Continuing to invest in something because of what's already been spent, even when quitting would be the smarter move.
Imagine you buy a big tub of ice cream, take a few bites, and realize you don't like the flavor at all. But instead of just throwing it away, you force yourself to eat the whole thing because you already paid for it — even though every spoonful makes you feel worse. That's the Sunk Cost Fallacy: doing something you don't want to do just because you already spent something on it, even when quitting would make you happier right now.
The Sunk Cost Fallacy describes a systematic pattern in which individuals escalate their commitment to a course of action precisely because they have already invested heavily in it, treating irrecoverable past expenditures as though they obligate future behavior. Rather than evaluating decisions purely on their prospective costs and benefits, people mentally 'carry forward' what they have already spent and feel compelled to justify those past outlays through continued investment. This creates a self-reinforcing cycle: each new investment increases the psychological pressure to continue, which in turn triggers more investment, even when objective evidence clearly indicates that stopping would minimize total losses. The bias is remarkably robust, affecting not only naive individuals but also highly trained professionals, sophisticated organizations, and even governments operating with full economic analysis at their disposal.
The same glitch looks different depending on the terrain. Finance, medicine, a relationship, a team — same mechanism, different costume.
Investors hold onto declining stocks or failing positions far too long because selling would 'lock in' the loss, while continuing to hold preserves the psychological fiction that the original investment might still pay off. This manifests as the disposition effect, where investors sell winners too early and ride losers too long, directly eroding portfolio returns.
Physicians may persist with a treatment plan that isn't producing results because significant time, cost, and patient suffering have already been invested in it, rather than switching to an alternative therapy. Patients similarly resist changing doctors or treatment approaches after extensive prior engagement, even when outcomes are poor.
Students continue pursuing majors or degree programs they've lost interest in because of the years and tuition already invested, rather than switching to a field that better fits their evolving goals. Institutions continue funding underperforming programs because of the historical investment in faculty, facilities, and curriculum development.
People remain in unfulfilling or even harmful relationships because of the years, emotional energy, and shared experiences they've invested, treating the relationship's history as a reason to continue rather than evaluating its future potential independently.
Teams continue building on legacy codebases or outdated technology stacks because of the enormous development effort already invested, rather than migrating to superior platforms. Product managers resist killing features that consumed significant development resources but show poor user adoption metrics.
Managers retain underperforming employees far longer than warranted because of the time and resources spent on hiring, onboarding, and training them. Organizations persist with failing strategic initiatives because of the reputational and political costs already invested in championing them.
Governments continue funding failing public projects and military operations because admitting failure would imply that prior expenditures of public money and, in wartime, lives were 'wasted.' Politicians frame continuation as honoring past sacrifice, making it politically toxic to advocate for rational withdrawal.
Formalized by Hal R. Arkes and Catherine Blumer in their 1985 paper 'The Psychology of Sunk Cost' published in Organizational Behavior and Human Decision Processes. Earlier conceptual groundwork was laid by Richard Thaler (1980) on mental accounting, Barry Staw (1976) on escalation of commitment, and Kahneman & Tversky (1979) on prospect theory and loss aversion. The term 'Concorde Fallacy' was introduced by Richard Dawkins and Tamsin Carlisle in a 1976 Animal Behaviour article.
In ancestral environments, persistence with a partially completed goal — finishing a hunt, completing a shelter, or seeing a seasonal food-gathering expedition through — was usually the optimal strategy because abandoning mid-effort genuinely wasted scarce resources with no alternative uses. The 'don't waste' instinct promoted goal persistence that enhanced survival when options were few and switching costs were genuinely high.
In deep reinforcement learning, agents exhibit sunk cost-like behavior by persisting with suboptimal episode trajectories to completion rather than terminating early, 'polluting' replay buffers with uninformative data and wasting compute. More broadly, organizations exhibit sunk cost reasoning toward AI systems themselves — continuing to invest in underperforming ML pipelines, outdated model architectures, or poorly designed training datasets because of the resources already spent building them, rather than starting fresh with better approaches. LLMs trained on human-generated text can also reproduce sunk cost reasoning patterns in their outputs when advising on decisions.
50 cards are free to preview. Buyers unlock the rest of the deck plus the interactive training — Spot-the-Bias Quiz unlimited, Swipe Deck with spaced repetition, My Blindspots, Decision Pre-Flight, the Printable Deck + Cheat Sheets, and the Field Guide e-book. $29.50$59.
