{"meta":{"slug":"cognitive-biases-decision-making","area":"strategia","data":"2026-10-03","autore":"Redazione Prodability","meta_title":"Cognitive biases in decision making at work","meta_description":"The most common cognitive biases in the decisions of those who run a company, the behaviors that reveal them and the countermeasure that reduces each one.","keyword_principale":"cognitive biases in decision making","keywords_secondarie":"confirmation bias, types of cognitive bias, cognitive bias examples, anchoring bias, sunk cost fallacy, outcome bias","tags":["Decision-making","Problem solving","Leadership","Meetings"],"title":"Business decisions: the cognitive biases to recognize","lunghezza":"17 min read","featuredVisual":{"kind":"image","src":"/article-assets/bias-cognitivi/en/cognitive-biases-decision-making.jpg","alt":"Business decisions: the cognitive biases to recognize"}},"content":"# Business decisions: the cognitive biases to recognize\n\nIn a meeting, a proposal is approved in twenty minutes and the question of what might disprove it is never asked: is that consensus or a flaw in the procedure?\n\nThe two look alike, and telling them apart means looking at how the decision was made.\n\nCognitive biases are systematic distortions of judgment: the Treccani dictionary records *bias* as a \"cognitive distortion, caused by prejudices, that leads to wrong predictions\" [5].\n\nSystematic means they repeat in the same direction, and that they also affect people who are experienced and rewarded for answering correctly [1].\n\nThe issue concerns the professional with two team members, the department manager and the person running a hundred-person company, and it weighs most where a second opinion is structurally missing: in 2022, 80.9% of Italian companies with at least three employees were controlled by an individual or a family [6].\n\n> **Cognitive biases are not corrected by the willpower of whoever decides: they are reduced by changing the process — criteria written beforehand, data gathered before the discussion, a role that argues the opposite, decisions recorded with their reasons.**\n\nWhat follows: what distinguishes a bias from a slip, the four that recur in the decisions of those who run a company — confirmation, anchoring, sunk costs, outcome — and the mistakes that cancel out their countermeasures.\n\n## Telling an occasional mistake from a cognitive bias\n\nGetting a sum wrong is one thing; getting it wrong in the same direction every time the conditions repeat is another: the second has a name and a defense procedure.\n\nThe word that carries the weight, in the definition recorded by the Treccani dictionary, is distortion: not a random stumble, but a directed deviation that repeats in the same direction [5].\n\nThree boundaries separate it from neighboring words.\n\n- **Slip.** A mistake caused by distraction changes direction each time and is corrected by rereading; a bias resists rereading, because it is the criterion that is skewed, not the execution.\n- **Opinion.** An opinion is stated, and for that reason it can be discussed in a meeting; a bias acts without being stated, and whoever is subject to it is convinced they have assessed the facts.\n- **Self-interest.** Tversky and Kahneman already observed in 1974 that the distortions they described cannot be attributed to motivational effects: they appear even when the person judging is encouraged to be accurate and rewarded for correct answers [1].\n\nThe third boundary is the most useful, because it shifts the problem: if the flaw does not lie in the good faith of whoever decides, asking for more attention won't correct it.\n\nThe types of bias cataloged in the literature number in the dozens, and a complete list would be of little use.\n\nThis article deals with four, chosen because they show up as behaviors you can observe in a meeting or in a year-end review: seeking only confirmation, anchoring to the first number, persisting with what has already been paid for, judging once the outcome is known.\n\nThe same pattern applies to each: a visible behavior and a countermeasure that lives in the process — and the fact that structured management practices are the subject of institutional surveys shows that the process is measurable ground, not a metaphor [7].\n\nThe criteria for comparing alternatives already on the table are the subject of the [decision matrix](https://blog.prodability.com/matrice-decisionale/): here we look at what distorts judgment before that.\n\n## Recognizing the signs of confirmation bias in a meeting\n\n> Biases are not corrected by asking whoever decides to pay more attention: they are reduced when the process produces, by design, information that the decision-maker did not choose.\n\nA proposal approved in twenty minutes and one approved after someone brought the contrary data produce the same minutes, and they are not worth the same.\n\nNickerson defines confirmation bias as seeking or interpreting evidence in ways that are partial to existing beliefs, expectations or hypotheses in hand [2].\n\nThe clarification that follows is the useful part: it is unwitting selectivity in the use of evidence, which is different from deliberately building a one-sided case [2].\n\nSomeone who looks only for confirmation is not lying: they are looking where they expect to find it.\n\nThe observable behaviors are few and recognizable.\n\n- Data are requested after the decision has taken shape, and serve to back it up.\n- Among customers, the one who has already responded well gets the call, and the field test starts with the most favorable line or sales agent.\n- The question \"what fact would make us change our minds\" does not appear on the agenda.\n- Whoever raises the objection is heard last, when the meeting has run out of time.\n\nThe documented countermeasure consists of bringing alternative hypotheses forward: Nickerson reports evidence in favor of the habit of formulating early, before the assessment, the reasons why the opposite hypothesis might be true [2].\n\nTranslated into organization, bringing them forward takes three forms.\n\n- Whoever gathers the data is not the person who proposed the decision.\n- The disconfirming question is written on the agenda and answered before the discussion.\n- A designated role argues the opposing position, on rotation, and its contribution goes into the minutes like everyone else's.