You did not stick with the safe option because it was better. You stuck with it because you could already picture the specific face you would make if the other one failed.
That misses the mechanism. Regret aversion is not about the size of the risk. It is about anticipated regret: the pain you expect to feel if a particular choice goes wrong, weighed before you make it. This post's lens is vividness, one factor that amplifies that anticipated pain. A career switch and a promotion may carry similar downside odds but only one comes with an easy-to-imagine future self saying 'I should never have left.' The easier that scene is to picture, the more it weighs, often out of proportion to its actual probability.
Vividness is not the only route. Norm Theory (Kahneman and Miller, 1986) proposes that abnormal or exceptional events make their alternatives more available and so draw stronger emotional reactions. It describes reactions after the outcome. Carrying it over to regret anticipated before a decision is this post's inference. Put plainly: vividness is how concretely you can feel the failure, counterfactual ease is how easily you can mentally undo the choice. The two usually travel together, since acting against the default is both easier to picture failing and easier to undo in hindsight but they are separable. Neither makes inaction regret-free. What they do is make your responsibility and the forgone alternative more salient when you act, especially when the action is unusual or self-authored.
How it works
Zeelenberg (1999) showed that anticipated regret depends on expected feedback: people anticipate more regret when they expect to learn how the option they rejected would have turned out. The driver is prospective emotion: before deciding, you simulate how bad you will feel under each outcome.
The process is asymmetric. When an action makes your responsibility, the forgone alternative or future feedback especially salient, anticipated regret can be sharper than for a comparable omission. That produces a tilt toward whichever choice minimises imaginable regret rather than the one that maximises expected value. The tilt is most likely to dominate when the action branch offers a single, vivid, easily narrated failure and the stakes feel personally attributable. Real expected-value differences, strong prior preferences or social pressure can swamp it when those conditions are absent. Laboratory studies show that anticipated regret can shift choices, though effect sizes vary by paradigm and the field evidence is thinner. Anticipated regret is a robust contributor to choice, not a universal dominant driver.
Also shows up in
- This fires hardest where outcomes are salient and easy to narrate. Medical decisions show a closely related action-inaction pattern. In hypothetical vaccination scenarios, Ritov and Baron (1990) found people reluctant to vaccinate a child when the vaccine itself could cause death, even when that risk was far smaller than the risk from the disease. They called it omission bias. Their later work (Ritov and Baron, 1995), also with hypothetical scenarios, found the bias toward not vaccinating grew when potential regret was present for example when the decision maker would learn how things turned out. How much of omission bias is regret and how much is moral evaluation is not settled.
- Knowing about it does not dissolve it, because the regret simulation can run before the analysis does. You may feel the vivid branch first, then build reasons that make the safer choice easier to defend. Organisations amplify this: when a decision can be blamed on one person, that person tends to route toward the option that is defensible in hindsight rather than the one with the best expected result. Requesting more data is the corporate ritual for outsourcing the regret.
Countermeasure
Start with a written expected-value pass before any commitment. List the plausible outcomes of each option, assign rough probabilities, then compare across the whole set. Where credible base rates or a reference class exist, set the probabilities from them before you describe any outcome in detail so the vivid scene cannot anchor the numbers.
Regret lives on one branch of the future; the correct decision is made across all of them.
Then add a rule for the recurring pattern, the request for more data. Ask three questions in order.
1. Could the specific data requested change the decision? If not, the request is regret management wearing an analytical costume. Decide on what you have.
2. If it could, is the decision reversible? A wrong reversible call can be corrected at reasonable cost so set a low bar: act unless the data is quick and cheap to get.
3. If the decision is irreversible and the data could change it, gather the data. Serious safety, health, legal or financial calls that cannot be undone belong here and caution in those cases is not regret aversion in disguise.
Set your own stopping rule in advance. For example: if two consecutive requests fail the first question, decide on what you have. Two is a working rule, not a tested threshold.
Where it hurts most
- Product and engineering: a team might delay a reversible launch to request data that could not change the call, while the cost of waiting grows.
- Medical decisions: people judge a harmful action worse than an equally harmful omission (Ritov and Baron, 1990) and potential regret can strengthen that pull (Ritov and Baron, 1995).
- Personal finance and career: an investor might hold an underperforming default or an employee stay in a stagnant role, because switching creates an 'I should not have' moment that staying does not.
The uncomfortable version
The choice that feels safe from regret is often the one that quietly compounds the largest cost. Inaction often leaves no equally vivid failure scene to replay. In the short run, you are less likely to lie awake rehearsing the promotion you did not chase, because there is no sharp authored moment to pin the regret on.
