Doing nothing feels safer than acting, and it is not the same as being safer

The asymmetry, demonstrated plainly

The standard experiments present two versions of a case with identical outcomes and different routes to them. In one, somebody does something that causes a harm. In the other, somebody declines to do something and the same harm follows. Participants consistently judge the action worse than the omission, rate the actor as more responsible, and report that they would feel worse having acted.

This shows up across a range of vignettes and populations, and the basic pattern isn’t seriously disputed. What is disputed is what it means, and that dispute is more interesting than the effect. Some researchers describe it as a bias — a departure from the consequences, which are the same either way. Others argue it tracks something real about causation and responsibility, since an action inserts you into a chain of events that would otherwise have run without you.

Both readings are partly right

The moral reading has a genuine case. Acting establishes a clearer causal role, it usually involves more information about what you were doing, and legal systems nearly everywhere distinguish the two for reasons that are not merely sentimental. Treating an action and an omission as equivalent in every respect would flatten distinctions most people are unwilling to give up.

What is harder to defend is the size of the gap and its indifference to stakes. The same asymmetry appears when the decision is entirely private, when nobody will ever know, and when the outcomes differ substantially rather than being identical. At that point it’s doing something other than tracking responsibility: it’s reducing the expected regret of the decider at some cost to the expected outcome.

Where the default absorbs the decision

The practical consequence is that whatever counts as inaction gains an advantage that has nothing to do with its merits. Keeping the current supplier, staying in the current role, leaving the arrangement as it is — each is a choice, and each is experienced as the absence of one. The status quo does not have to win the argument, because it’s never entered into the argument.

This is why the framing of which option counts as doing nothing is worth attending to closely. Whoever gets to define the default has already won a significant advantage, and the definition is frequently arbitrary. Renewing a contract and switching are both actions; one of them is merely automatic, and the automation is a design decision somebody made.

The cost is invisible in exactly the way you would expect

Harms caused by acting produce a visible event with an identifiable author. Harms caused by not acting produce nothing at all — no incident, no date, no person to name. This means the two error types are recorded completely differently, and any system that learns from its records will learn to avoid action.

That has real consequences in organisations, where the personal risk of a decision that goes wrong is much higher than the personal risk of a decision never taken. The resulting caution is individually reasonable and collectively expensive, and it is not fixed by encouraging people to be bolder. It is fixed, if at all, by recording the decisions not taken alongside the ones that were, so that the two categories become comparable at all.

Testing whether the asymmetry is doing legitimate work

The check is to restate both options as actions and see whether the preference survives. Not choosing to switch is choosing to remain. Not deciding is deciding to let the current arrangement continue for another quarter. If the preference flips under that restatement, the framing was carrying the decision.

A second check concerns whose interest the asymmetry serves. If the reason to prefer inaction is that you would feel worse having acted, that is a real cost and it belongs in the reckoning — you are the one who has to live afterwards. If the reason is that nobody could blame you, the consideration is about attribution rather than outcome, and it deserves to be named as such rather than allowed to pass as prudence.

Where inaction is genuinely the right call

Frequently, and this is not a manifesto for doing things. Where the situation is poorly understood, where interventions have a history of unintended effects, or where the system will self-correct given time, waiting is a substantive strategy rather than a failure of nerve. Some of the worst outcomes on record came from confident action in a system nobody modelled correctly.

The distinction worth holding is between deciding to wait and failing to decide. The first has a reason, a review date and a condition that would change the answer. The second has none of those and looks identical from outside, which is precisely why it survives so comfortably.