You may let a patient die by doing nothing rather than risk killing them by doing something, even when the numbers favor action. Your conscience is not only tracking harm. It is tracking your fingerprints: who can be identified as the cause, not just who bore the cost.
What shifts is who gets recorded as the author of the harm. We treat a harm we caused by moving as categorically worse than an identical harm we allowed by staying still. The body count can be equal or the inaction can cause more harm and the mind still ranks the actor as the greater wrongdoer.
This is not a fringe quirk. Spranca, Minsk and Baron documented it in 1991 across vaccine, inheritance and competition scenarios. The asymmetry survives even when subjects agree the outcomes are identical. A 2022 preregistered meta-analysis by Yeung, Yay and Feldman pooled 49 effects across 21 samples and found a significant overall omission-commission asymmetry, though the effect was stronger for moral judgment and blame than for decisions.
How it works
The driver is a causal-attribution shortcut: when you act, you become the direct cause of the result. When you refrain, the world's baseline gets the blame and you feel merely present. In laboratory scenarios, people tend to treat inaction as less causally responsible even when the consequences are equally bad. Spranca, Minsk and Baron (1991) framed omission bias as an asymmetry in causal attribution, not as an expression of loss aversion. Baron and Ritov (2004) went further: the asymmetry rests largely on the distinction between direct and indirect causation rather than on action versus inaction as such.
A reconstruction, not a tested process model: • you classify one option as 'doing' and the other as 'not doing' • you treat only the 'doing' branch as something you authored • you assign greater responsibility and blame to the branch where your action directly produces the harm
The result is systematic bias toward inaction whenever both paths carry cost. Spranca, Minsk and Baron (1991) established the core finding. Ritov and Baron (1990) found that people were reluctant to vaccinate even when the vaccine's risk was far lower than the risk from the disease - a result the authors interpreted as reflecting the omission-commission asymmetry.
Also shows up in
- A regulator who approves a drug owns every side effect directly. A regulator who delays approval faces a different distribution of visibility: the patients who die waiting rarely appear in any file with the regulator's signature on them. This creates an incentive structure in which cautious inaction can be the institutionally rational move even when it causes serious harm. Whether regulatory delay kills more people than premature approval is contested and this post does not settle it - the narrower claim is that blame asymmetry pulls the incentives in this direction.
- Why knowing does not help
- Knowing about the bias does not dissolve it. The asymmetry is baked into how blame is socially assigned so acting against it exposes you to real punishment. You are not only fighting your own instinct - you are fighting everyone else's. Visible action can attract blame that invisible inaction does not. The incentive structure rewards the omission even when the ethics condemn it.
Countermeasure
The Symmetric Ledger is an author-constructed heuristic, not a validated debiasing intervention. It targets the causal-attribution asymmetry described above: if the bias operates by letting inaction escape authorship, a protocol that forces authorship onto both branches is the obvious counter. In high-stakes medical or institutional contexts, treat it as a structured prompt for deliberate reflection rather than a proven tool.
Threshold condition: apply this when the two options can be meaningfully compared on expected outcomes and the intervention risk is real but not obviously dominant. Do not apply it when natural recovery is likely and intervention carries real iatrogenic risk or in legal or contractual contexts where the act/omission distinction carries real normative weight rather than being a bias artifact. There the asymmetry may be justified, not distorted.
When the threshold is met, run it like this: write both options with their full expected outcomes and avoid the words 'act' and 'do nothing'. Both branches are decisions. Both have authors. You are the author of both.
Failure to act is not the absence of a decision. It is a decision you have hidden from yourself so you never have to sign it.
Where it hurts most
- Medicine: physicians under-treat conditions where intervention risk is visible and untreated decline is diffuse. This is what the leading account predicts, not a causal finding from the studies cited here.
- Content platforms: leaving harmful misinformation up draws less censure than removing accurate content. The leading account predicts systematic under-moderation; whether platforms actually under-moderate on net is contested.
The uncomfortable version
The uncomfortable version: reluctance to act can function as self-protection wearing the costume of caution. That is not the motivational account already set aside. The motivational story says inaction is chosen because it is easy; this is different: inaction can get chosen because it is blameless. You are not weighing the patient's welfare against the intervention's risk. You are weighing the intervention's risk against your own exposure to blame. The victim of your inaction is invisible, no signature, no lawsuit, no name in the file, which is exactly why you can let them go.
There is experimental support for exactly this. In one study, DeScioli, Christner and Kurzban (2011) ran an economic game in which one person could take money from another either by acting or by staying still. Taking-by-omission rose when a third party was watching and able to punish. Their reading is that the omission effect is better described as a strategy than as a bias.
This pattern is not uniform. Baron and Ritov (2004) found meaningful individual differences in who exhibits omission bias and how strongly so the pull toward blameless inaction is stronger in some decision-makers and some domains than others.
