Amazon's "Are Right, A Lot": You're Reading It Backwards

You see the name and think the principle rewards people who are correct. Confident. Decisive. The person in the room who calls the shot and nails it.

That reading is exactly wrong. Amazon's fourth leadership principle, Are Right, A Lot, is less about being right and more about the process that makes you right over time. The full text gives it away: "Leaders are right a lot. They have strong judgment and good instincts. They seek diverse perspectives and work to disconfirm their beliefs." The second sentence is the one interviewers actually score.

Walk into an Amazon behavioral round with a story about how you trusted your gut and it paid off, and you will probably get a polite "not inclined." That's Amazon for "please leave."

The society if people admitted when they lost an argument, showing a futuristic utopia skyline

Amazon hiring committees if candidates stopped bragging about being right.

What Does "Disconfirm Their Beliefs" Actually Mean?

In 1960, psychologist Peter Wason gave participants a number sequence (2, 4, 6) and told them it fit a hidden rule. Their job: propose new triples, get feedback, guess the rule. Most hypothesized "even numbers increasing by two" and tested only confirming sequences. 8, 10, 12. Correct. 20, 22, 24. Correct. They announced their rule, confident. Wrong.

The actual rule was "any three ascending numbers." The participants never tested a sequence designed to break their theory. They sought confirmation, not disconfirmation, and 80 to 90 percent got it wrong. Not because they were dumb. Because they were human.

This is the cognitive trap Amazon's LP filters for. Confirmation bias is not a personality flaw. It is the default operating system of human reasoning. The principle asks: do you actively fight it, or do you just run your happy-path tests and ship?

Liz Jones, an Amazon Bar Raiser, put it bluntly on Amazon's blog: "This principle is about being wrong a lot, too. It's about having the confidence and good judgment to know when you're right and when you're wrong."

Software engineer tweeting that after decades of experience and teaching in a dozen countries, he asked a colleague for help debugging and the reason was he hadn't hit compile

Decades of expertise, hundreds of thousands of users, and the bug was not pressing a button. Confirmation bias does not care about your seniority.

Bezos Said the Quiet Part Out Loud

During a visit to Basecamp's offices, Bezos did a Q&A where Jason Fried recounted the key line:

"People who are right a lot of the time are people who often changed their minds."

He went further. Consistency of thought is not a positive trait. The smartest people constantly revise their understanding, reconsider problems they thought they had solved, and remain open to contradictions. If you've never publicly said "I was wrong about this," you're either a genius or you're not paying attention. Statistically, it's the second one.

The corollary matters just as much. People who are wrong a lot share a common trait: they obsess over details that support only one point of view. They cannot zoom out. They are, in debugging terms, staring at the wrong file.

This is the intellectual posture Amazon probes for. Not "I was right." The posture is: "I hold strong views, but I treat them as hypotheses that new data can revise."

Which Amazon Behavioral Interview Questions Target This LP?

Amazon interviewers are each assigned two to three LPs to evaluate. If yours drew "Are Right, A Lot," expect questions from these clusters.

The incomplete-data decision:

Tell me about a time when data and analysis were not sufficient and you had to rely on judgment and instincts.

The wrong call:

Tell me about a time you made a bad decision. What did you learn?

The belief revision:

Tell me about a time you changed your mind based on new data or a different perspective.

The diverse-perspectives probe:

Tell me about a time you incorporated a diverse set of perspectives into solving a problem.

Notice the pattern. Half these questions ask about being wrong. The principle called "Are Right, A Lot" tests whether you have a productive relationship with being wrong. You're preparing for a principle about correctness. Amazon is preparing to ask you about your failures.

The Three Signals Interviewers Score

Your interviewer is not listening for a happy ending. They are scribbling notes against three dimensions.

1. Judgment Process, Not Judgment Outcome

A strong answer shows how you arrived at your decision, not just that it worked. Did you identify what data you had and what was missing? Did you name assumptions explicitly? Did you build a validation mechanism before committing?

Dave Anderson, a former Amazon VP and Bar Raiser, described a principal engineer who insisted on creating hypotheses backed by data, instrumenting metrics upfront, and testing with a controlled audience before rollout. She did this even when Anderson offered to skip validation. She demonstrated the principle by refusing to trust her instincts without a system to test them.

Interviewers score the validation system you built, not whether your gut was right. Your gut has a 50/50 track record. Your process is what they can evaluate.

2. Belief Revision Under Pressure

This signal requires showing vulnerability. The interviewer wants a moment where you held a strong position, encountered disconfirming evidence, and changed your mind. Not grudgingly. Not after being overruled. Willingly.

Weak version: "My manager told me to change direction, so I did." That is compliance, not judgment. That is also basically every day at work.

Strong version: "I believed X. I deliberately tested it by doing Y. The results contradicted my assumption. I revised my position." You tested to break your own theory. Like a scientist, except you get paid more and have worse posture.

3. Institutional Learning

Most candidates stop at "I learned from this experience." Amazon wants more. Did you create something that outlives your personal memory? A checklist, a decision framework, a post-mortem template the team still uses?

The principle tests whether you build systems that increase your team's correctness rate over time. "I learned a lot" is not a system. A runbook is a system.

Why This LP Collides With Other Principles

A tension that catches candidates off guard: several Amazon LPs pull in opposite directions.

