Cognitive biases are systematic ways our judgment deviates from rational thinking, mental shortcuts that speed decisions but introduce predictable errors. In UX they cut both ways: users have them, so design has to account for how people actually decide, and designers have them, so research has to guard against skewed conclusions.
Use cases
Ignore biases and you design for a rational user who doesn't exist. Miss your own and you ship conclusions your data never supported. Both quietly wreck products.
Anchoring on price: Show a $100 plan next to a $70 one and the $70 reads as a deal, even if it's steep on its own. Pricing pages live or die on this.
The framing trap in research: "16 of 20 users found it" and "4 of 20 failed" describe the same test, but NN/g found designers push for a redesign far more often when they hear the failure framing.
Confirmation bias in a study: A researcher who expects a feature to test well writes leading questions and remembers the praise, not the confusion. The report confirms the hypothesis, and the flaw ships.
How it's used in practice
Design with the grain: Social proof, sensible defaults, and honest anchors reduce effort and help people decide. Work with how minds actually run.
Draw the ethics line: Helping someone decide is good; manufacturing false urgency or hiding the opt-out is a dark pattern. Same psychology, opposite intent.
Neutralize your research framing: Present findings as raw counts and both framings so the numbers, not the wording, drive the call.
Bring in other eyes: Diverse reviewers and a devil's advocate catch the blind spots your own biases hide.
🪄
Pro-tip: The bias most likely to burn you is your own. Confirmation bias runs hottest when you're excited about your design, exactly when you most want the research to say yes. Go in trying to prove yourself wrong, and treat easy agreement as a warning sign rather than a win.
Challenges & limitations
You can't switch them off: Biases are how brains save energy. Awareness helps, but nobody debiases fully, so build process around it instead of relying on willpower.
The exploitation slope: The same knowledge that helps you guide users can be turned against them. Persuasion and manipulation share a toolkit; only intent separates them.
Too many to memorize: There are dozens of named biases with fuzzy overlaps. Chasing the full list matters less than knowing the handful that hit your product and your process.
