The Mechanics of ``Tacit Knowing''

The term ``tacit knowing'' has been much used to denote certain aspects of human knowledge that are only partly verbalizable. The claim which it presupposes, that ``we know more than we can tell,'' has sometimes been taken to imply that mechanistic approaches to the understanding or simulation of human thought are in principle foredoomed to failure. This correspondence argues that a general-purpose goal-pursuing information system, equipped to represent the structure of its world in terms of conditional constraints on action and the planning of action, would automatically show signs of ``tacit knowledge.'' Examples are cited ofmechanistic principles (including the use of interactive elements that combine digital and analog information processing) which would lend themselves to the operation of such a system and would be able to handle information unverbalizable by the system itself. It is suggested that programs seeking to simulate only introspectively verbalizable thinking are methodologically inadequate to represent the full range of human intelligence.