\n\nThe third point has a facilitation cost and lives inside the meeting's rules, not alongside them: how the meeting is set up is covered in the [guide to effective business meetings](https://blog.prodability.com/riunioni-aziendali-efficaci/).\n\nThe first point costs one calendar entry: whoever prepares the case is named before the proposal is formulated.\n\n![Diagram of the four practices that reduce bias in decisions: criteria written beforehand, data gathered by someone other than the proposer, a role that argues the opposite, the decision recorded with its reasons](/article-assets/bias-cognitivi/bias-cognitivi-quattro-gesti.jpg)\n\n## The first number spoken in a meeting weighs more: anchoring\n\nTwo identical quotes, one read after hearing \"we should stay around forty thousand\" and one read before any figure is mentioned: does the judgment on which one is expensive stay the same?\n\nIt doesn't, and the direction of the gap is predictable.\n\nTversky and Kahneman describe the mechanism: estimates start from an initial value that is then adjusted, the adjustments typically turn out to be insufficient, and different starting points produce estimates biased toward the initial value [1].\n\nThe name of the phenomenon is anchoring.\n\nIn the experiment that documents it, different groups were shown a number drawn in front of them on a wheel of fortune, and were then asked to estimate a quantity that had nothing to do with that number.\n\nThe median estimates were 25 and 45 for the groups that had received 10 and 65 as their starting points [1].\n\nThe detail that matters to decision-makers is a different one: paying participants for accuracy did not reduce the effect [1].\n\nAn arbitrary number, recognized as arbitrary, still shifts the estimate.\n\nAt work, anchors go by ordinary names.\n\n- The first quote received becomes the yardstick by which the other three are read.\n- Last year's budget becomes the starting point for the new one, and the discussion revolves around the variance instead of the amount.\n- The first date spoken out loud — \"can it be done by the end of the month?\" — becomes the deadline around which the plan is built.\n\nThe countermeasure is about sequence, not character: numbers are formed before the discussion, not during it.\n\nIn practice, participants hand in their estimate in writing before the meeting, with a minimum-maximum range; the estimates are opened together and the discussion is about the spread, not about who spoke first.\n\nWhen the number comes from outside, the sequence is preserved: the acceptable range is written down before opening the envelope.\n\nThe range is better built on historical data than on impressions, and that is the job of [performance indicators](https://blog.prodability.com/kpi-aziendali-pmi/).\n\n## Carrying on so as not to waste: the sunk cost bias\n\nA project that has absorbed two years and the same project to be started from scratch today should get the same answer to the question \"is it worth continuing?\".\n\nThey get opposite answers, and the gap has been measured.\n\nArkes and Blumer define the sunk cost effect as a greater tendency to continue an endeavor once an investment in money, effort or time has been made, and attribute its justification to the desire not to appear wasteful [4].\n\nIn one of their scenarios, a development project already 90% funded was carried through to completion by 41 respondents out of 48, even though a competitor was already selling a better and cheaper product.\n\nThe exact same remaining commitment, presented instead as new spending on a project not yet started, got 10 votes in favor out of 60 [4].\n\nThe amount still to be spent was the same: the only thing that changed was what had already been spent and could not be recovered.\n\nIn a second scenario, the same faster and cheaper machine, offered to someone running a printing company, was bought by 49 respondents out of 64 when the recent purchase had been of a different kind, and by 43 out of 81 when it had been a similar machine [4].\n\nWhat changed was not the advantage, but the feeling of turning the previous purchase into a waste.\n\nThe scenario is not exotic for a family business that installed equipment last year with a loan still being paid off, and that today sees an opportunity go by.\n\nThere are three observable signals.\n\n- In progress meetings, people cite how much has been invested, and the figure appears among the arguments for continuing.\n- Estimates of success rise as spending rises, without any new fact having emerged: in the experimental data, those who had already incurred a cost inflated the probability of success they attributed to the project [4].\n- The person who decided to start the project is the same person who decides whether to continue, and nobody raises the point.\n\nThe countermeasure is written beforehand, not afterward: at the start, you set the stopping condition and the date on which to check it.\n\nThe condition must be stated in observable terms — a volume, a margin, a cycle time — and the check is a calendar entry.\n\nOn that date, the question to ask is not about how much has been spent: if we had spent nothing today, would we start this project?\n\n## Evaluating a decision without looking at how it turned out\n\nDid the salesperson who closed the contract decide better than the one who lost it?\n\nIt depends on what they knew at the moment they decided, and that is exactly the point the instinctive answer skips.\n\nBaron and Hershey presented groups of subjects with decisions described identically, varying only the reported outcome.\n\nThe same decision received on average 0.85 on a scale from −3 to +3 when the outcome was favorable and −0.05 when it was unfavorable [3].\n\nThe most instructive detail comes afterward: when asked, the subjects stated that the outcome should not have been taken into account, and they took it into account anyway [3].\n\nThat is why outcome bias resists good intentions.\n\nAt work it produces three observable effects.\n\n- A lucky choice made without any criterion becomes the model to repeat, and the criterion is never written down.\n- A supplier selected after a serious review, which then fails for an unforeseeable reason, enters the company's memory as an error of judgment.