Omission Bias

Omission bias is when people prefer not to act because doing nothing feels safer than making a choice. They avoid taking actions that might cause harm even if acting would likely help.

Omission bias denotes a decision tendency where agents favor inaction over action due to perceived responsibility asymmetries and anticipated outcome attribution. It reflects an evaluative distortion in which commission risks are overweighted relative to omission risks within cognitive valuation processes.

A parent notices their child has a mild fever but hesitates to give them over-the-counter medicine, worrying about side effects from the drug. Even though leaving the fever untreated might be more risky, doing nothing feels safer because if something goes wrong from giving medicine, the parent feels more at fault than if the fever simply worsens on its own.

In a clinical trial safety monitoring committee, interim data reveals a modest but statistically significant increase in adverse events in the control arm. Despite equipoise having been broken, committee members resist halting the trial early—an active commission—fearing reputational liability for "stopping a trial prematurely," even though continuation exposes control-arm participants to preventable harm. This reflects asymmetric responsibility weighting: perceived causal attribution for harm from commission (stopping) outweighs perceived attribution for harm from omission (continuing), distorting expected utility computations against the statistically warranted intervention. The feedback sparsity inherent in sequential monitoring (sparse interim looks, wide confidence intervals) further suppresses belief updating toward the action branch, reinforcing default_preservation dynamics and perpetuating activation_asymmetry between action and inaction pathways.

People feel more blame for actions than for inaction, so they avoid choosing to act. This fear of being blamed causes them to stick with the default option.

Omission bias operates through asymmetric causal attribution and responsibility weighting across action versus inaction nodes, with stronger perceived liability tied to commission than omission. Structural features like default states and feedback sparsity constrain belief updating and bias expected utility computations toward passivity.

Change the default so action is normal and recommended so people act. Make consequences of not acting clear and visible.

Reframe accountability and redesign defaults to normalize proactive choices, calibrating feedback to equalize perceived responsibility. Use explicit risk comparisons and salience cues to rebalance expected utility estimates.

missed beneficial interventions; status quo perpetuation; underreaction to risks

Adversarial actors can weaponize omission bias by deliberately designing choice architectures where harmful defaults persist unchallenged—for example, structuring regulatory or policy frameworks so that inaction protects incumbent interests while burdening challengers with the perceived agency of "making a change." Propagandists can suppress accountability for harmful status quos by consistently framing inaction as morally neutral, ensuring populations tolerate ongoing harm through normalized passivity. In medical, financial, or security contexts, omission bias can be exploited by withholding salient feedback about the costs of inaction, causing decision-makers to systematically underreact to emerging threats or crises.

Redesign choice architectures so that beneficial interventions are the default, explicitly requiring active opt-out rather than opt-in, which rebalances commission-omission asymmetry at the structural level. Provide explicit, salient comparisons of omission risks alongside commission risks in decision briefs, using concrete outcome metrics to counteract feedback sparsity. Train decision-makers in asymmetric responsibility attribution—specifically through scenario exercises where inaction consequences are tracked and made attributable—to recalibrate expected utility computations toward parity.