eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

How Intelligence Works | Escalation of Commitment — Sunk Costs, Exit Rules and Better Decisions

—

by

eduKate Asia

Did you know that a decision can become harder to reverse precisely because someone has worked so diligently on it? Escalation of commitment happens when people devote additional resources to a course of action after troubling evidence appears, partly because the earlier investment makes withdrawal feel wasteful, embarrassing or disloyal. The solution is not to stop everything at the first setback. It is to judge the next decision on credible future costs, future benefits, live alternatives and a stated rule for reconsideration, rather than letting unrecoverable past spending decide the future.

A student continues a revision method that is not improving retrieval because she has already filled three notebooks. A school retains a platform because staff spent a term configuring it. A company spends more on software that will soon be obsolete because its first budget was large. In each case, the sensible question is not whether the past effort was sincere; it is whether the next hour, dollar and opportunity are likely to do more good here than elsewhere. Good intelligence can honour the learning from past work without demanding that its original plan survive forever.

Jump to a question

Why intelligent people keep backing a failing plan

Escalation is not simply stubbornness. People operate under incomplete information, real deadlines, social expectations and commitments to others. Occasionally another investment genuinely changes the outlook. The task is to separate those situations from a cycle in which every setback becomes a reason to spend still more. We will work through the causal mechanisms, an original numerical decision, practical exit rules, organisational safeguards and a transferred school-life problem.

What counts as a sunk cost and what does not

A sunk cost is a past expenditure that cannot be recovered regardless of what is chosen now. It may be money, time, effort, prestige or relationship capital. Calling something a sunk cost does not say the earlier decision was stupid. Research can be worthwhile even when a project is stopped, because the information it yields may prevent larger future losses. The distinction concerns the decision boundary: when choosing tomorrow’s actions, money already irretrievably spent is not an incremental cash flow of tomorrow’s options.

Not every old investment is unrecoverable. A machine might have resale value; unused subscription months might be refundable; trained employees may transfer skills to a different project; datasets might be reusable. Such salvage and transfer values are relevant because they change what can happen next. The error is either to count an unrecoverable cost as if it could be undone, or to ignore a genuine recovery option merely because the accounting record describes the original expenditure as historical.

Ask three separate questions. Which resources were already consumed and cannot be changed? Which future costs will follow from continuing versus stopping? Which future benefits, salvage values and alternative opportunities can still change? Write the answers in distinct columns. A decision that merges all three into one emotionally charged total is difficult to audit. A decision that names each amount can be challenged, updated and explained to someone who did not make the original choice.

Costs can also be nonfinancial. Continuing a stressful project may crowd out family time, sleep, professional development or other responsibilities. Assigning a precise currency value to these effects may be inappropriate, but pretending they do not exist is also not rational. Record them as bounded qualitative criteria where reliable quantification is unavailable.

The difference between persistence and escalation

Persistence can be intelligent when evidence still supports the objective, the next action has a plausible mechanism for success, and the expected benefits remain worth the future cost. The fact that earlier stages were difficult does not automatically make continuation a mistake. Scientific experiments, rehabilitation programmes, language learning and engineering builds often involve temporary setbacks. A disappointing early result can be part of an ordinary learning curve rather than evidence that the whole objective is unsound.

Escalation becomes the concern when the action is defended mainly by the past: ‘We have already spent too much to stop’, ‘I cannot let the team down after everything we did’, or ‘Another investment must surely recover the previous one’. Those sentences omit the marginal question: what improvement is expected from this specific additional investment, and how will we know? Asserting that success is just around the corner is not evidence unless a credible model explains what will change.

Distinguish a strategy from its objective. The objective might be to help students understand fractions. The current strategy might be a costly digital game that distracts students. Stopping the game need not mean abandoning the educational objective. It may free time to test a simpler explanation, a paper-based exercise or peer tutoring. A team that can separate mission from method avoids turning every method change into an identity crisis.

