Boundary Conditions (Polanyi)

Boundary conditions are Polanyi's term for the way lower organizational levels constrain higher ones without fully determining them. In engineering, physics provides boundary conditions: materials must support loads, energy must be conserved, forces must balance. But these constraints do not specify which machine gets designed—they permit vast ranges of possible configurations, and engineering introduces higher-level principles (purpose, efficiency, aesthetics) to select among them. In biology, chemistry provides boundary conditions: molecules must bond stably, reactions must be thermodynamically favorable. But chemistry does not specify which molecular sequences code for viable organisms—natural selection operating over evolutionary time introduces the higher-level organizational principle of fitness. The relationship is neither reduction (higher level determined by lower) nor disconnection (higher level independent of lower) but constraint: the lower level sets limits, the higher level selects within those limits. AI training data provides boundary conditions for outputs: statistical patterns constrain what is probable. But probability does not determine truth, significance, or quality—these require higher-level principles (epistemic standards, evaluative judgment, committed assessment) that humans supply and machines lack.