\n- Decision-makers learn to choose the option that can be defended after the fact rather than the best one at the time, because they have understood how they will be judged.\n\nThe third effect is the most costly, because it changes the behavior even of those who have not been judged.\n\nThe countermeasure is a line written while deciding, not after the fact, and the place where it lives is the meeting minutes: no new document is needed.\n\nAgainst outcome bias, two fields filled in beforehand are enough — the criterion applied and the data available at the time — next to the name of whoever decided and the date of the review.\n\nOn that date, you reread the line and judge the criterion with the information available then, noting separately what happened afterward.\n\nThe difference between \"wrong decision\" and \"unfavorable outcome\" becomes a written distinction, not a nuance lost in the discussion.\n\nThe reverse also holds: a decision made without any criterion that turned out well still needs correcting, and the written line is where the correction doesn't get forgotten.\n\nThe process in which that line finds its place — from the problem to the alternatives, from the criteria to the mandate and on to the review — is described in the guide to [decision making in five steps](https://blog.prodability.com/decision-making/).\n\n![Table of the four cognitive biases covered in the article with the corresponding observable behavior and organizational countermeasure](/article-assets/bias-cognitivi/bias-cognitivi-tabella-bias.jpg)\n\n## Why countermeasures to cognitive biases stop working\n\nA company that started writing down the reasons for its decisions six months ago and never rereads them, and one that never started: which of the two has changed anything in its meetings?\n\nThe four mistakes below are not about the biases, they are about the countermeasures, and you recognize them by looking at the calendar and the documents, not at the people.\n\n**Relying on awareness.** Knowing the name of a bias does not switch it off: Tversky and Kahneman found that even experienced researchers fall into the same errors when they judge intuitively [1], and Arkes and Blumer that the sunk cost effect was not reduced among those who had taken economics courses [4].\n\nThe practice is to shift the burden from the head to the document: a question written on the agenda, a criterion filed beforehand.\n\n**Asking for objections without assigning the role.** According to Italy's permanent business census, management is in most cases in the hands of the business owner or a family member, and the use of an internal or external manager is concentrated in medium and large companies [6].\n\nIn a family business with long-standing team members, a generic invitation to object produces silence, because the objection costs the person who raises it.\n\nThe practice is to name the role on rotation and record it in the minutes: arguing the opposite becomes an assigned task, not an act of courage.\n\n**Applying the procedure to every decision.** On reversible, low-value choices the time cost exceeds the benefit, and the procedure is abandoned wholesale.\n\nThe practice is to set two thresholds — amount and reversibility — and apply the process only above them.\n\n**Recording the outcome and not the criterion.** A line that notes how things went reproduces the very outcome bias it should contain.\n\nThe practice is to check that the criterion field was filled in beforehand, and to consider any line without it incomplete.\n\nThe final check is a matter of reading: if after a quarter the lines show different criteria for different decisions, the process exists; if they show the same criterion written after the fact, it is a box-ticking exercise.\n\n## Limitations and conditions of applicability\n\nThe experimental studies cited were conducted in the laboratory, largely on university students and with written scenarios [3] [4]: they document the existence and direction of the effect, not its intensity in any single company.\n\nThe Italian permanent business census figure describes an ownership structure [6]: it says how common it is for the decision to rest with one person, not that those who decide alone decide worse.\n\nThe Bank of Italy paper cited for the measurement of management practices is explicitly descriptive [7]: it finds associations between ownership structures, practices and results, not causal links.\n\nThe countermeasures described have a time cost, and on reversible, low-value decisions that cost exceeds the benefit.\n\nWhen a decision has to be made within a few hours — an equipment shutdown, a contract deadline, a safety issue — the sequence is shortened and the recording takes place afterward, with the criterion reconstructed and declared as such.\n\nThe procedures described here do not eliminate biases: they reduce the probability that a single distortion reaches the decision undisturbed.\n\n## FAQ\n\n### What is confirmation bias?\n\nIt is the tendency to seek and interpret evidence in ways that are partial to beliefs, expectations or hypotheses already held [2].\n\nIt is not bad faith: it is unwitting selectivity in gathering and using information [2].\n\nYou can recognize it by a simple signal: data are requested after the decision has taken shape.\n\n### How many types of bias are there?\n\nThe literature has cataloged dozens, mostly grouped according to the judgment heuristic they derive from [1].\n\nThe number matters less than the selection criterion: at work, the useful ones are those that produce an observable behavior and for which a process countermeasure exists.\n\nThe four covered in this article meet both requirements.\n\n### What are some examples of cognitive biases in work decisions?\n\nThe first quote that becomes the yardstick for judging the others is anchoring [1].\n\nThe outdated project that is carried through because it is already 90% funded is the sunk cost effect [4].\n\nThe supplier chosen after a serious review and then filed away as a mistake because it failed is outcome bias [3].\n\n### Is knowing about biases enough to avoid them?\n\nIt isn't, and the studies cited on this page show it on two different fronts.\n\nEven experienced researchers are subject to them when they judge intuitively [1], and the sunk cost effect was not reduced among those who had taken economics courses [4].