Reversible and Irreversible / Decision-Making

Two kinds of door

Some decisions are two-way doors: you walk through, and if the room is wrong, you walk back out, a little time lost. Others are one-way doors: they lock behind you, and there is no returning to the choice you had a moment before. Most of the skill in deciding is telling which door you are standing in front of — and applying a completely different standard to each.

The common mistake is to use one standard for both: agonizing over trivial reversible choices as though they were permanent, and rushing through permanent ones as though they could be undone. Both errors come from not asking the one question that sorts them.

Speed for the reversible

For a two-way door, the cost of being wrong is small — you reverse, having lost only the detour. This means deliberation is usually overpriced. The time and anxiety spent perfecting a reversible decision often exceeds the entire cost of just getting it wrong and correcting. Reversible choices reward speed, experiment, and cheap error: try it, learn from reality rather than from analysis, and undo it if it fails.

This is the convex strategy from the last essay in operational form. A reversible decision is a convex bet — bounded downside (the detour), open upside (it might be great) — and the right policy is to take many of them quickly without much agonizing.

The asymmetry of ruin

The one-way door is a different kind of object, and it breaks the logic of averages in a way worth understanding precisely.

Expected value implicitly assumes you can play many times and collect the average. That assumption dies the moment a single outcome can remove you from the game. If one possible result is ruin — bankruptcy, death, an unrecoverable loss — then you do not get to keep playing until the odds even out, because you are gone. The average over many trials is irrelevant to someone who does not survive the first bad one.

A bet that pays handsomely on average but carries a small chance of ruin is not a good bet you should sometimes lose. It is a countdown. Play it enough times and the absorbing outcome is not unlucky — it is certain.

This is the difference between the average across many parallel gamblers and the trajectory of one gambler across time. They coincide only when there is no ruin. Introduce an absorbing barrier — a loss you cannot come back from — and the time-path of any individual diverges catastrophically from the rosy ensemble average. Avoiding that barrier is not conservatism. It is the precondition for every other strategy to have a chance to work.

The meta-historical echo

This is the same irreversibility that haunted the previous topic, now at the scale of a single life. History’s ratchet meant some changes never reverse; the warning there was that rising capability lets ordinary folly trigger consequences that cannot be undone. A decision under uncertainty faces the identical structure in miniature: most errors are recoverable detours, but a few are absorbing, and those few deserve a caution out of all proportion to their probability.

The practical rule that falls out is simple to state and hard to live: be aggressive with reversible decisions and conservative with irreversible ones. Take the two-way doors fast and often. Approach the one-way doors slowly, and never let a merely-probable gain talk you through one whose downside you could not survive.

Toward the cost of error

Reversibility, then, is really a claim about the cost of being wrong — bounded and recoverable on one side, unbounded and final on the other. That suggests the question underneath all of this was never about outcomes at all, but about the structure of error: not how likely you are to be wrong, but what it costs you when you are, and how that cost is shaped. That is what the next essay takes apart.