I once wrote a “pure empiricist” sequential learning program that took this sort of approach to the data in its input stream.
Here is the manual, that will give some idea —
The program integrated a sequential learning module and a propositional reasoning module that I thought of as The Empiricist and The Rationalist, respectively.
The learning module was influenced by ideas from the psychologists Thorndike and Guthrie and the statisticians Fisher and Tukey. The reasoning module made use of ideas about logical graphs from C.S. Peirce. There is a kind of phase transition as we pass from finite state adaptation covered by the learning module to context-free hypothesis generation covered by the reasoning module, but it happens that some aspects of the latter are already anticipated in the former.