Dr. Wallace's Favorite Mental Models

Motivated Reasoning

Why smart people defend conclusions they want to be true — and how to catch yourself doing it.

130-Second Overview

The fast version, if that's all you need right now.

What it is: Motivated reasoning is the tendency to process information in a way that is biased toward a conclusion you want to reach, rather than the conclusion the evidence actually supports.
Why it happens: Psychologist Ziva Kunda showed in 1990 that reasoning serves two different goals. An accuracy goal pushes you toward whatever conclusion the evidence supports. A directional goal pushes you toward a specific, preferred conclusion. Both goals use real reasoning strategies, so the person doing it rarely feels biased.
The catch: You cannot just believe anything you want. Kunda's key finding was that a desired conclusion still has to be supported by a case that feels reasonable. Motivated reasoning works by searching harder for the evidence, memories, and standards that build that case, not by ignoring the need for one.
Why it matters at work: It explains why a team keeps defending a strategy after the market has turned, why a hiring manager "discovers" reasons to like the candidate they met first, and why a leadership team's own sunk cost in a decision quietly reshapes what evidence they're willing to call credible.

2Mental Model Compass

Where Motivated Reasoning sits next to three models already in this collection that are easy to confuse with it. Click a dot.

Mental Model Compass A scatter plot comparing four related mental models on two axes: how actively the person seeks out supporting information, and whether the underlying goal is accuracy or a preferred direction. Directional goal → ← Accuracy goal Actively seeks evidence ↑ Passively reacts ↓ Motivated Reasoning Confirmation Bias Cognitive Dissonance Self-Serving Bias
Click any dot on the compass to see how that model relates to Motivated Reasoning.

3Progressive Interactive Learning

Four short stages, each building on the last. Work through them in order the first time.

Ziva Kunda's 1990 paper The Case for Motivated Reasoning (Psychological Bulletin, 108(3), 480–498) made a precise claim: motivation does not let people believe whatever they like. It changes which reasoning strategy they reach for.

An accuracy goal makes you reach for whatever strategy you think is most likely to get the right answer, even if the answer is unwelcome. A directional goal makes you reach for whatever strategy is most likely to get you to a specific answer you already want.

Because it means the fix is not "try harder to be unbiased" in the abstract. It is noticing which goal is actually running the search: are you asking "what's true here?" or quietly asking "how do I justify what I already decided?" Those two questions pull from different evidence, even when you are not aware of switching between them.

Charles Taber and Milton Lodge tested this directly with partisans evaluating political arguments (Motivated Skepticism in the Evaluation of Political Beliefs, American Journal of Political Science, 50(3), 2006, pp. 755–769). People with strong, well-informed views:

  • Rated arguments that agreed with their position as stronger than arguments that disagreed — even when told to be objective.
  • Spent more time picking apart the arguments they disagreed with (a "disconfirmation bias") than evaluating the ones they liked.
  • When allowed to choose their own sources, chose like-minded sources 70 to 75 percent of the time.
After reading a balanced mix of arguments for and against their position, the most informed and most committed participants did not move toward the middle. They became more extreme than before. More information, filtered through a directional goal, produced more polarization — not less. Weakly committed, less knowledgeable participants did not show this effect.

Motivated reasoning is not an abstract lab curiosity. Wikipedia's documented cases include smokers who focus on information that makes smoking seem less dangerous while discounting the evidence against it, and people who reject masking or vaccination evidence in favor of misinformation that fits what they already wanted to believe.

A 2024 UK study found that people who received either the Moderna or the Pfizer COVID-19 vaccine tended to rate their own vaccine as the safest option — and a meaningful share misremembered having preferred that brand all along, when in most cases they had simply been offered whichever vaccine was available. The preference came first; the "reasons" for it were constructed afterward.

The mechanism hasn't changed, but the environment feeding it has. Recommendation algorithms learn what keeps you engaged, which is often exactly the content that confirms what you already lean toward. That turns a psychological tendency into a technically reinforced loop: the directional goal and the information supply now reinforce each other automatically, with no deliberate choice required on your part.

An AI assistant that is optimized to produce answers you'll approve of can become a motivated-reasoning amplifier: it will find a plausible-sounding case for almost any conclusion you steer it toward. The discipline this model teaches — separate "what do I want to be true" from "what does the evidence support" — applies directly to how you prompt and evaluate AI output, not just to how you read the news.
Stage 1 of 4

4Diverse Examples

The same mechanism, four different settings.