\n\nThe reduction comes through the process — criteria written beforehand, data gathered by someone other than the proposer, reviews on a fixed date — not through awareness.\n\n## Key takeaways\n\nThe starting point is a distinction: a mistake caused by distraction changes direction each time, a bias repeats in the same direction and resists rereading.\n\nFor decisions above an amount threshold or that are hard to reverse, it pays to apply four practices, each aimed at a different bias.\n\nAgainst seeking only confirmation, whoever gathers the data is not the proposer, and the disconfirming question is on the agenda with an answer before the discussion.\n\nAgainst anchoring, estimates are handed in writing before the meeting, with a minimum-maximum range, and opened together.\n\nAgainst sunk costs, the stopping condition and the review date are written at the start, in observable terms.\n\nAgainst judging once the outcome is known, every decision leaves a line in the minutes with the criterion applied, the data available at the time and the review date.\n\nThe role of whoever argues the opposite should be assigned on rotation and recorded in the minutes, because a generic invitation to object produces silence.\n\nThe last item on the agenda is the date on which the criteria are reread, and on which the reasoning is judged with the information available at the time.\n\n## Conclusion\n\nCognitive biases are not attention failures to be fixed with effort: they are deviations that repeat in the same direction, and they are reduced by changing the process through which the decision is made.\n\nEach of the four described here has a behavior that makes it visible, and each has a countermeasure that lives in a document and a date, not in a resolution.\n\nCriteria written beforehand, data gathered before the discussion, a role that argues the opposite, decisions recorded with their reasons: four practices that cost minutes and leave a trace in the minutes.\n\nThe four practices don't stand on their own: they are the points where the [five-step decision-making process](https://blog.prodability.com/decision-making/) adds friction before the choice is closed.\n\nWhen the alternatives are already on the table and you need a criterion to compare them, the next step is the [decision matrix](https://blog.prodability.com/matrice-decisionale/); when instead the decision keeps getting put off, the problem changes nature and is described in [putting off decisions](https://blog.prodability.com/rimandare-decisioni/).\n\nIf you're looking for the framework in which these practices sit, you'll find it in the [guide to business problem solving](https://blog.prodability.com/problem-solving-aziendale/).\n\nAfter a few quarters of filled-in lines, the Monday meeting changes its subject.\n\nNo longer the defense of a choice already made, but the reading of a criterion written beforehand and the comparison with what happened afterward.\n\nIt is the shift from a company in which decisions are argued out loud to one in which they can be reread, and in which those who make them no longer have to defend themselves against the outcome in order to defend their reasoning.\n\n\n\n## Sources and references\n\n[1] Tversky, A., Kahneman, D., \"Judgment under Uncertainty: Heuristics and Biases\", Science, New Series, vol. 185, no. 4157, September 27, 1974, pp. 1124-1131, DOI 10.1126/science.185.4157.1124. Full text consulted at: https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf (University of California, Irvine)\n\n[2] Nickerson, R. S., \"Confirmation Bias: A Ubiquitous Phenomenon in Many Guises\", Review of General Psychology, vol. 2, no. 2, 1998, pp. 175-220, DOI 10.1037/1089-2680.2.2.175. Full text consulted at: https://pages.ucsd.edu/~mckenzie/nickersonConfirmationBias.pdf (University of California, San Diego)\n\n[3] Baron, J., Hershey, J. C., \"Outcome Bias in Decision Evaluation\", Journal of Personality and Social Psychology, vol. 54, no. 4, 1988, pp. 569-579. Author's copy available at: https://www.sas.upenn.edu/~baron/papers/outcomebias.pdf (University of Pennsylvania; archived copy: https://web.archive.org/web/2020/https://www.sas.upenn.edu/~baron/papers/outcomebias.pdf)\n\n[4] Arkes, H. R., Blumer, C., \"The Psychology of Sunk Cost\", Organizational Behavior and Human Decision Processes, vol. 35, no. 1, February 1985, pp. 124-140, DOI 10.1016/0749-5978(85)90049-4 (canonical reference verified on Crossref). The publisher does not offer open access and no copy on an academic domain could be found: the full text consulted, with the original journal pagination and header, is hosted by a private archive — https://www.communicationcache.com/uploads/1/0/8/8/10887248/the_psychology_of_sunk_cost.pdf (archived copy: https://web.archive.org/web/20260918070345/http://www.communicationcache.com/uploads/1/0/8/8/10887248/the_psychology_of_sunk_cost.pdf)\n\n[5] Treccani, entry \"bias\", Neologismi (2024), Istituto della Enciclopedia Italiana. Available at: https://www.treccani.it/vocabolario/neo-bias_(Neologismi)/\n\n[6] ISTAT, \"Censimento permanente delle imprese 2023: primi risultati\", press release, November 14, 2023 (reference year 2022; about 280,000 responding companies, representative of 1,021,618 companies with at least 3 employees). Available at: https://www.istat.it/comunicato-stampa/censimento-permanente-delle-imprese-2023-primi-risultati/ — PDF: https://www.istat.it/it/files/2023/11/REPORTCensimprese.pdf\n\n[7] Baltrunaite, A., Formai, S., Linarello, A., Mocetti, S., \"Ownership, governance, management and firm performance: evidence from Italian firms\", Questioni di Economia e Finanza (Occasional Papers) no. 678, Banca d'Italia, March 2022, DOI 10.32057/0.QEF.2022.0678. Available at: https://www.bancaditalia.it/pubblicazioni/qef/2022-0678/index.html — PDF: https://www.bancaditalia.it/pubblicazioni/qef/2022-0678/QEF_678_22.pdf","path":"content/articles/art-0165/en.md","routePath":"cognitive-biases-decision-making","wordCount":3698,"imageMeta":{"/article-assets/bias-cognitivi/bias-cognitivi.jpg":{"w":1200,"h":825},"/article-assets/bias-cognitivi/bias-cognitivi-quattro-gesti.jpg":{"w":893,"h":1600},"/article-assets/bias-cognitivi/bias-cognitivi-tabella-bias.jpg":{"w":1600,"h":873},"/article-assets/bias-cognitivi/en/cognitive-biases-decision-making.jpg":{"w":1200,"h":825}},"html":"