Write down the condition under which you would continue if you were deciding from scratch today, with the same known information but without personal ownership of the earlier plan. The fresh-start question is not infallible, because historical commitments can create real obligations, but it exposes when sunk costs are quietly carrying more weight than prospective evidence.

Why the mind tries to rescue past effort

People often want their past decisions to appear coherent. Admitting a strategic mistake can threaten competence, reputation or group membership. When the same person who launched a project must judge whether to cancel it, self-evaluation becomes entangled with project evaluation. A later decision may then serve two purposes at once: improving the future outcome and defending the earlier choice. Those purposes can conflict.

Another mechanism is framing. A project that is described as ‘almost recovering our losses’ encourages a different reaction from one described as ‘spend another S$40,000 for an uncertain incremental benefit’. A loss frame can make a risky attempt to break even feel psychologically attractive. The same mathematical probabilities do not automatically receive the same attention in both descriptions. The framing hypothesis is developed by Glen Whyte in the Academy of Management Review analysis of escalating commitment.

Effort can also change subjective value. A student may treasure notes because so many evenings were spent making them, even if the notes are poor for practice. This is understandable; effort can create skill, attachment and pride. The analytical error occurs when that attachment is treated as proof that spending more time on the same method will deliver the needed learning outcome.

Group structures can amplify the pressure. A manager who asks for contrary evidence but rewards only supportive reports is unlikely to hear the real case for stopping. Staff may infer that loyalty means agreement. Over time, weak signals disappear from meetings while increasingly polished progress presentations remain. An effective review system therefore needs permission to discuss bad news and a place to record what would change the conclusion.

What the original research supports and what it cannot settle

The classic paper by Hal Arkes and Catherine Blumer, The Psychology of Sunk Cost (1985), reports evidence that higher past investment can influence willingness to continue an activity. Their study’s theatre-ticket example illustrates how a prior payment can affect later attendance. That is evidence about a behavioural tendency under studied conditions, not a formula that determines what every individual will do. The lesson is to look for the influence of irrecoverable investment when making a fresh choice.

Barry Staw’s 1981 review of escalating commitment discusses cycles in which setbacks can be followed by further commitment intended to recover earlier losses. The review also describes multiple interacting determinants rather than a single universal cause. For organisational design, that matters: a budgeting problem may not respond to the same intervention as a reputation problem or a missing-information problem.

Research has also examined limiting conditions. Thomas Bateman’s 1986 sequential decision experiments caution against treating escalation as inevitable; decision alternatives and future prospects matter. This should make us more precise, not less concerned. A sensible organisation can create conditions in which people update their judgement as new information arrives instead of assuming everybody is incapable of changing course.

These sources do not license invented prevalence figures for today’s workplaces, nor do they establish that any particular failed project was caused by sunk-cost reasoning. Case-specific evidence is needed. In an actual review, the diagnosis should explain the sequence of choices, available alternatives, genuine constraints and decision-makers’ information at the time rather than relying on a fashionable psychological label.

Worked case: the school learning platform at a decision point

Consider a fictional school team piloting a learning platform. It has already spent S$120,000 on development, setup and training. That sum is non-refundable. To complete the next year requires S$50,000 of additional licence, support and adaptation costs. The team forecasts S$42,000 worth of next-year benefits, using an agreed valuation method for staff time saved and useful learning materials produced. We are treating those benefits as hypothetical planning estimates, not observed outcomes.

The incremental financial value of continuing is S$42,000 minus S$50,000 = negative S$8,000. The already spent S$120,000 does not belong in that subtraction because neither continuing nor stopping can change it. A presentation claiming that the team must spend S$50,000 ‘to recover the first S$120,000’ incorrectly turns an unrecoverable expenditure into a reason to incur more. It might feel responsible, but it does not establish a future advantage.

Suppose stopping releases the same staff capacity for a simpler classroom-support project whose estimated net value is positive S$18,000 over the same next-year horizon. The two alternatives are not automatically comparable unless their estimates include consistent staff, timing and risk assumptions. Given our stated hypothetical assumptions, their difference is S$18,000 minus negative S$8,000 = S$26,000 in favour of stopping and redirecting. The lost previous expenditure remains S$120,000 in both scenarios.

