Every leader reading this can already name the traps. Sunk cost. Anchoring. Confirmation bias. The vivid recent example that distorts everything. The vocabulary has been in general circulation for forty years.
They fall into all of them anyway, and at roughly the same rate as people who have never heard the terms. This is the fact that most writing on the subject skips past, and it is the only fact that changes what you should do.
Awareness is not a countermeasure
The reason is structural rather than a matter of insufficient effort. These errors do not announce themselves as errors. From the inside, sunk cost reasoning feels like commitment and follow-through. Confirmation feels like having done the reading. The anchor feels like a reasonable starting point that everyone agreed to.
You cannot catch, by introspection, a process that produces no distinctive sensation. Which means the entire genre of advice consisting of “be aware of your biases” is close to useless, and the useful interventions all have a common shape: they change the procedure or the sequence in which information arrives, so that the trap has no opportunity to operate. None of them require anybody to be more vigilant.
Sunk cost
Money and time already spent are gone regardless of what you do next, and should therefore have no weight in the decision. Everyone knows this. Almost nobody acts on it, because abandoning a project means publicly converting an investment into a loss, and the person who must announce that is usually the person who authorised it.
Two things work. The first is a question, and it has to be asked by somebody with no stake in the answer: if we were not already in this, and the money were sitting in the account today, would we start it now? The question is trivial and the honesty required to answer it is not, which is why the asker matters more than the wording.
The second is stronger and has to be done at the beginning. Write the abandonment conditions into the project at approval. If this milestone slips past this date, or the cost exceeds this figure, the project stops and does not require anyone to argue for stopping it. Deciding while you are disinterested is the only reliable protection against deciding while you are committed.
Anchoring
The first number spoken in a room owns the room. Everything afterward is an adjustment from it, adjustments are consistently too small, and the effect operates even when everyone present knows the first number was arbitrary.
This is the easiest of the four to defeat and the one most often left in place, because the fix costs four minutes and looks slightly odd the first time. Before any number is said aloud, everyone who has a view writes their estimate down. Then they are read out. That is the whole intervention.
The value is not only accuracy. A group that has written first reveals its actual spread, and the spread is information. Six estimates clustered tightly means something. Six estimates ranging across a factor of four means the group does not know, and would never have discovered this had the senior person spoken first and watched everyone converge on his figure within ten minutes.
Confirmation
Having formed a view, you will find supporting evidence more readily, weigh it more heavily, and interrogate contrary evidence far more rigorously than the evidence you like. The asymmetry is not in what you find. It is in how hard you looked and how hard you pushed.
Inviting disagreement does not fix this, and leaders who invite it and receive none conclude wrongly that there was none. Disagreement carries a cost to the person offering it, the cost rises with the seniority of the person being disagreed with, and no amount of encouragement changes the arithmetic.
What works is assignment. A named person is given the job of building the strongest case against the proposal and presenting it, and everyone understands it is an assignment rather than an opinion. This removes the cost entirely, because the person is not disagreeing with you, they are doing what you asked. Rotate who holds the role, so it does not become a personality. The output is frequently thin and occasionally saves a great deal of money.
Availability
What comes to mind easily feels more probable than it is. A recent event, a vivid one, or one that happened to somebody you know will distort the estimate badly and in a direction you cannot feel.
A single lawsuit will reshape an organization’s entire risk posture for four years. One bad hire from a particular background quietly closes off a whole category of candidate. One competitor’s failure in a market becomes proof the market is impossible. In every case the sample size is one, and in every case it is operating with the force of a general finding.
The countermeasure is to ask for the rate rather than the instance. How often has this actually happened, across how many attempts, over what period? The answer is often available and almost never looked up, because the vivid case feels like sufficient evidence and looking it up feels redundant.
A warning about the vocabulary
One further caution, because this knowledge has a common misuse and it does real damage in meetings.
Naming a bias in somebody else’s argument is not a refutation. “That’s just confirmation bias” is a rhetorical move that requires no engagement with the substance and is nearly impossible to answer, since any defence can be characterised as further confirmation. People who have recently learned this vocabulary use it this way constantly, and the effect on a room is to make disagreement more expensive rather than less.
The material in this entry is for use on your own procedures. Applied outward as a diagnostic on colleagues, it becomes a way of dismissing people while sounding rigorous, and an organization that has learned to argue this way has acquired the language of careful thinking and lost the practice of it.
Edited by Patrick J. Wolf, PhD