Reversibility Assessment¶

Why It Matters¶

Risk analysis often treats all errors as equivalent. But reversible errors are fundamentally different from irreversible ones. You can afford to be wrong about reversible decisions; you can't afford to be wrong about irreversible ones. This tool corrects the systematic error of applying the same decision threshold to decisions with different reversibility profiles.

The Key Move¶

For any contemplated action, explicitly assess: "If this turns out to be wrong, can I return to the current state?" If reversible, act on moderate confidence. If irreversible, demand high confidence or create reversibility (staged implementation, backup plans). The key is making reversibility an explicit input to the confidence threshold you require.

Classic Application¶

Intubation decisions. Intubating a patient (inserting a breathing tube) is high-risk but reversible - if you intubated unnecessarily, you can remove the tube. NOT intubating a patient who needs it is often irreversible - they stop breathing and die. The asymmetric reversibility means the threshold for intubation is lower than absolute certainty. Better to intubate and extubate if wrong than to wait for certainty and have the patient die.

Surprising Application¶

Database migrations and schema changes. Deploying a new feature flag is highly reversible - flip the flag off if problems emerge. Dropping a database column is irreversible - once data is deleted, it's gone. Reversibility determines strategy: feature flags can deploy at 80% confidence, but dropping columns requires 99% confidence plus extensive backups. Same product, different decisions based on reversibility.

Failure Modes¶

Treating theoretically reversible actions as practically reversible. Reputation damage, broken trust, or organizational momentum can make nominally reversible decisions effectively permanent. The tool fails when reversal costs are high even though reversal is possible - surgery can be reversed with more surgery, but the cumulative harm may exceed the original problem. Also fails when it's used to justify reckless action: "we can always reverse it" becomes an excuse for inadequate analysis rather than a reason for appropriate confidence thresholds.

Go Deeper¶

Gawande, A. (2002). Complications: A Surgeon's Notes on an Imperfect Science, particularly essays on error and reversibility; Reason, J. (1997). Managing the Risks of Organizational Accidents, on recovery paths and reversibility in complex systems.