"Bias for Action" says move fast. "Are Right, A Lot" says validate before committing. "Have Backbone; Disagree and Commit" says defend your position. "Are Right, A Lot" says seek to disconfirm your beliefs. It's like being told to run faster and also to stop and look both ways.

These are calibration tests, not contradictions. Amazon wants people who can hold two things simultaneously: the confidence to make a call with 70 percent of the information (Bezos's threshold from his 2016 shareholder letter) and the humility to keep testing that call after making it.

The resolution depends on decision type. Type 1 (irreversible, one-way doors) deserve deep validation. Type 2 (reversible, two-way doors) deserve speed. The interviewer checks whether you can tell the difference. Most candidates treat every decision like a one-way door. A few treat everything like a two-way door. Both are wrong.

A strong STAR story often shows both: you moved fast on a reversible decision, monitored results, caught a signal your direction was wrong, and pivoted.

How to Build Your STAR Story

Situation and Task (15 to 20 percent of your time): What decision needed to be made? What was at stake? Two to three sentences. Resist the urge to set the scene like a novel.

Action (55 to 60 percent): Walk through four beats:

Your initial position and reasoning. Show a data-informed hypothesis, not a random guess.
The disconfirming signal. What challenged your belief? Did you seek it out, or did it find you? Actively seeking scores higher.
Your belief revision. Name the old belief, name the new one, explain why the evidence was compelling.
The adjusted action. What did you do differently?

Result (25 to 30 percent): State the outcome, quantify if possible, then land on the systemic change.

A skeletal example:

"We were evaluating two architectures for a payments service. I was convinced we needed synchronous design for consistency. Before committing, I set up POCs for both approaches and invited the team's most skeptical engineer to run load tests. Her tests revealed the synchronous approach would hit a latency wall at 3x current volume. I was wrong. We switched to event-driven with idempotency keys. The service launched on time and handled 4x traffic without rearchitecting. I documented the evaluation framework, and the team now runs that comparison for any new service above a traffic threshold."

That story hits all three signals: judgment process, belief revision, and institutional learning.

Five Killers That Sink Your Answer

1. The "I was right" brag. Your instinct was correct and everyone else was wrong. Congratulations. This demonstrates confidence, not judgment. Amazon has plenty of confident people. They are hiring for something harder.

2. Compliance disguised as flexibility. "My manager told me to do it differently, so I did." Following orders is not belief revision. That is payroll, not principle.

3. No disconfirming evidence. You made a decision, it worked, end of story. Where is the moment you tried to break your own hypothesis? If you never stress-tested the plan, the interviewer has nothing to write down.

4. Trivial stakes. Choosing a meeting time or picking a CSS framework does not demonstrate judgment under uncertainty. Unless the CSS framework was a deeply controversial decision at your company, in which case, you may need to leave that company.

5. Learning without a mechanism. "I learned to always consider other perspectives." That is a fortune cookie. Name the specific process change, the document, the checklist. Something concrete that still exists.

This is fine dog meme showing vibe coders after sending AI code to production while everything is on fire

Your STAR story when you skip the "what went wrong" beat and just say everything worked out.

The Bar Raiser Is Listening Differently

In every Amazon loop, one interviewer is a Bar Raiser: a specially trained interviewer from outside the hiring team with effective veto power. For "Are Right, A Lot" specifically, they probe whether you actively sought disconfirming evidence or just happened to encounter it. The first demonstrates the principle. The second demonstrates luck. And Amazon does not hire for luck, despite what it sometimes feels like from the outside.

One number that should focus your prep: roughly 25 percent of candidates who pass Amazon's technical bar are still rejected on behavioral grounds (interviewing.io). The LP round is scored as rigorously as your coding round. Possibly more so, because there's no partial credit.

How to Prep for "Are Right, A Lot" Interview Questions

You do not need 32 separate stories (two per LP). That produces weak answers across the board, like studying for 16 exams the night before each one. Build 10 to 12 strong stories that each map to three or four LPs.

For "Are Right, A Lot," you need at least two:

One where you were right, but only because you validated rigorously. The story should make clear your initial instinct could have been wrong, and you built a system to check it.
One where you were wrong and revised your position. This scores higher. Show the old belief, the disconfirming evidence, the revision, and the durable change.

Both stories should survive probing. Amazon interviewers ask follow-ups like "What data did you use?" and "What was your specific role?" If your story collapses under two follow-ups, it is not ready.

Practice out loud. The behavioral round is a spoken performance. If you have only rehearsed in your head, the first time you say your story to another person will feel completely different. SpaceComplexity runs voice-based mock interviews that score you on structure and communication in real time, so you can pressure-test your LP stories before the real loop.

The Recap

"Are Right, A Lot" tests whether you actively fight confirmation bias through diverse perspectives and disconfirming evidence.
Three scored signals: judgment process, belief revision, institutional learning.
Half the questions ask about being wrong. Prepare a story where you were wrong, revised, and created a systemic improvement.
The Bar Raiser probes for active disconfirmation. Stumbling into the right answer is not the same as building a process to find it.

Your LP stories need the same rigor as your coding interview communication. The write-up your interviewer submits is built from what you said out loud, and a strong "Are Right, A Lot" answer gives them something concrete to quote. For more on how that write-up shapes the committee's decision, see our breakdown of what your interviewer is writing while you think.