Now suppose the platform owner obtains new, credible evidence that the next-year benefit could be S$70,000 rather than S$42,000, with no extra cost beyond the S$50,000. The incremental continuing value would be positive S$20,000. It would then exceed the alternative’s S$18,000 by S$2,000 on these simplified estimates. This is why an exit rule must allow evidence to change a decision. ‘Always quit after a setback’ is no more intelligent than ‘never quit after an investment’.

However, a tiny S$2,000 margin may not be robust to forecast error, delivery delays or obligations to existing students. A responsible team would test sensitivity and ask how certain the revised benefits really are. If the S$70,000 figure came only from an unverified vendor promise, it should not receive the same weight as observed trial results. A decision table is the beginning of deliberation, not an excuse to replace judgement with fragile arithmetic.

A second case: probabilities matter more than a compelling story

Imagine a community programme that requires a further S$30,000 to finish. If successful, the additional programme stage would deliver benefits valued at S$60,000; if unsuccessful, it would deliver benefits valued at S$5,000. The programme team estimates a 40% probability of success and 60% of failure. These probabilities are hypothetical subjective estimates for illustration, not measured base rates. Expected additional benefit is 0.40 times S$60,000 plus 0.60 times S$5,000, which equals S$24,000 plus S$3,000 = S$27,000.

The expected incremental value of completing the stage is therefore S$27,000 minus S$30,000 = negative S$3,000 before considering opportunity cost. If a credible alternative has positive expected value, the case for continuation weakens further. Notice that the successful outcome is attractive. A persuasive success story does not erase the unsuccessful outcome or make it appropriate to use 100% success in the calculation.

Now change the evidence, not the emotion. Suppose a carefully conducted pilot lifts the justified probability of success from 40% to 60%, with the same payoffs. Expected benefit becomes 0.60 times S$60,000 plus 0.40 times S$5,000 = S$36,000 plus S$2,000 = S$38,000. Against the S$30,000 future cost, expected incremental value becomes positive S$8,000. That could warrant continuation depending on the best alternative, distribution of harm and confidence in the pilot.

Real programmes seldom have precisely known probabilities. Document who estimated them, which observations support them and what ranges are plausible. If the decision changes sharply when success probability moves by a few percentage points, say so. This makes uncertainty visible rather than wrapping a personal preference in a precise-looking forecast.

Opportunity cost: the hidden option that changes the verdict

The cost of doing something is not limited to the invoice for that action. A team has finite people, meeting hours, trust, managerial attention and space to experiment. If continuing one initiative prevents a valuable alternative, the foregone alternative is an opportunity cost. It is often missing from progress reports because existing projects have budgets, owners and dashboards while unrealised alternatives do not.

In the platform case, the school’s S$18,000 alternative was relevant because staff could not deliver both. If the team could deliver both using separate resources without conflict, the comparison would be different. That constraint must be stated. Likewise, a student choosing to spend three hours formatting notes gives up some chance to practise questions, sleep or seek feedback. The alternative does not have to be glamorous to be real.

Opportunity costs change over time. A project that was the best available option six months ago may no longer be best when new tools, teachers or evidence appear. Reassessment is not evidence that the earlier team acted carelessly. It may be a rational response to a changing environment. The reverse also holds: a cheaper alternative can disappear, leaving continuation more attractive.

Make at least one serious alternative explicit at each review: stop and redeploy, pause and gather evidence, narrow scope, hand over, or continue under revised conditions. An agenda that includes only ‘continue as planned’ and ‘abandon everything’ is an invitation to exaggerate the cost of stopping.

How to design exit rules before emotions peak

An exit rule is a pre-agreed condition that triggers a review, pause or decision to stop. It need not be a mechanical cutoff. Its function is to make uncomfortable evidence discussable before reputations and schedules become entrenched. Write the rule at project approval, when participants can more easily consider failure scenarios without feeling that the proposal is already under personal attack.

