Motivated reasoning
Motivated reasoning is reasoning that starts from the conclusion you want and works backward to a justification for it, while feeling from the inside exactly like ordinary careful thought.
What motivated reasoning is
Motivated reasoning is the use of real reasoning skill in the service of a predetermined answer. The reasoner is not lying and is usually not aware of doing anything unusual: the evidence that supports the wanted conclusion is accepted at face value, and the evidence against it is examined until a flaw is found. Since every real study, forecast and dataset has flaws, the search almost always succeeds.
What distinguishes it from the rest of the cognitive bias family is the presence of a goal. Confirmation bias can run on a belief you have no stake in at all. Motivated reasoning needs something you want to be true: a decision you already announced, a product you already bought, a candidate you already recommended. The psychologist Ziva Kunda, in a 1990 review in Psychological Bulletin, set out the constraint that makes the whole thing work: people arrive at the conclusions they want, but only if they can construct a case that would look reasonable to a fair observer. The desire sets the destination, and the reasoning has to build a road that passes inspection.
The one sentence version, and the asymmetry that gives it away
In one sentence, motivated reasoning applies two different standards of proof to two sides of the same question, and the asymmetry, not the conclusion, is the diagnostic sign. A conclusion reached by motivated reasoning may well be correct. That is what makes the bias so hard to catch by looking at answers.
The asymmetry shows up as two silent questions. For evidence you like, the question is whether you are permitted to believe it. For evidence you dislike, the question is whether you are compelled to believe it. Those are very different bars, and moving between them takes no conscious effort at all.
- Welcome evidence is read once, accepted, and filed.
- Unwelcome evidence is read twice, checked for methodology, and filed under pending.
- Welcome sources are described by what they found.
- Unwelcome sources are described by who funded them.
The evidence: unwelcome results get more scrutiny, not less
The clearest experimental signature of motivated reasoning is a difference in processing time, and Peter Ditto and David Lopez reported it in 1992 in the Journal of Personality and Social Psychology. Participants who received an unfavorable result from a test looked for reasons the test might be unreliable, took longer over it, and were more willing to repeat it, while participants who received a favorable result from the same procedure accepted it quickly and moved on. Nobody in the unfavorable group refused the result. They simply held it to a standard the favorable group never had to meet.
A second line of evidence comes from the "consider the opposite" work of Charles Lord, Mark Lepper and Elizabeth Preston, published in 1984. Simply telling people to be fair and unbiased did not change their evaluations. Instructing them to ask specifically whether they would have judged the study the same way had it come out the other way did change them. That result is the single most useful thing known about this bias: the general instruction fails and the specific one works.
Where motivated reasoning shows up in ordinary work
Motivated reasoning appears wherever a person has already committed to something in public, such as a forecast, a hire, a purchase, a design or a published number. Three worked examples, none of them exotic:
- The supplier trial. A purchasing manager who recommended a new supplier runs a six month trial. Defect rates rise slightly. The manager notes that the trial period included two weeks of unusual demand and proposes extending it. The extension may be reasonable. The test is whether the manager would have proposed extending a trial that came back slightly better than expected.
- The sales forecast. A team forecasts 4,000 units and sells 2,900. The shortfall is attributed to a competitor's discounting, which is real. The same team, when it beat its forecast the previous year, attributed that to its own campaign, and did not mention that the competitor was out of stock for a month.
- The exam results. A school changes its math teaching and its pass rate rises. The report highlights the rise and mentions in a footnote that the entry cohort was stronger. Had the pass rate fallen, the cohort would have been in the headline.
In all three, the reasoning offered is not invalid. Nothing here is a formal error of the kind that a deductive reasoning check would catch. The premises are true and the inference is respectable. What is wrong is the selection: the reasoner searched for exactly one kind of explanation and stopped when it was found.