The second-order cost most people miss: omission bias scales into institutions and hardens into policy. When every individual actor prefers the blameless inaction, the organization systematically under-intervenes on exactly the problems where waiting is fatal. The bias is not just yours. At institutional scale, the same asymmetry can become embedded in organizational decision processes.
When it doesn't apply
The preference for inaction can be rational when intervention carries genuine risk and natural recovery is likely. Most self-limiting illnesses resolve without treatment and well-meant intervention often causes iatrogenic harm. Here 'first, do no harm' is not a bias, it is correct decision theory. The dual framing: the preference becomes suspect when the action/omission distinction is doing more work than the expected outcomes. It is wisdom when the intervention's downside is real and you have concrete reason to believe recovery without it is the more probable outcome. A condition with a documented natural resolution rate well above the intervention's complication rate, say or a domain where watchful waiting is the evidence-based standard precisely because most cases self-correct. The error is treating the asymmetry as a universal rule rather than testing it against actual expected values.
Danger zone
Status Quo Bias + Regret Aversion + Omission Bias - together they push toward inaction feeling safe, correct and blameless all at once. The default is preferred, the imagined regret of acting is amplified and the harm of not acting is morally discounted. These three may share mechanisms rather than operating independently so their combined effect is hard to quantify. The directional pull of all three toward inaction is consistent and the practical result is a strong disposition to do nothing even when inaction causes serious harm.
Also mapped in Cross-Domain Transfer: Medicine & Healthcare →
Stress test
1. What outcome will occur if you do nothing, stated as concretely as the outcome of acting?
2. Who or what gets named as the cause of that outcome - and is it you?
3. Would you judge the same outcome differently if your action had caused it directly?
4. Are you comparing expected outcomes or comparing your exposure to blame?
5. Would you choose differently if both options required your explicit signature?
Testable prediction
If decision-makers are required to state the expected harm of inaction in the same units as the expected harm of action, then the rate at which they choose the higher-expected-harm inaction branch should fall relative to an unprompted control group. This is a hypothesis derived from the causal-attribution mechanism, not an established intervention effect.
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 distorted by omission bias - the tendency to treat harm you cause by acting as worse than equal or greater harm you allow by not acting. MY DECISION: [paste here] Audit: 1. Restate both options as active decisions with authors. Strip the words 'act' and 'do nothing' - what are the two outcomes I am choosing between? (Note: avoid this relabeling in legal or contractual contexts where the act/omission distinction carries real normative weight - see question 6.) 2. In the same units, what is the expected harm of the inaction branch versus the action branch? 3. Who bears the harm of the inaction and why is that harm less visible to me than harm caused by the action would be? 4. Am I weighing the outcomes or weighing my own exposure to blame between the two paths? 5. If I had to sign my name to whatever result followed either choice, which would I actually defend? 6. Before applying any symmetry correction: is this a situation where the action genuinely carries more moral or causal weight than the inaction - for example because of legal duty, contractual obligation or a context where the act/omission distinction is normatively meaningful rather than a bias artifact? If so, the audit above may not apply. Keep this in your stack. Next trap in the chain: Action Bias.
Related concepts
- Status Quo Bias - stacks with omission bias so that not-acting is doubly protected: it escapes the blame of causing harm and it preserves the intrinsically preferred default.
- Action Bias - the mirror-image failure, where people intervene to feel in control. The same person can over-act in one domain and under-act in another depending on where blame lands: whichever error is more visible in a given context is the one they learn to avoid.
- Moral Luck - highlights why being the visible cause of a bad outcome can attract blame even when intent is identical, which makes avoiding causal responsibility psychologically attractive.
- Regret Aversion - anticipated regret of a self-caused harm looms larger than the unfelt regret of permitted harm, tilting the ledger toward inaction before the outcomes are even compared.
Also connected in the mapPrevention Bias·The Just-World Hypothesis
Origin
Sources
- Spranca, M., Minsk, E. & Baron, J. 1991 - Omission and commission in judgment and choice. Journal of Experimental Social Psychology 27(1), 76-105.
- Ritov, I. & Baron, J. 1990 - Reluctance to vaccinate: Omission bias and ambiguity. Journal of Behavioral Decision Making 3(4), 263-277.
- Baron, J. & Ritov, I. 2004 - Omission bias, individual differences, and normality. Organizational Behavior and Human Decision Processes 94(2), 74-85.
- Yeung, S. K., Yay, T. & Feldman, G. 2022 - Action and inaction in moral judgments and decisions: Meta-analysis of omission bias omission-commission asymmetries. Personality and Social Psychology Bulletin 48(10), 1499-1515.
- DeScioli, P., Christner, J. & Kurzban, R. 2011 - The omission strategy. Psychological Science 22(4), 442-446.
Read next
Entry #69 of 631 in The Lexicon · see the full Lexicon