A useful rule names the objective, measurement window, evidence threshold, decision owner and permitted responses. For example: after eight weeks, if fewer than 60 of 100 independently assessed students can explain the target concept, review the teaching method before purchasing another term of software. The number 60 here is a fictional design choice; a real school must choose a defensible target tied to curriculum, cohort and measurement quality. The rule triggers reconsideration rather than automatically punishing teachers or students.

Separate safety stop conditions from performance reviews. A privacy breach, serious safeguarding concern or invalid permission can require immediate containment regardless of a project’s expected financial return. Ordinary underperformance may justify a controlled comparison, smaller pilot or alternative approach. Conflating these categories creates either excessive rigidity or dangerous tolerance of unacceptable conditions.

A good exit rule also specifies what counts as disconfirming evidence. If every disappointing observation can be labelled a temporary obstacle and every favourable observation a breakthrough, the rule is not operational. Reviewers need access to raw measures, denominator definitions and a record of exceptions rather than only a narrative summary.

Decision ledgers: preserve reasons, not just outcomes

A decision ledger records what was known, which alternatives were available, who owned the judgment, and what would trigger reconsideration. It should be short enough to maintain but specific enough to check later. A useful entry includes the date, question, measurable goal, current evidence, uncertain assumptions, estimated future costs and benefits, selected option and reason for rejecting alternatives.

Do not retroactively edit the old rationale to make it appear prescient. Later information belongs in a new entry linked to the earlier one. This makes learning possible: a forecast can be compared with reality, a missing assumption can be noticed, and reviewers can distinguish a sensible decision that had bad luck from a weak decision whose flaws were visible at the time.

Evidence should be linked to its actual source. If a vendor promises that performance will double, record the claim, date and conditions; do not convert it into a proven benefit. If teachers report poor usability, note sample size and the context in which the feedback was collected. A small voluntary survey may reveal problems without estimating their prevalence in the whole school.

The related discussion of outcome bias and fair decision reviews helps prevent a later successful or unsuccessful outcome from rewriting the quality of the original reasoning. A good decision process can sometimes lose, and a poor process can sometimes get lucky.

Independent review without turning it into a blame exercise

Independent review is not a ritual of assigning someone to say no. It adds a different perspective and reduces the chance that the same emotional commitments govern both initial investment and subsequent evaluation. The reviewer needs access to the actual decision criteria, outcomes, constraints and alternatives, not just a polished presentation prepared by the original project owner.

One effective practice is to let the champion explain the case for continuation and then ask a second reviewer to write the strongest legitimate case for stopping. A third person can audit assumptions and units: are costs annual or total, gross or net, paid or committed? Do the benefit estimates cover the same population and time horizon? A disagreement over a denominator is often more consequential than a disagreement over tone.

Review should protect dignity. A team that fears humiliation when a pilot fails will hide uncertain or negative results. Leaders can reward an honest early stop when evidence contradicts the hypothesis, while still examining whether the initial experiment was designed responsibly. Stopping a weak method can be evidence that an organisation is learning, not evidence that it lacks ambition.

Keep the roles clear. The person who has authority to stop must be identifiable, and the people most affected should have channels to submit evidence. A process that collects staff or student feedback but gives nobody decision rights is likely to generate frustration rather than better judgment.

Why dashboards can strengthen the wrong commitment

A dashboard can make continuation feel objective even when it reports the wrong indicators. The number of hours used, pages produced, meetings held or features delivered may measure effort without measuring the benefit that justified the project. When teams are pressured to prove progress, activity can become a substitute for outcomes. Heavy previous activity then becomes further evidence that stopping would waste everyone’s work.

Define leading and lagging measures carefully. A leading measure such as students attempting practice questions may indicate engagement, but it does not by itself establish durable understanding. A lagging assessment score may reflect external factors. Use multiple relevant measures, specify when they are expected to change, and avoid inventing a causal story from a single improvement after implementation.

