Formal Assessment of Brachman–Levesque Common Human Intelligence

The level of reasoning capacity seen in, for example, Kurt Gödel cannot be rationally called “common,” by any reasonable metric. Gödel, regarded by many to be the greatest human logician, was able to routinely discover solutions to reasoning challenges that, realistically, would perpetually paralyze most human beings, however high their educational level might reach. What then is the nature of common (= everyday) human reasoning (CHR)? Some scholars have proposed substantive answers to this question; for example, Johnson-Laird et al. have proposed what they declare to be an explicitly anti-logic one (according to which common reasoning is based on mental objects that are “iconic possibilities”). Very recently, in the context of an attempt on their part to contribute to the engineering of AIs with common-sense, or—as we prefer to say—common human intelligence (CHI), Brachman and Levesque (B&L) have ventured an extremely interesting logic-based answer to this question. We analyze their answer—CHRB-L—herein, and find that it posits a level of intelligence too low to do justice to intuitive, pre-analytic concepts of CHR, and much too low to do justice to our own (admittedly elevated) CHRB-F. At the heart of matters is the level of expressivity that a given reasoning challenge displays, and correspondingly, the level of expressivity that a given theory of common reasoning enjoys.