Survey ethics: 8 rules for asking without breaking trust
Every survey asks someone for their time and their honest opinion. If that person feels misled, identifiable, or that their answers were used for something else, they won't answer honestly again. Survey ethics is not paperwork: it is what makes the data useful.
I write this from psychology and Human Resources, where I saw engagement surveys stop getting answers because someone promised anonymity and didn't keep the promise. These rules apply to workplace surveys and to research.
This is not legal advice. Data protection laws vary by country. If your survey asks for personal or sensitive data, have the text reviewed by someone who knows the rules that apply to you.
The 8 rules
- People know what they are answering. Who runs the survey, why, how long it takes and what happens to the answers. If there is personal data or it is a study, ask for explicit consent before starting.
- It is truly voluntary. Nobody should gain or lose anything by answering or not. At work, that means not tying bonuses, reviews or time off to the survey.
- Don't promise anonymity if it isn't anonymous. If you ask for a name or email, or the link is personal, the survey is confidential, not anonymous. Say so.
- Ask only for data you will use. Every profile question (age, department, tenure) adds a risk of identifying someone. If you won't analyze it, don't ask it.
- Protect small groups. Don't show results for groups under 5 people. In a team of 3, "the team average" is each person's opinion.
- Ask without manipulating. No questions that push an answer and no surveys that are really advertising. Options should cover every position, including the negative one.
- Use the data only for what you said. If the survey was to improve service, don't use it to sell or share it with third parties. Publish only aggregated results.
- Give something back. Share a summary of the results and what you will do about them. It is the simplest way to respect respondents' time, and what raises participation most next time.
Anonymous and confidential are not the same
- Anonymous: nobody, not even whoever runs the survey, can know who answered what. No names or emails, the same link for everyone, and no small groups shown.
- Confidential: whoever runs the survey does know who answered, but commits not to reveal it and to show only aggregated results. It is normal in a 360 review or when you need follow-up.
Both are valid. What is not ethical is calling a confidential survey "anonymous". If people find out they could be identified, you lose their trust in every survey that follows.
Step by step, with the mistakes that give respondents away:
How to run a truly anonymous surveyHard cases at work
- A manager wants to see the team's answers. Show team results only with 5 or more people, and never individual answers. Agree on it before launch and say it in the introduction.
- Open comments that identify people. "My manager on the night shift at store 3" says who wrote it. Review comments before sharing them and remove identifying details.
- Questions about harassment, health or personal topics. Always offer "Prefer not to answer" and, if the survey may reveal a serious situation, say where to get help. Don't ask what you can't act on.
- Incentives. A small thank-you is fine. A big prize or pressure to reach 100% makes people answer just to answer, which is also an ethical problem with the data.
- Tools and AI. Before passing answers to another tool, especially an AI tool, check what it does with the data and remove names and personal data you don't need.
For reviews with few people per group:
How to protect anonymity in a 360 reviewIn research or a thesis
Research with people usually follows three principles: respect (informed consent and freedom to withdraw), beneficence (minimal risks and clear benefits) and justice (not always burdening the same groups with answering). If your university has an ethics committee, check what it requires before launching: it may need to approve the questionnaire and the consent text.
- With minors you need permission from parents or guardians, plus the minor's assent.
- Keep data only as long as you need it and explain how you will store it.
- In the results, present groups, not people. If you quote an open answer, check it doesn't identify anyone.
The text that goes at the start of the survey:
Informed consent in surveys, with an exampleChecklist before sending
- The introduction says who runs the survey, why and how long it takes.
- It says whether it is anonymous or confidential, and that is true.
- Every profile question will be used in the analysis.
- No result will be shown for groups under 5 people.
- Sensitive questions have "Prefer not to answer".
- No question pushes an answer (have someone outside review it).
- You know who will see the results and what you will do with them.
- You have a date to share a summary with respondents.
For point 6:
How to write unbiased survey questions