Denominators matter. If a trial started with 100 participants and only 60 completed a later questionnaire, reporting that 48 of 60 completers were satisfied describes 80% of respondents, not 80% of the original cohort. The programme may need to understand the 40 missing voices before treating satisfaction as a success signal. Selection into follow-up can create a misleadingly favourable result.

A useful dashboard includes a place for failure evidence, exceptions and unresolved harms. It also includes what the system is not measuring. When hard-to-measure outcomes are important, use careful qualitative review rather than declaring them zero by omission.

When sunk-cost reasoning hides inside moral language

People often describe a commitment in ethical terms: do not abandon the team, keep promises, protect those who relied on the plan. Those can be genuine considerations, not simply rationalisations. A school cannot irresponsibly discontinue a service on which vulnerable learners depend just because a spreadsheet shows a slight deficit. The decision must account for obligations that remain live and for the cost of a safe transition.

The remedy is to separate the obligation from the chosen implementation. If students need continued support, perhaps the organisation should maintain transitional access while moving to a better tool. If a staff member has been promised training, perhaps that promise can be fulfilled through an alternative course. A change in method need not mean abandonment of the people involved.

Ask which commitments are legally, ethically or practically binding; what their cost is under each available alternative; and whether the commitment can be honoured in a different way. This turns an all-or-nothing argument into a solvable design problem. ‘We owe people continuity’ is a stronger claim when accompanied by a credible continuity plan than when used only to protect a failing procurement decision.

Good institutional intelligence can express gratitude for earlier effort while changing course. Appreciation is about people and learning; future funding is about prospective value and duties. Confusing the two can damage both.

Practical diagnosis: seven signals to investigate

The first signal is a justification dominated by phrases about how much has already been spent. The second is a repeated forecast that moves the expected success date forward without explaining why the underlying mechanism will now work. The third is a changing definition of success, especially when hard outcomes are replaced by easier activity metrics. None alone proves escalation; each warrants a closer look.

The fourth signal is a narrow choice set: decision-makers are shown only continuing or total failure, not a pause, smaller scope or alternative method. The fifth is a loss of independent challenge: contrary evidence is absent from meeting records even when frontline workers report problems privately. The sixth is a growing gap between what the dashboard declares and what users actually experience. The seventh is a refusal to define the next test that could falsify the plan.

Diagnosis should be fair to operational constraints. A team might need time to meet regulatory milestones, honour transition commitments or collect a sufficiently large sample. An apparently unchanged project may be following a legitimate staged plan. Ask for the relevant dates, decisions and evidence before applying a psychological label.

The best response to an emerging signal is a concrete information request: What has changed since the previous decision? Which earlier forecast failed? What is the incremental expected value now? What evidence would make us stop? Can someone independent reproduce the calculation? These questions convert concern into testable work.

Repair a struggling programme without rewriting its history

Begin by freezing the current decision snapshot. Document what has actually been delivered, spent and committed, and separate the remaining contractual obligation from optional spending. Preserve the original plan; do not overwrite it with a revised story. Identify who may lawfully and operationally change the plan. This establishes a trustworthy baseline for the next choice.

Next define the attainable objective. If the original target is unrealistic, decide whether a smaller target still has value. The team may change the implementation without changing the educational purpose. Explore at least two plausible alternatives and estimate the marginal resource requirements for each. Where estimates are uncertain, state plausible ranges and test whether the preferred choice survives them.

Run a time-boxed decision review. Assign someone not personally invested in the launch to check the calculations, assumptions and evidence. Include frontline users who understand costs that budget spreadsheets may miss. Decide to continue, narrow, pivot, pause or stop. Write the reason in terms of future value, obligations and risk, not reputational defence.

Finally set a post-decision check. If a pilot is continued for four more weeks, decide now which observations would count as success, failure or inconclusive evidence. An extension without a learning condition is simply another increment of unexamined commitment. A bounded extension can be rational when it buys information worth more than its cost.

