A vivid research session can become the story everyone remembers. A quieter set of contradictory observations may struggle to compete, particularly when the memorable session supports a decision the team already favours. Cognitive biases are systematic tendencies in attention, memory and judgement that can shape how people interpret information.
In research, they matter on both sides of the conversation. Participants bring expectations to an experience, and researchers bring expectations to the evidence. Knowing the names of biases can help identify a concern, but it does not establish that a particular person or finding is biased.
Examine the researcher’s interpretation first
Confirmation bias involves favouring information that supports an existing belief. Nielsen Norman Group’s discussion of confirmation bias in UX describes its relevance to research design and interpretation. A team convinced that navigation is the problem may ask questions, notice incidents and select quotations that keep the explanation centred on navigation.
Availability can make a recent or memorable incident feel more common than it is. Anchoring can make the first interpretation in a meeting disproportionately influential. Framing can change how the same result is understood, such as emphasising successful attempts while leaving the failures and their consequences unexamined.
These are reasons to inspect the process, not convenient labels for dismissing disagreement. A colleague may remember an unusual session because it exposed a serious failure, and the seriousness may justify action even if the event was uncommon.
Use procedures that make alternatives visible
Write down research questions, consequential measures and planned comparisons before inspecting results. Record later changes as changes, with their reasons. Exploratory work is legitimate, but it should remain distinguishable from an analysis planned in advance.
Review the full relevant dataset rather than relying on memory or a selection of clips. Seek cases that challenge the emerging interpretation and ask what other explanation could produce the observation. Independent initial notes can help a team compare readings before a group discussion settles around one account.
For a fictional checkout study, repeated hesitation near delivery options might reflect price, terminology or uncertainty about arrival dates. Coding every pause as “price resistance” would make the analysis reproduce the initial belief. Returning to the words, actions and conditions of each attempt helps test that interpretation.
Keep judgement accountable rather than pretending to remove it
Structured methods still involve choices about categories, significance and practical consequences. A second reviewer can challenge those choices, but agreement between reviewers does not make a conclusion objectively certain. Preserve the evidence and reasoning so others can assess them.
Response bias provides a closer account of influences on participants’ answers, while research bias covers systematic problems across the study. Thematic analysis can help organise interpretation when used with an approach appropriate to the research question.
The useful outcome is a conclusion that has survived relevant alternatives and accurately expresses its uncertainty. The goal is not to produce a team that considers itself immune to bias, because that confidence would make the next difficult interpretation less likely to receive scrutiny.