A VP championed a reorganization eighteen months ago. Engagement scores have slipped since, and two senior managers have quietly left. In the quarterly review, the VP highlights the two teams whose output improved and attributes the departures to "unrelated personal reasons" without checking exit interview notes that say otherwise. The VP isn't lying. The directional goal — "my reorg worked" — is deciding which data points feel relevant enough to mention.

An investor bought a stock at $80. It's now at $45. Every piece of bad news gets reframed as "temporary" or "already priced in," while a single analyst upgrade gets forwarded to three friends as vindication. The original $80 purchase, not the current evidence, is quietly setting the bar for what counts as convincing.

Two people look at the same economic data release. One, who favors tighter immigration policy, reads it as confirming that immigration suppresses wages. The other, who favors looser policy, reads the same release as confirming the opposite. Both can cite real numbers from the report. Taber and Lodge's finding explains why: each spent more effort stress-testing the interpretation they didn't want, and less effort stress-testing the one they did.

Someone trying a new diet reads every success story in detail and skims past articles describing rebound weight gain as "those people probably didn't follow it properly." The desired conclusion — "this will work for me" — is shaping how much scrutiny each piece of evidence receives, not whether evidence is allowed in at all.

5Worked Application

A six-step check you can run on yourself before a decision you already feel strongly about.

  1. Name the preferred conclusion out loud. Before reviewing the evidence on, say, whether to keep funding a stalled product line, write down in one sentence what you already hope the answer turns out to be.
  2. Ask which goal is actually running. Are you trying to find out whether the product line is viable, or trying to build a case for the decision you've already half-made?
  3. Time-check your scrutiny. Notice whether you're spending more time finding holes in the evidence against your preferred answer than in the evidence for it. Taber and Lodge's "disconfirmation bias" shows up as unequal effort, not as ignoring one side outright.
  4. Find someone with the opposite preference and ask them to make the case. Not to "win" against them — to see which evidence they treat as central that you treated as noise.
  5. Pre-commit to a standard before you look at new evidence. Decide in advance what result would actually change your mind, and write it down before you start, so the standard can't quietly move afterward.
  6. Separate the decision from the evaluation of the person who made it. If you were the one who championed the original choice, that is exactly when directional motivation is strongest — the goal quietly shifts from "is this working" to "was I right."

6Common Mistakes

Treating it as someone else's problem

The most common error is using "motivated reasoning" as a label for why the other side is wrong, while assuming your own conclusions were reached cleanly. Kunda's and Taber and Lodge's findings apply to everyone with a stake in the outcome — including the person doing the labeling.

Confusing it with simple dishonesty

Someone engaged in motivated reasoning usually believes their own conclusion completely. It is not lying. Treating it as a character flaw rather than a cognitive pattern makes the conversation defensive instead of useful.

Collapsing it into confirmation bias

Confirmation bias — seeking or favoring evidence that fits an existing belief — is one tool motivated reasoning uses. Motivated reasoning is the broader motive underneath it. Fixing one symptom (actively seek disconfirming evidence) without addressing the underlying goal often just produces a more sophisticated-looking rationalization.

Assuming more information always helps

Taber and Lodge found that balanced information made strongly motivated, knowledgeable people more extreme, not less. Simply handing someone "the facts" can backfire when a directional goal is already engaged.

7When Not To Use It

Don't use it to dismiss disagreement

Saying "you're just motivated-reasoning this" is not an argument and shouldn't substitute for actually engaging with someone's evidence. It can become a way to avoid a case you don't want to answer.

Don't use it to explain every wrong belief

Plenty of mistaken conclusions come from incomplete information, honest misunderstanding, or bad luck, with no self-serving goal involved at all. Reach for motivated reasoning specifically when a conclusion conveniently protects something the person already values, not for every disagreement.

Don't use it where someone has already updated

If a person has genuinely changed their mind in light of new evidence, re-litigating their past view as "motivated" ignores the update and discourages the next one.

9Try It Yourself

Fill in a real decision you're currently leaning one way on. This builds a short, specific prompt you can paste into any AI assistant or use to interview yourself.


    

      
      Copied
    

  






  
10Save as Skill

  
A portable, reusable version of this model you can hand to an AI assistant or your future self.

  

    
Preview

    

    

      
      
      Copied
    


    
Apply to My Problem

    
Describe a live situation and generate a ready-to-paste prompt that applies this model to it.

    
    
    

    

      
      Copied