1.1.1—Bias

Bias is a consistent tendency for thinking, measurement or evidence to lean in one direction. It is not simply a random mistake: some feature of the person, method or context repeatedly makes one interpretation, group or outcome more likely.

Bias can enter at different stages:

• cognitive bias changes how information is noticed or interpreted; an initial number can anchor later estimates
• researcher, participant or sampling bias changes how evidence is produced or who is represented
• cultural, gender or publication bias changes which experiences and findings become visible

Suppose a researcher expects a memory strategy to work. If ambiguous responses are more often scored as correct in the treatment group, the expectation has influenced measurement. A concrete anchoring study found that people given a low starting sequence made a much smaller estimate than people given a high starting sequence, even though both sequences had the same final answer. If the sample also comes from one highly selective school, the evidence may not represent other learners. Neither problem requires dishonesty; bias can operate without awareness.

A surprising result is not automatically biased. To justify that judgement, identify the mechanism that systematically favours one interpretation, group or outcome. The anchoring example is one form of cognitive bias, not a definition of bias as a whole.