Towards Model-Based Model of Cognition

By Karlis Podnieks

In data mining, people are working ... with a variety of computer models that are extremely useful in practice, but hardly resemble their target objects. "All models are wrong, but some are useful." - as put by the statistician George E. P. Box...

This wide experience with "theory-less, wrong and useful" models is provoking three radical ideas. None of them is completely new, the novelty might be the composition proposed below: let's consider philosophy of all kinds of cognition, and try a completely model-based rewriting of it.

The first idea: internally, there is no difference between "true" models and "wrong" models... This corresponds well to "The Dappled World" picture (Nancy Cartwright)...

And, if we wish to answer non-trivial questions like as "Is it true that quarks really exist?", then our philosophy of cognition shouldn't introduce the notion of "truth" too early, as something primary, and therefore - mystical. Truth is emerging later - on top of cognition...

In terms of theories, a similar idea was proposed as the “Deepening Maxim” by Paul Thagard ... And by Jeffrey Alan Barrett ... .

The second idea: models are the ultimate results of all (scientific, non-scientific, and anti-scientific) kinds of cognition...

The third idea: to cover all kinds of cognition, the notion of model should be defined as broadly as possible... Among philosophers, this broadest possible notion of model was stated by Paul Teller... But, among computer scientists, it can be traced back to Marvin Minsky...

... My general impression: despite many brilliant insights, generated by philosophers for many years, the field (philosophy of cognition) remains unordered for too long a time...

Full text:
K. Podnieks. Towards Model-Based Model of Cognition. The Reasoner, Vol. 3, N 6, June 2009, pp. 5-6 (available online).