Over longer horizons, Gilovich and Medvec (1995) found the pattern shifts: people come to regret inactions more than actions. A 2023 replication (Richardson and Gilovich) reproduced the interaction between action versus inaction and time perspective, though the precise pattern differed from the original. So inaction is not regret-free, it is regret-deferred. This post's reading is that the bias bites hardest when the long-run regret is furthest away and least imaginable, because deferral lets inaction slip past the decision you are making right now.
Bold actions generate memorable, replayable regret. Passive drift generates diffuse, unattributed loss that surfaces later, often too late to correct. Your regret ledger overcounts action and undercounts inaction.
Over a career, this shapes which mistakes you learn from. You correct for the visible failures and stay blind to the far more expensive quiet ones.
When it doesn't apply
Anticipated regret is a real planning instrument when it models consequences accurately. A pre-mortem that rehearses future regret uses this response well: you simulate failure in advance to strengthen commitment and catch weak spots. The pattern serves you when imagined regret is proportional to the actual outcome distribution. It fails you when a single vivid branch crowds out the rest. The test is whether you have calibrated imagined regret to probability or whether vividness has hijacked it.
Danger zone
Status Quo Bias + Omission Bias + Regret Aversion - the default is pre-defended, inaction is pre-forgiven and the consequences of action are pre-rehearsed so the threshold for intervention rises.
Each bias can operate alone. Together they can reinforce one another in a way that makes intervention harder than any one of them would. The signature: the cost of staying may never become as vivid as the cost of acting so the bar for acting keeps rising while the situation quietly worsens.
Stress test
1. Think of a recurring approval you own. When you last asked for more information before deciding, would that information have changed which option you picked or did it just delay the moment of commitment?
2. In your last major reversible choice, did you write down the range of outcomes for each option or did you picture one bad result and stop there?
3. Name a slow decline you are currently tolerating in your work or finances. What imagined action-regret is keeping you from intervening?
4. When you chose the option that others had chosen, was it because it scored higher on expected value or because it would be easier to defend if it failed?
Testable prediction
If a team must state, before requesting data, which decision that data could change, it will make fewer delay-only requests and reach reversible decisions faster. If the expected-value pass works by reducing vividness-driven overweighting, its effect should be largest when one option has a particularly vivid downside but the higher expected value.
Run this audit with your AI
Paste your decision in, and let your assistant audit it through this bias.
Use this to audit a decision you suspect is driven by regret aversion rather than expected value. MY DECISION: [paste here] Audit: 1. Which single vivid failure am I picturing? Using base rates or a reference class, not the image itself, roughly how likely is it? A rough outside anchor beats a confident number drawn from the scene. 2. If I list the plausible outcomes of each option with rough probabilities, which choice has the better expected result across the whole set? 3. Am I choosing this option because it scores highest or because it would be easiest to defend in hindsight if it went wrong? 4. If I am requesting more data or delaying, what specific piece of information would change my choice and does it actually exist? 5. What is the compounding cost of inaction here and why does it not generate an imaginable regret scene the way action does? 6. Am I choosing the default or the crowd's path and would I still choose it if I had to choose from scratch with no default present? Keep this in your stack. Next in this series: Omission Bias, why harmful inaction is judged more leniently than equally harmful action.
Related concepts
- Status Quo Bias - it supplies the default that regret aversion then defends, since the current state needs no justification while any change does.
- Omission Bias - it grades harmful inaction more leniently than equally harmful action. Related but separable; the next post covers where the boundary sits.
- Loss Aversion - it can make the downside of an action loom larger than an equivalent upside, giving anticipated regret more negative material to work with.
Also connected in the mapHyperbolic Discounting·Norm Theory·The Default Effect·Non-Adaptive Choice Switching (Hot Stove Effect)·Gold Medalist Syndrome·The Monty Hall Problem·The Yes-Damn Effect·Regret Minimization
Origin
Sources
- Zeelenberg, M. 1999 - Anticipated regret, expected feedback and behavioral decision making. Journal of Behavioral Decision Making 12(2), 93-106.
- Ritov, I. and Baron, J. 1990 - Reluctance to vaccinate: Omission bias and ambiguity. Journal of Behavioral Decision Making 3(4), 263-277.
- Gilovich, T. and Medvec, V. H. 1995 - The experience of regret: What, when and why. Psychological Review 102(2), 379-395.
- Kahneman, D. and Miller, D. T. 1986 - Norm theory: Comparing reality to its alternatives. Psychological Review 93(2), 136-153.
- Ritov, I. and Baron, J. 1995 - Outcome knowledge, regret and omission bias. Organizational Behavior and Human Decision Processes 64(2), 119-127.
- Richardson, J. and Gilovich, T. 2023 - A very public replication of the temporal pattern to people's regrets. Royal Society Open Science 10(6), 221574.
Read next
Entry #80 of 631 in The Lexicon · see the full Lexicon