Bias

A tendency to see or report things in a way that is not fully objective — a tilt away from the truth.

Judging based on facts, not on personal feelings or expectations.

When the people studied don't represent the wider group you want to describe.

When a researcher's expectations shape how they run or interpret a study.

When people change their behaviour because they know they are being studied.

Noticing evidence that fits what you expect and ignoring evidence that doesn't.

When mostly positive results get published, hiding the studies where nothing happened.

It threatens objectivity, so the findings become less trustworthy.

Clues in a study that let participants guess the aim, so they behave to fit it.

Three places: the design of the study, the researcher, and the participants.

Giving the answer that makes you look good rather than the true one. It is a kind of response bias.

Misremembering the past without meaning to, so the answers are inaccurate.

A question whose wording pushes people towards one answer.

When the people who do not take part differ from those who do, so the sample is tilted.

Implicit: the person does not notice it. Explicit: the person is aware of it.

Researcher bias in an interview: the interviewer's words or reactions steer the answers.

People perform better or try harder simply because they know they are in a study.

Another name for participant bias: what participants think the study wants changes how they act or answer.

Judging or measuring one culture by the standards of another.

Seeing your own culture as the normal or correct one, and judging others by it.

Studying a culture from the inside, on its own terms and in its own concepts.

Comparing cultures from the outside using shared, general measures.

Using a method built in one culture on another as if it were neutral for everyone.

Diagnosis, prevalence statistics and treatment.

Use emic methods, local researchers, and translate then back-translate the measure.

Translating a measure into a language and back again to check the meaning survived.

Western, Educated, Industrialised, Rich, Democratic — over-used and wrongly treated as universal.

An etic approach compares cultures with a shared measure, which science needs. Bias comes when the measure is imposed without checking that it fits.

They count diagnoses. If symptoms are expressed differently or not recognised, a disorder looks rarer or more common than it is.

The clinician reads the patient's behaviour against her own cultural norms, and the symptom lists were written mostly in Western countries.

How common a disorder is in a population.

Research treating one gender unfairly — as better, worse, or the default.

Exaggerating the differences between genders, often locking in stereotypes.

Ignoring or minimising real gender differences, often by studying one gender and applying it to all.

Treating male behaviour as the normal standard for everyone.

Testing a heart drug only on men, then giving it to everyone at the same dose.

Diagnosis, prevalence statistics and treatment.

Use a balanced sample of all genders and report results for each group.

The same problem shows differently in men and women, and clinicians read it through their expectations of each gender.

'Treating everyone the same' ignores real differences, which can quietly disadvantage the untested group.

Alpha exaggerates gender differences; beta ignores them.

A recorded rate mixes how many people have a disorder with how readily it is noticed in them.

No. A difference the evidence supports is a finding. Assuming a difference, or assuming none, is the bias.

Protecting objectivity, so findings reflect the truth and can be trusted.

It makes the sample mirror the wider group, so results generalise — fixing sampling bias.

Giving every participant the same instructions and conditions, so differences don't creep in.

Neither the participant nor the tester knows the condition — reduces participant and researcher bias.

A researcher's awareness of her own views and values, and of how they may influence the research.

Others repeat the study; a one-off, biased result usually fails to repeat.

Emic methods, local researchers, and translating then back-translating the measure.

A balanced sample of all genders, with results reported for each group.

Posting the plan before data collection — reduces publication bias, so failures can't vanish.

Match the fix to the bias, and explain why it restores objectivity.

A small trial run that finds confusing instructions and leading questions before the real study.

Two or more researchers collect or analyse the data, so one person's bias is checked by another.

Sampling decides who is in the study. Allocation decides which group each person joins.