The sunk cost fallacy: continuing an action because of past investment (time, money, effort), even when the marginal returns no longer justify it.

The correct frame: sunk costs are sunk. The decision should be based entirely on expected future costs and benefits. The past is irrelevant to the forward decision.

The fallacy is adaptive in some contexts (reputation for commitment deters exploitation) and pathological in others (staying in a failing project because you've already spent three years on it). Distinguishing which context you're in requires stepping outside the decision.

Related: [[loss-aversion]], [[long-run-short-run]]