Hawthorne Effect

The Hawthorne effect is the change in people's behavior that happens because they know they are being observed or studied. Workers may work harder during an audit, patients may follow advice more closely while in a trial, and users may try harder in a usability test. Researchers need to account for it because it can make an intervention look effective when the real cause was the attention of the study itself.

The effect is a form of reactivity: measurement changing the thing being measured. It matters in experimental research, field studies, workplace programs, clinical trials and UX research.

Where the name comes from

The name refers to studies run from 1924 to 1932 at Western Electric's Hawthorne Works, a factory near Chicago. Researchers first changed lighting levels to see how output responded, then ran longer experiments with small groups of workers in separate test rooms, changing breaks and working hours. Output often went up whatever was changed, and the researchers, including Elton Mayo, Fritz Roethlisberger and William Dickson, credited the attention and the social setting of the experiments. The term "Hawthorne effect" was coined later, by Henry Landsberger in 1958.

The original evidence is weaker than its fame suggests. Economists Steven Levitt and John List re-examined the surviving lighting data in 2011 and found little support for the classic story. A 2014 systematic review by Jim McCambridge and colleagues concluded that research participation can change behavior, but that the size and conditions of the effect vary widely. Today the term is used as a general label for observation effects, not as a precise, constant law.

Examples of the Hawthorne effect

• Workplace: a team's error rate drops during a "quality month" with extra supervision, then returns to normal.
• Healthcare: hand-hygiene compliance can look higher when staff can see an auditor on the ward than when it is measured unobtrusively.
• Customer service: call quality scores rise in weeks when agents know calls are being reviewed.
• UX research: participants in a moderated usability test persist longer on a task than they would at home, so the product seems easier than it is.
• Surveys: employees who believe their answers can be traced back to them give more positive feedback.

How it can mislead a study

Example. A company introduces a new checklist for warehouse pickers and assigns a researcher to watch the first two weeks. Picking errors fall from 2.0% to 1.2%. Three months later, with no one watching, errors are back at 1.9%. Most of the early improvement came from being observed, not from the checklist.

• False optimism: results improve while the study runs and fade afterward.
• Wrong cause: the improvement is credited to the new tool, training or process.
• Weak generalization: results from an observed setting do not carry over to everyday conditions.

How to reduce it

• Use a control group observed in the same way. With random assignment, both groups are equally aware of being studied, so the difference between them reflects the intervention. Randomization keeps the effect from favoring one group, although it does not remove it.
• Measure a baseline before the study is announced, and keep measuring after observation becomes routine.
• Run the study longer. People get used to being observed, and the early spike flattens.
• Prefer unobtrusive measures such as existing logs or system data, within ethical and legal limits and with appropriate consent.
• Reduce evaluation pressure in surveys. Explain how answers are used, collect them anonymously where possible, and avoid wording that sounds like a performance review.

Related effects

• Placebo effect: improvement caused by believing you received a treatment.
• Observer-expectancy effect: the researcher's expectations influence participants or how results are recorded.
• Demand characteristics: participants guess the study's purpose and act to fit it.
• Social desirability bias: respondents give answers that make them look good.

Common mistakes

• Judging a new policy only by the first weeks after launch, while attention is high.
• Comparing an observed pilot group with unobserved historical data.
• Assuming the effect always improves results. Observation can also make people nervous and perform worse.
• Presenting the Hawthorne story as settled fact. Cite it as an illustration of reactivity.