The two look alike, and telling them apart means looking at how the decision was made.

\n

Cognitive biases are systematic distortions of judgment: the Treccani dictionary records bias as a \"cognitive distortion, caused by prejudices, that leads to wrong predictions\" [5].

\n

Systematic means they repeat in the same direction, and that they also affect people who are experienced and rewarded for answering correctly [1].

\n

The issue concerns the professional with two team members, the department manager and the person running a hundred-person company, and it weighs most where a second opinion is structurally missing: in 2022, 80.9% of Italian companies with at least three employees were controlled by an individual or a family [6].

\n
\n

Cognitive biases are not corrected by the willpower of whoever decides: they are reduced by changing the process — criteria written beforehand, data gathered before the discussion, a role that argues the opposite, decisions recorded with their reasons.

\n
\n

What follows: what distinguishes a bias from a slip, the four that recur in the decisions of those who run a company — confirmation, anchoring, sunk costs, outcome — and the mistakes that cancel out their countermeasures.

\n

Telling an occasional mistake from a cognitive bias

\n

Getting a sum wrong is one thing; getting it wrong in the same direction every time the conditions repeat is another: the second has a name and a defense procedure.

\n

The word that carries the weight, in the definition recorded by the Treccani dictionary, is distortion: not a random stumble, but a directed deviation that repeats in the same direction [5].

\n

Three boundaries separate it from neighboring words.

\n\n

The third boundary is the most useful, because it shifts the problem: if the flaw does not lie in the good faith of whoever decides, asking for more attention won't correct it.

\n

The types of bias cataloged in the literature number in the dozens, and a complete list would be of little use.

\n

This article deals with four, chosen because they show up as behaviors you can observe in a meeting or in a year-end review: seeking only confirmation, anchoring to the first number, persisting with what has already been paid for, judging once the outcome is known.

\n

The same pattern applies to each: a visible behavior and a countermeasure that lives in the process — and the fact that structured management practices are the subject of institutional surveys shows that the process is measurable ground, not a metaphor [7].

\n

The criteria for comparing alternatives already on the table are the subject of the decision matrix: here we look at what distorts judgment before that.

\n

Recognizing the signs of confirmation bias in a meeting

\n
\n

Biases are not corrected by asking whoever decides to pay more attention: they are reduced when the process produces, by design, information that the decision-maker did not choose.

\n
\n

A proposal approved in twenty minutes and one approved after someone brought the contrary data produce the same minutes, and they are not worth the same.

\n

Nickerson defines confirmation bias as seeking or interpreting evidence in ways that are partial to existing beliefs, expectations or hypotheses in hand [2].

\n

The clarification that follows is the useful part: it is unwitting selectivity in the use of evidence, which is different from deliberately building a one-sided case [2].

\n

Someone who looks only for confirmation is not lying: they are looking where they expect to find it.

\n

The observable behaviors are few and recognizable.

\n\n

The documented countermeasure consists of bringing alternative hypotheses forward: Nickerson reports evidence in favor of the habit of formulating early, before the assessment, the reasons why the opposite hypothesis might be true [2].

\n

Translated into organization, bringing them forward takes three forms.

\n\n

The third point has a facilitation cost and lives inside the meeting's rules, not alongside them: how the meeting is set up is covered in the guide to effective business meetings.

\n

The first point costs one calendar entry: whoever prepares the case is named before the proposal is formulated.

\n

\"Diagram

\n

The first number spoken in a meeting weighs more: anchoring

\n

Two identical quotes, one read after hearing \"we should stay around forty thousand\" and one read before any figure is mentioned: does the judgment on which one is expensive stay the same?

\n

It doesn't, and the direction of the gap is predictable.

