Anticipated regret is a real input and a poor master

Regret is part of the decision, not only its aftermath

People don’t merely experience regret; they predict it and choose to avoid it. Given a choice between an option that could turn out badly in a way you would feel responsible for and one that could turn out equally badly through no fault of yours, many people take the second, and will say plainly that the reason is how they would feel afterwards.

There is a substantial body of work on this and the basic observation holds up: anticipated regret influences choices in insurance, in whether to change an answer, in whether to act on incomplete information. Whether it should is a separate question, and the answer isn’t simply no. A decision you will spend two years turning over has a real cost attached, and pretending otherwise isn’t rigour.

The forecast is the unreliable part

What is much less reliable is the prediction itself. Research on how people forecast their own future feelings has found a consistent pattern: the intensity and especially the duration of emotional reactions are overestimated, for both good and bad events. People expect to be more affected, and for longer, than they turn out to be.

The proposed explanation is that when imagining a future state you focus on the event and leave out everything else that will be happening at the same time, and you fail to anticipate how quickly the mind assimilates a change and moves on. The broad finding that predicted reactions are too extreme has reasonable support; the specific mechanisms have been argued about, and some of the neater demonstrations in this area have proved less robust than the general pattern.

Regret about action and about inaction behave differently

A frequently cited distinction is that in the short term people regret what they did, while over longer periods they regret what they failed to do. The evidence is a mixture of surveys and recall studies, which have their own problems — long-term recall of regret is filtered by everything that happened since, and the categories are hard to define consistently.

The pattern is nonetheless worth holding, because it points at an asymmetry in visibility. A regretted action produces a concrete outcome you can examine and eventually explain, and explanation dissolves regret over time. A regretted omission has no outcome at all, only an open imagined alternative, and an imagined alternative can be improved indefinitely because nothing constrains it.

Where anticipated regret helps

It works well as a detector for decisions where you have not been honest with yourself about the stakes. Imagining the outcome vividly, and noticing a strong reaction, frequently surfaces a consideration that the analytical version of the decision had quietly dropped. Used that way, it’s a prompt to look again rather than a verdict.

It is also a reasonable input where the decision is genuinely one-off and irreversible, because in that situation there is no long run for the average to assert itself and how you live with the outcome is a real part of the outcome. The tie-break version — when two options are close, take the one you would defend more comfortably in a year — costs nothing and uses the signal at roughly the weight it deserves.

Where it does damage

It does damage when it drives towards inaction as a general policy, since the harms of inaction never arrive as identifiable events and therefore never generate the anticipated regret that would have balanced the calculation. The forecast is taken over one side of the ledger only, which guarantees a systematic tilt.

It also does damage in repeated decisions. Optimising each instance to minimise how bad you might feel produces a strategy that’s worse over the sequence, because avoiding a small chance of a vivid, blameable loss usually means accepting a large chance of a mild, unattributed one. Over enough repetitions the mild losses dominate, and nobody ever regrets them because nothing about them is memorable.

A workable weighting

Make the decision on its merits first, in whatever terms you would use if nobody had to live with it, and write that down. Then ask what you would regret, and treat the answer as evidence about what you value rather than as an instruction about what to do. When the two conflict, the question is whether the regret is pointing at a consideration you left out or at a preference for not being blamed.

And discount the forecast. Whatever intensity you are imagining is probably too high and the duration almost certainly too long, which is a general finding rather than reassurance about your particular case. That does not make the feeling irrelevant. It makes it a signal with a known bias, which is the only kind of signal any of this ever produces.