Polanyi’s Hidden Rulebook

Watch a child learn to ride a bicycle, and you are witnessing the limits of human speech. You can explain how to pedal, how to pull the handlebars, and how to maintain forward momentum, but no amount of instruction will capture the exact micro-adjustments happening in their muscles. We simply know more than we can tell.

This gap between action and articulation is known as tacit knowledge. It is practical, action-oriented know-how born from experience rather than explicit rules. You cannot download it from a textbook. To acquire it, a novice has to undergo what philosopher Michael Polanyi called indwelling—extended, hands-on practice within a specific form until the skill becomes second nature.

For a long time, this inarticulate dimension of mastery set a hard ceiling on automation. Because machines run on explicit instructions, tasks that required human intuition or physical knack remained stubbornly out of reach. This barrier, known as Polanyi’s paradox, suggested that certain human capabilities could never be coded away.

Except, modern technology has tried to write around the rulebook. Technological optimists argue that machine learning has effectively bypassed Polanyi’s paradox by learning from vast oceans of data and context instead of relying on programmer-defined algorithms.

Yet, a machine that mimics the shape of tacit knowledge is not the same thing as one that understands it. Even as algorithms absorb the patterns of human expertise, we are left wrestling with the consequences. When systems operate on learned data rather than transparent rules, we lose our ability to fully interpret their decisions, introducing persistent anxieties over safety and hidden bias.

Ultimately, the paradox has not vanished. It has simply shifted shape. We have built machines that can perform what we cannot explain, leaving us to wonder how much we truly understand about the systems we rely on.