\n

Tversky and Kahneman describe the mechanism: estimates start from an initial value that is then adjusted, the adjustments typically turn out to be insufficient, and different starting points produce estimates biased toward the initial value [1].

\n

The name of the phenomenon is anchoring.

\n

In the experiment that documents it, different groups were shown a number drawn in front of them on a wheel of fortune, and were then asked to estimate a quantity that had nothing to do with that number.

\n

The median estimates were 25 and 45 for the groups that had received 10 and 65 as their starting points [1].

\n

The detail that matters to decision-makers is a different one: paying participants for accuracy did not reduce the effect [1].

\n

An arbitrary number, recognized as arbitrary, still shifts the estimate.

\n

At work, anchors go by ordinary names.

\n\n

The countermeasure is about sequence, not character: numbers are formed before the discussion, not during it.

\n

In practice, participants hand in their estimate in writing before the meeting, with a minimum-maximum range; the estimates are opened together and the discussion is about the spread, not about who spoke first.

\n

When the number comes from outside, the sequence is preserved: the acceptable range is written down before opening the envelope.

\n

The range is better built on historical data than on impressions, and that is the job of performance indicators.

\n

Carrying on so as not to waste: the sunk cost bias

\n

A project that has absorbed two years and the same project to be started from scratch today should get the same answer to the question \"is it worth continuing?\".

\n

They get opposite answers, and the gap has been measured.

\n

Arkes and Blumer define the sunk cost effect as a greater tendency to continue an endeavor once an investment in money, effort or time has been made, and attribute its justification to the desire not to appear wasteful [4].

\n

In one of their scenarios, a development project already 90% funded was carried through to completion by 41 respondents out of 48, even though a competitor was already selling a better and cheaper product.

\n

The exact same remaining commitment, presented instead as new spending on a project not yet started, got 10 votes in favor out of 60 [4].

\n

The amount still to be spent was the same: the only thing that changed was what had already been spent and could not be recovered.

\n

In a second scenario, the same faster and cheaper machine, offered to someone running a printing company, was bought by 49 respondents out of 64 when the recent purchase had been of a different kind, and by 43 out of 81 when it had been a similar machine [4].

\n

What changed was not the advantage, but the feeling of turning the previous purchase into a waste.

\n

The scenario is not exotic for a family business that installed equipment last year with a loan still being paid off, and that today sees an opportunity go by.

\n

There are three observable signals.

\n\n

The countermeasure is written beforehand, not afterward: at the start, you set the stopping condition and the date on which to check it.

\n

The condition must be stated in observable terms — a volume, a margin, a cycle time — and the check is a calendar entry.

\n

On that date, the question to ask is not about how much has been spent: if we had spent nothing today, would we start this project?

\n

Evaluating a decision without looking at how it turned out

\n

Did the salesperson who closed the contract decide better than the one who lost it?

\n

It depends on what they knew at the moment they decided, and that is exactly the point the instinctive answer skips.

\n

Baron and Hershey presented groups of subjects with decisions described identically, varying only the reported outcome.

\n

The same decision received on average 0.85 on a scale from −3 to +3 when the outcome was favorable and −0.05 when it was unfavorable [3].

\n

The most instructive detail comes afterward: when asked, the subjects stated that the outcome should not have been taken into account, and they took it into account anyway [3].

\n

That is why outcome bias resists good intentions.

\n

At work it produces three observable effects.

\n\n

The third effect is the most costly, because it changes the behavior even of those who have not been judged.

\n

The countermeasure is a line written while deciding, not after the fact, and the place where it lives is the meeting minutes: no new document is needed.

\n

Against outcome bias, two fields filled in beforehand are enough — the criterion applied and the data available at the time — next to the name of whoever decided and the date of the review.

\n

On that date, you reread the line and judge the criterion with the information available then, noting separately what happened afterward.

\n

The difference between \"wrong decision\" and \"unfavorable outcome\" becomes a written distinction, not a nuance lost in the discussion.

\n

The reverse also holds: a decision made without any criterion that turned out well still needs correcting, and the written line is where the correction doesn't get forgotten.

\n

The process in which that line finds its place — from the problem to the alternatives, from the criteria to the mandate and on to the review — is described in the guide to decision making in five steps.

\n

\"Table

\n

Why countermeasures to cognitive biases stop working

\n

A company that started writing down the reasons for its decisions six months ago and never rereads them, and one that never started: which of the two has changed anything in its meetings?

\n

The four mistakes below are not about the biases, they are about the countermeasures, and you recognize them by looking at the calendar and the documents, not at the people.

\n

Relying on awareness. Knowing the name of a bias does not switch it off: Tversky and Kahneman found that even experienced researchers fall into the same errors when they judge intuitively [1], and Arkes and Blumer that the sunk cost effect was not reduced among those who had taken economics courses [4].

\n

The practice is to shift the burden from the head to the document: a question written on the agenda, a criterion filed beforehand.

\n

Asking for objections without assigning the role. According to Italy's permanent business census, management is in most cases in the hands of the business owner or a family member, and the use of an internal or external manager is concentrated in medium and large companies [6].

\n

In a family business with long-standing team members, a generic invitation to object produces silence, because the objection costs the person who raises it.