What motivated reasoning is confused with, beginning with confirmation bias
Motivated reasoning is confused most often with confirmation bias, and the two are not the same. Confirmation bias favors what you already believe; motivated reasoning favors what you want to be true. They overlap constantly, because people usually want their beliefs to be right, but they come apart in a way that matters for the fix.
| Bias | Driven by | Comes apart when |
|---|---|---|
| Confirmation bias | An existing belief | You have a belief with no stake in it, and still favor it |
| Motivated reasoning | A wanted outcome | You want something you currently believe is unlikely |
| Self serving bias | Protecting your own standing | The credit and blame land on you specifically |
| Rationalization | An action already taken | The reasoning is constructed entirely after the fact |
Motivated reasoning is also confused with lying, and it is not lying. The reasoner believes the case they are making. That is why arguing against it with better evidence so often fails: better evidence gives the motivated reasoner more material to work with, and a person with more knowledge of a subject can build a more sophisticated defense of the answer they wanted. Expertise raises the quality of the road, not the choice of destination, which is one of the few places where the pattern behind the Dunning Kruger effect runs in reverse.
One question tells the three apart: if the answer made no difference to me, would I still be reading the evidence this way? If the answer is no, the driver is something you want, and it is motivated reasoning. If the answer is yes, and you would still favor the evidence that fits, it is confirmation bias. If what is being protected is specifically your own record rather than the conclusion, it is self serving bias.
What actually reduces motivated reasoning
What reduces it is structure applied before the evidence arrives, because after it arrives you are no longer a neutral party. Five moves, in rough order of effectiveness:
- Write down, in advance, which result would count as a failure and what you would do about it.
- Ask whether you would accept this same evidence if it pointed the other way.
- Give one person in the room the explicit job of building the case against.
- Separate the person who ran the analysis from the person who decides what it means.
- Expect to explain your reasoning to someone whose own view you do not yet know.
That last item comes from the research on accountability reviewed by Jennifer Lerner and Philip Tetlock in 1999: knowing you will have to justify a judgment to an audience whose position is unknown tends to produce more even handed thinking, while knowing the audience's position tends to produce a defense of it. The design of the accountability matters more than the amount.
Be honest about the ceiling. None of these moves eliminates the bias, and the person best placed to apply them is the person least able to see they are needed. Structure works because it does not rely on the reasoner noticing anything.
The test to run before your next conclusion
Ask the question in this form, and ask it out loud:
- What result would have made me change my mind, and did I decide that before or after I saw the data?
- Am I asking whether I can believe this, or whether I must?
- If the numbers were reversed, would I be criticizing the method or citing it?
If you cannot answer the first question with something concrete, you did not have a hypothesis. You had a preference.
Motivated reasoning in psychology: directional goals and the parent concept
In psychology, motivated reasoning is the parent concept, and confirmation bias is one of the mechanisms through which it operates. That is the opposite of the popular ordering, where motivated reasoning is treated as a variety of confirmation bias, and getting it the right way round makes the literature much easier to read.
The standard vocabulary comes from Ziva Kunda's 1990 review in Psychological Bulletin, which sorted the research into two kinds of goal. An accuracy goal makes a person think harder and more evenly, because being right is what they want. A directional goal makes a person think harder in one direction, because a particular answer is what they want. The same effort, the same intelligence, two different destinations. Confirmation bias, selective recall, and skeptical scrutiny of unwelcome evidence are then the tools a directional goal reaches for.
The research tradition is social psychology, specifically the study of motivation and cognition, and it borders judgment and decision making on one side and attitude research on the other. How firmly established is it? The core finding is solid: people apply more scrutiny to conclusions they dislike than to conclusions they like, and this shows up in processing time and in the reasons they generate. What is not established is the strong popular version in which people simply believe whatever suits them. Kunda's own constraint says otherwise: a directional goal can only deliver a conclusion the reasoner can build a plausible case for, which is why evidence still matters and why the reality constraint is the most useful part of the theory to remember.
The backfire effect, and why it is weaker than its reputation
The backfire effect is the claim that correcting a false belief can strengthen it, and the honest current answer is that it is real in some conditions but rare, and much weaker than the popular version suggests. Brendan Nyhan and Jason Reifler described the pattern in 2010 and gave it the name, and it spread quickly, partly because it is a satisfying story about other people.
What came next matters more. Later work using many more corrections and much larger samples, including studies by Thomas Wood and Ethan Porter, generally found that corrections moved people toward the facts rather than away from them, even on beliefs they held firmly. Nyhan himself has written since that the effect is not a general rule. Backfire effects in psychology are now best described as a possible outcome under particular conditions, not as the reason to give up on correcting a false claim.
The practical consequence is worth stating plainly, because the popular version of the backfire effect has been used to argue that evidence never changes minds. That conclusion does not follow from the research, and believing it is comfortable in exactly the way this page has been describing.