Cognitive bias in user research
I keep returning to this topic because user research is supposed to reduce guesswork, yet the process can still be shaped by what we expect, notice, remember, and choose to ask.
What stayed with me most is that bias is not a rare research mistake. It can sit inside very normal-looking decisions: who I invite to an interview, how I phrase a question, which answer feels important in the moment, and which quote I later choose as evidence.
That is why I do not see bias as something to check only at the end. I would rather treat it as part of the research setup itself, because once a team turns a biased read into product confidence, the decision can feel far more solid than it really is.
The biases I keep watching for
These are the traps I would keep closest to the research plan, because each one can quietly bend a session in a different direction:
- Confirmation bias
- It pushes me to notice the feedback that supports the idea I already like and to explain away the feedback that makes it weaker. In product work, this is especially dangerous because teams often enter research with a preferred direction already in mind.
- Peak-end rule
- A participant's strongest moment, or the way a session ends, can become louder in memory than the full experience. If I rely only on what felt most intense, I can overvalue an emotional quote and miss the quieter repeated pattern.
- Observer-expectancy bias
- This is about the influence I bring into the room. Tone, facial expression, follow-up questions, or even small confirmations can tell people what kind of answer I expect.
- Anchoring
- An early number, example, or assumption can set the frame for everything that follows. Once that anchor is there, later answers may orbit around it instead of starting from the user's own reference point.
- Order effect
- The sequence of questions can prime people, make later answers easier or harder to give honestly, or make one topic feel more important than another.
How I try to keep it grounded
The practical lesson for me is to design the research process against my own assumptions: write neutral questions, avoid leading examples too early, separate observation from interpretation, and make space for evidence that contradicts the first idea.
I also trust synthesis more when it is collaborative. When more than one person reviews the same raw notes, hidden assumptions are easier to catch. The goal is not to remove every bias completely, but to slow down the jump from user signal to product decision.
The bigger thought I took from the article is simple: research is not just a method, it is a discipline of attention. The quality of the outcome depends on how carefully I protect the process from the shortcuts my mind naturally wants to take.