Transfer problem: choosing a study course after a non-refundable fee

Consider a student who prepaid S$240 for ten optional online revision sessions. After two sessions, the student discovers that the material repeats what they already know. The S$240 fee cannot be refunded or transferred. Eight sessions remain. Attending them would require twelve additional hours. A different free school-supported study group is available for twelve hours and appears better matched to the student’s current difficulties. Which option should the student choose?

The relevant future cash fee for either option is zero; the S$240 is already unrecoverable. The time commitment is twelve hours either way, so compare the expected learning benefit and any real obligations, not the amount already paid. If a credible diagnostic shows that the group targets the student’s weak topics while the prepaid lessons mostly repeat mastered material, switching is sensible on the stated facts. The prepaid fee is not ‘recovered’ by attendance; attendance adds further time consumption.

Now change the context. Suppose the student discovers that the remaining paid sessions will include individual feedback on exactly the weak topics, whereas the free group is overcrowded. The available evidence about future benefits has changed. Attending the paid sessions may now be rational, not because the S$240 was spent but because the remaining sessions are more useful. A good decision rule survives both versions of the problem.

Students can use this principle for textbooks, memorisation methods, extracurricular activities and revision schedules. Parents can help by separating encouragement to persevere from pressure to justify a past purchase. The goal is to learn well, not to prove that every earlier expense was perfect.

From personal judgment to organisational intelligence

In teams, intelligence is not only the quality of an individual prediction. It includes how evidence travels, how authority is exercised and how an organisation notices when conditions have changed. A group can contain brilliant people and still make poor continuation decisions if it rewards deference, treats uncertainty as weakness or makes stopping procedurally impossible.

Organisations can use rotating reviewers, written alternatives, pre-agreed review dates and clear decision rights to reduce that risk. None guarantees a correct outcome. The point is to make information visible and enable appropriate revision before costs accumulate. Another helpful practice is to recognise the people who detect a weak assumption early, rather than celebrating only those who present good news.

The neighbouring article on pluralistic ignorance explores why people may privately disagree while publicly appearing to support a plan. Together, those mechanisms explain how a project can acquire the appearance of unanimous confidence just when it most needs dissenting evidence.

A healthy institution can say: we made a reasonable choice on the evidence available, the situation has changed, and therefore our next choice will change too. That sentence preserves accountability without pretending that a previous decision must remain right forever.

Limits of forecasts and the dangers of false precision

Prospective calculations are useful only when their inputs deserve some confidence. An estimated S$18,000 benefit may conceal assumptions about adoption, training time, teacher workload and student outcomes. If those assumptions are weak, the appropriate result is a range or scenario rather than a single supposedly exact verdict. An organisation should avoid using financial notation to disguise evidence gaps.

Qualitative benefits may be important even when they are hard to price. A programme could support inclusion, safety or community trust. Evaluate those effects explicitly with relevant stakeholders. The inability to turn them into dollars does not make them worthless, and a claimed social benefit does not justify ignoring serious harm or a better alternative. Rigour is about making the reasons inspectable.

Do not confuse sunk-cost awareness with perpetual impatience. Some objectives require long investment horizons and do not show immediate progress. A genuine long-horizon case should specify milestones or intermediate evidence consistent with the mechanism. It should also name the conditions under which new evidence would justify changing course.

The method applies imperfectly when stopping would violate binding commitments or create abrupt harm. Include transition requirements and statutory duties in the future-cost comparison. A responsible exit can be staged rather than instantaneous.

Reader pathways and a small rehearsal exercise

If you are a student, choose one study habit you maintain mainly because you have invested effort in it. Do not discard it reflexively. Define the learning objective, test your current method on a small set of questions and compare it with a plausible alternative. Decide on the next week based on evidence about learning, not the thickness of your old notes.

If you are a parent, ask what your child gains from the next session, book or activity rather than what the family already paid. Respect commitments, preferences and emotional context. Invite your child to explain what would change their mind. This is a rehearsal in revisable judgement, not a licence for adults to impose a favourite option.

