Research Ethics
Date: 2026-08-17
Consent, privacy, welfare and honesty in research with people. In the UK it’s partly a legal obligation under data protection law rather than only a professional norm — and recordings of participants are personal data with everything that implies.
Research ethics covers the obligations owed to people who take part in research: informed consent, privacy, welfare, fair compensation and honest reporting.
Informed consent
Informed is the operative word — agreeing to something not understood is not consent.
BEFORE THEY AGREE, THEY KNOW
who is running it, and for whom
what will happen, and how long
what is recorded — screen, voice,
face
who will see the recording
how long it is kept
that they can stop at any time,
without giving a reason, and
still be paid
how to withdraw afterwards
how to contact you
“You can stop at any time and still be paid” is the sentence that makes the rest real. Without it, an incentive becomes pressure to continue when someone is uncomfortable.
Consent is ongoing. Someone who agreed at the start can withdraw mid-session, and a facilitator should notice distress rather than wait to be told.
Recordings are personal data
The part that’s legal rather than only ethical.
A SESSION RECORDING CONTAINS
a face, a voice
possibly a name, address, order
history on screen
possibly health information, in a
medical or pharmacy context
Under UK GDPR this is personal data, and where it touches health it’s a special category with a higher bar.
YOU NEED
a lawful basis
a stated retention period, and
actual deletion
security controls on storage
a route for a participant to
request deletion
a record of what you hold
[CHECK: your organisation’s lawful basis, retention policy and any DPIA requirement for recorded research before running sessions — take advice rather than relying on a summary.]
Screen-sharing sessions capture whatever is on screen, including other tabs, notifications and real customer data if the participant is using a live account. Ask them to close everything else, and pause recording if something sensitive appears.
Minimise what you collect
DON'T RECORD faces unless you need
them — voice and screen is usually
sufficient
ANONYMISE in notes and reports
→ "P4", not the person's name
STORE clips, not full recordings,
once analysis is done
REDACT identifying detail before
anything is shared internally
NEVER put raw recordings in a
general-access repository
— Research Repositories
“P4 said…” costs nothing and removes most of the risk from an internal report that will be forwarded further than you expect.
Welfare
NO DECEPTION about the purpose
→ "we're testing the site, not you"
must be true and demonstrated
NO PRESSURE to continue
REALISTIC TASKS
don't ask someone to enter their
real card details
WATCH FOR DISTRESS
particularly on sensitive topics —
health, money, bereavement
DEBRIEF
answer their questions afterwards
In a health or pharmacy context this is heavier than it looks. A participant discussing a condition is disclosing sensitive information, and the session design should let them decline specifics without derailing the task.
Compensation
PAY FAIRLY
for their time, at a rate
reflecting it
PAY EVERYONE who turns up
including those who withdraw and
those screened out after arriving
PAY PROMPTLY
DON'T pay only in vouchers for your
own product
→ that's a discount, not
compensation
Under-paying skews the sample towards people with time and towards your most engaged customers, which is a research quality problem as well as a fairness one — User Interviews.
Honest reporting
The ethics that apply after the sessions:
REPORT what you found, including
what contradicts the plan
DON'T cherry-pick the quote that
supports the roadmap
— Qualitative Coding
STATE the sample and the method
→ "5 participants, moderated, all
existing customers"
DON'T inflate to percentages
— Sample Size in Qualitative Research
DISTINGUISH observation from
interpretation
See: Qualitative Coding · Sample Size in Qualitative Research
Presenting a finding without its sample size is the most common reporting failure, and it lets a five-person study be read as a survey.
A workable minimum
1 a plain-language consent form,
agreed before recording starts
2 recording only what's needed
3 a stated retention period, actually
enforced
4 anonymised notes and reports
5 fair, prompt payment to everyone
6 a named contact for withdrawal
Point 3 is the one that quietly fails. Retention periods are agreed and then nobody deletes anything — which is exactly the kind of drift a repository makes invisible.