system, then iteratively (automatically) modify the language according to some schema, then brute force evaluate all the formable sentences in the new language. Etc. Take that to its extreme and you get fully automated theory construction (even if the "theories" make no sense to any humans). On 09/09/2016 07:18 PM, Nick Thompson wrote:
Glen wrote: *There's no doubt that any form of inference done by humans is subject to premature registration or even apophenia. But the inverted claim, that _all_ registration is premature (or imaginary) is way too strong, and perhaps a case of tu quoque.* Narcissist that I am, I assume you are punishing me for all the weird language I have inflicted on the list over the last 12 years. I humbly acknowledge the punishment. Now: Could you explain what you meant? (};-)]
-- ☣ glen ============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