\n

The practice is to name the role on rotation and record it in the minutes: arguing the opposite becomes an assigned task, not an act of courage.

\n

Applying the procedure to every decision. On reversible, low-value choices the time cost exceeds the benefit, and the procedure is abandoned wholesale.

\n

The practice is to set two thresholds — amount and reversibility — and apply the process only above them.

\n

Recording the outcome and not the criterion. A line that notes how things went reproduces the very outcome bias it should contain.

\n

The practice is to check that the criterion field was filled in beforehand, and to consider any line without it incomplete.

\n

The final check is a matter of reading: if after a quarter the lines show different criteria for different decisions, the process exists; if they show the same criterion written after the fact, it is a box-ticking exercise.

\n

Limitations and conditions of applicability

\n

The experimental studies cited were conducted in the laboratory, largely on university students and with written scenarios [3] [4]: they document the existence and direction of the effect, not its intensity in any single company.

\n

The Italian permanent business census figure describes an ownership structure [6]: it says how common it is for the decision to rest with one person, not that those who decide alone decide worse.

\n

The Bank of Italy paper cited for the measurement of management practices is explicitly descriptive [7]: it finds associations between ownership structures, practices and results, not causal links.

\n

The countermeasures described have a time cost, and on reversible, low-value decisions that cost exceeds the benefit.

\n

When a decision has to be made within a few hours — an equipment shutdown, a contract deadline, a safety issue — the sequence is shortened and the recording takes place afterward, with the criterion reconstructed and declared as such.

\n

The procedures described here do not eliminate biases: they reduce the probability that a single distortion reaches the decision undisturbed.

\n

FAQ

\n

What is confirmation bias?

\n

It is the tendency to seek and interpret evidence in ways that are partial to beliefs, expectations or hypotheses already held [2].

\n

It is not bad faith: it is unwitting selectivity in gathering and using information [2].

\n

You can recognize it by a simple signal: data are requested after the decision has taken shape.

\n

How many types of bias are there?

\n

The literature has cataloged dozens, mostly grouped according to the judgment heuristic they derive from [1].

\n

The number matters less than the selection criterion: at work, the useful ones are those that produce an observable behavior and for which a process countermeasure exists.

\n

The four covered in this article meet both requirements.

\n

What are some examples of cognitive biases in work decisions?

\n

The first quote that becomes the yardstick for judging the others is anchoring [1].

\n

The outdated project that is carried through because it is already 90% funded is the sunk cost effect [4].

\n

The supplier chosen after a serious review and then filed away as a mistake because it failed is outcome bias [3].

\n

Is knowing about biases enough to avoid them?

\n

It isn't, and the studies cited on this page show it on two different fronts.

\n

Even experienced researchers are subject to them when they judge intuitively [1], and the sunk cost effect was not reduced among those who had taken economics courses [4].

\n

The reduction comes through the process — criteria written beforehand, data gathered by someone other than the proposer, reviews on a fixed date — not through awareness.

\n

Key takeaways

\n

The starting point is a distinction: a mistake caused by distraction changes direction each time, a bias repeats in the same direction and resists rereading.

\n

For decisions above an amount threshold or that are hard to reverse, it pays to apply four practices, each aimed at a different bias.

\n

Against seeking only confirmation, whoever gathers the data is not the proposer, and the disconfirming question is on the agenda with an answer before the discussion.

\n

Against anchoring, estimates are handed in writing before the meeting, with a minimum-maximum range, and opened together.

\n

Against sunk costs, the stopping condition and the review date are written at the start, in observable terms.

\n

Against judging once the outcome is known, every decision leaves a line in the minutes with the criterion applied, the data available at the time and the review date.

\n

The role of whoever argues the opposite should be assigned on rotation and recorded in the minutes, because a generic invitation to object produces silence.

\n

The last item on the agenda is the date on which the criteria are reread, and on which the reasoning is judged with the information available at the time.

\n

Conclusion

\n

Cognitive biases are not attention failures to be fixed with effort: they are deviations that repeat in the same direction, and they are reduced by changing the process through which the decision is made.

\n

Each of the four described here has a behavior that makes it visible, and each has a countermeasure that lives in a document and a date, not in a resolution.

\n

Criteria written beforehand, data gathered before the discussion, a role that argues the opposite, decisions recorded with their reasons: four practices that cost minutes and leave a trace in the minutes.

\n

The four practices don't stand on their own: they are the points where the five-step decision-making process adds friction before the choice is closed.

\n

When the alternatives are already on the table and you need a criterion to compare them, the next step is the decision matrix; when instead the decision keeps getting put off, the problem changes nature and is described in putting off decisions.

\n

If you're looking for the framework in which these practices sit, you'll find it in the guide to business problem solving.

\n

After a few quarters of filled-in lines, the Monday meeting changes its subject.

\n

No longer the defense of a choice already made, but the reading of a criterion written beforehand and the comparison with what happened afterward.

\n

It is the shift from a company in which decisions are argued out loud to one in which they can be reread, and in which those who make them no longer have to defend themselves against the outcome in order to defend their reasoning.