If you teach, make review criteria visible before a classroom pilot begins. Include what should improve, how it will be measured, what obstacles are expected and when the class or team will revisit the approach. Document the first test and the repair, so students see that changing a method can be a sign of learning. The wider purpose of learning beyond the classroom is discussed in the Gold Standard of Education.

If you lead a team, give projects explicit exit, pivot and evidence-gathering options. Separate sponsor advocacy from independent review, and record denominators and alternatives in the same memo. Start with one decision where the team can compare predicted future value with the actual results later. Over time, the quality of those comparisons matters more than any single dramatic rescue story.

Frequently asked questions

Is every attempt to save a failing project irrational? No. Additional investment can be justified when credible future benefits exceed future costs and available alternatives. Escalation is a concern when unrecoverable history dominates the explanation despite deteriorating prospects.

Does sunk cost mean earlier work had no value? No. It can create knowledge, relationships, reusable materials or skills. Those present and future values are relevant. Only the nonrecoverable portion should be excluded from incremental cash-flow comparisons.

Should schools automatically cancel any pilot below target? No. An exit rule may trigger diagnosis, additional measurement or a change of method. Review the validity of the target, the student context, possible harms and transition obligations before choosing.

What is the difference between sunk cost and opportunity cost? A sunk cost has already been incurred and cannot change. Opportunity cost is the value of an alternative you forgo by making the next decision. Mixing them reverses the direction of a proper comparison.

Can a leader avoid escalation by delegating all decisions? Delegation alone is not sufficient. Reviewers require real authority, independent evidence, clear criteria and freedom to report unwelcome findings. Otherwise the same incentives can be reproduced in a new role.

What if the benefit cannot be measured in money? Describe the intended outcomes and evidence, include nonfinancial effects in the choice criteria, and compare alternatives honestly. Avoid treating unpriced benefits as either zero or infinite.

How do I disagree with a colleague without accusing them of irrationality? Ask for the incremental case: what benefit will the next investment create, what would change the forecast, and what alternatives are feasible? The discussion is more productive when it examines the next choice rather than attacking the past one.

Could ending a project too early also be a mistake? Certainly. Stopping prematurely can destroy future value. Use time horizons, milestones and uncertainty ranges appropriate to the mechanism. A rational process is revisable rather than automatically pessimistic.

What is the fastest practical check? Remove irrecoverable past expenditure from your forward-looking comparison; calculate future incremental outcomes for each feasible option; name one disconfirming observation; and give someone independent permission to challenge the result.

How is this different from outcome bias? Outcome bias judges the quality of a decision by what happened afterward. Escalation of commitment concerns the choice to keep investing despite changing information, often under the influence of past costs. A good review can guard against both.

Sources and the closing decision test

The central behavioural references are Arkes and Blumer, The Psychology of Sunk Cost (1985), Staw, The Escalation of Commitment to a Course of Action (1981), Whyte, Escalating Commitment: A Reinterpretation (1986), and Bateman’s 1986 experimental work on moderators and limiting conditions. These sources address related but not identical questions. Use them as research foundations rather than as a diagnostic label for someone you have not studied.

The reader can explore adjacent problems in the How Intelligence Works hub, including collective judgment and institutional memory. The general question of learning from decisions is broader than a financial spreadsheet: it concerns keeping purposes, evidence, obligations and authority aligned as circumstances change.

Before the next consequential commitment, write one paragraph that would convince a reasonable outsider to fund the next step today even if they had never paid for the earlier steps. Include a credible benefit mechanism, future resource needs, the best alternative and a condition that would make you reconsider. If the paragraph relies chiefly on the pain of earlier investment, the plan needs another review.

Intelligence is not the ability to defend every decision indefinitely. It is the capacity to keep a worthwhile aim in view while letting accurate new information change the path.

Discover more from eduKate Singapore

Subscribe now to keep reading and get access to the full archive.

Continue reading