The Tacit Dimension

Michael Polanyi argues that all knowledge rests on a foundation we cannot fully state: we recognize a face, ride a bicycle, or diagnose an X-ray by attending to countless subsidiary clues we could never list, let alone teach as a formal rule. This tacit knowing, not explicit propositions, is what lets the parts of an experience add up to a meaningful whole — captured in his famous line that we know more than we can tell.

For anyone thinking about machine intelligence, the claim is a challenge more than a comfort. If expert skill is built from clues we can’t articulate, systems trained only on what can be written down and labeled may be reproducing the shadow of a competence whose substance stays out of reach.