\n\n

Sources and references

\n

[1] Tversky, A., Kahneman, D., \"Judgment under Uncertainty: Heuristics and Biases\", Science, New Series, vol. 185, no. 4157, September 27, 1974, pp. 1124-1131, DOI 10.1126/science.185.4157.1124. Full text consulted at: https://sites.socsci.uci.edu/~bskyrms/bio/readings/tversky_k_heuristics_biases.pdf (University of California, Irvine)

\n

[2] Nickerson, R. S., \"Confirmation Bias: A Ubiquitous Phenomenon in Many Guises\", Review of General Psychology, vol. 2, no. 2, 1998, pp. 175-220, DOI 10.1037/1089-2680.2.2.175. Full text consulted at: https://pages.ucsd.edu/~mckenzie/nickersonConfirmationBias.pdf (University of California, San Diego)

\n

[3] Baron, J., Hershey, J. C., \"Outcome Bias in Decision Evaluation\", Journal of Personality and Social Psychology, vol. 54, no. 4, 1988, pp. 569-579. Author's copy available at: https://www.sas.upenn.edu/~baron/papers/outcomebias.pdf (University of Pennsylvania; archived copy: https://web.archive.org/web/2020/https://www.sas.upenn.edu/~baron/papers/outcomebias.pdf)

\n

[4] Arkes, H. R., Blumer, C., \"The Psychology of Sunk Cost\", Organizational Behavior and Human Decision Processes, vol. 35, no. 1, February 1985, pp. 124-140, DOI 10.1016/0749-5978(85)90049-4 (canonical reference verified on Crossref). The publisher does not offer open access and no copy on an academic domain could be found: the full text consulted, with the original journal pagination and header, is hosted by a private archive — https://www.communicationcache.com/uploads/1/0/8/8/10887248/the_psychology_of_sunk_cost.pdf (archived copy: https://web.archive.org/web/20260918070345/http://www.communicationcache.com/uploads/1/0/8/8/10887248/the_psychology_of_sunk_cost.pdf)

\n

[5] Treccani, entry \"bias\", Neologismi (2024), Istituto della Enciclopedia Italiana. Available at: https://www.treccani.it/vocabolario/neo-bias_(Neologismi)/

\n

[6] ISTAT, \"Censimento permanente delle imprese 2023: primi risultati\", press release, November 14, 2023 (reference year 2022; about 280,000 responding companies, representative of 1,021,618 companies with at least 3 employees). Available at: https://www.istat.it/comunicato-stampa/censimento-permanente-delle-imprese-2023-primi-risultati/ — PDF: https://www.istat.it/it/files/2023/11/REPORTCensimprese.pdf

\n

[7] Baltrunaite, A., Formai, S., Linarello, A., Mocetti, S., \"Ownership, governance, management and firm performance: evidence from Italian firms\", Questioni di Economia e Finanza (Occasional Papers) no. 678, Banca d'Italia, March 2022, DOI 10.32057/0.QEF.2022.0678. Available at: https://www.bancaditalia.it/pubblicazioni/qef/2022-0678/index.html — PDF: https://www.bancaditalia.it/pubblicazioni/qef/2022-0678/QEF_678_22.pdf

","headings":[{"level":2,"text":"Telling an occasional mistake from a cognitive bias","id":"telling-an-occasional-mistake-from-a-cognitive-bias"},{"level":2,"text":"Recognizing the signs of confirmation bias in a meeting","id":"recognizing-the-signs-of-confirmation-bias-in-a-meeting"},{"level":2,"text":"The first number spoken in a meeting weighs more: anchoring","id":"the-first-number-spoken-in-a-meeting-weighs-more-anchoring"},{"level":2,"text":"Carrying on so as not to waste: the sunk cost bias","id":"carrying-on-so-as-not-to-waste-the-sunk-cost-bias"},{"level":2,"text":"Evaluating a decision without looking at how it turned out","id":"evaluating-a-decision-without-looking-at-how-it-turned-out"},{"level":2,"text":"Why countermeasures to cognitive biases stop working","id":"why-countermeasures-to-cognitive-biases-stop-working"},{"level":2,"text":"Limitations and conditions of applicability","id":"limitations-and-conditions-of-applicability"},{"level":2,"text":"FAQ","id":"faq"},{"level":3,"text":"What is confirmation bias?","id":"what-is-confirmation-bias"},{"level":3,"text":"How many types of bias are there?","id":"how-many-types-of-bias-are-there"},{"level":3,"text":"What are some examples of cognitive biases in work decisions?","id":"what-are-some-examples-of-cognitive-biases-in-work-decisions"},{"level":3,"text":"Is knowing about biases enough to avoid them?","id":"is-knowing-about-biases-enough-to-avoid-them"},{"level":2,"text":"Key takeaways","id":"key-takeaways"},{"level":2,"text":"Conclusion","id":"conclusion"},{"level":2,"text":"Sources and references","id":"sources-and-references"}],"tldr":"In a meeting, a proposal is approved in twenty minutes and the question of what might disprove it is never asked: is that consensus or a flaw in the procedure?","tldrItems":null}