Module 10 Article: Ethics — Doing Research You Can Defend

The big idea

Research ethics is the social contract between researchers and society: you get the privilege of studying people and publishing, and in return you must protect participants and be honest. The three classic pillars — respect for persons, beneficence, and justice (from the 1979 Belmont Report) — still structure every ethics review today, whether for medical, social-science, or CS research.

The three Belmont principles

Principle	What it means	Research translation
Respect for persons	People can set their own limits	Informed consent; protect vulnerable groups
Beneficence	Maximise benefits, minimise harms	Risk/benefit assessment
Justice	Fairness in selection & burden	Fair subject selection; no exploitation

Informed consent

A participant must understand (and agree to) four things:

1. The purpose of the study.
2. The procedures (what they will do / data used).
3. The risks and benefits (realistic, not inflated).
4. Their rights — voluntary participation, withdrawal at any time, no penalty.

Documentation: usually a signed form; for online surveys, a checkbox + info page. Waiver of consent is allowed only by the ethics committee (IRB/REC/IEC) in exceptional cases.

For our stress study

Consent must say: PSS data will be de-identified and stored encrypted, used only for research, kept 5 years then destroyed; participation is voluntary and withdrawal is fine at any point. Students should know the exercise programme is optional and that their grades will not be affected.

Ethical committees and approvals

• IRB (US), REC (UK), IEC/ERB (many countries) — independent bodies reviewing protocol, consent, and risk.
• They ask: scientific merit (is it worth the risk?), risk-benefit, consent adequacy, and privacy safeguards.
• Continuing review: most approvals are annual; major changes require amendment.
• Expedited / exempt: minimal-risk studies may skip full board review (but still need ethics approval).
• A study is unethical to publish without approval if it involved human subjects — and many high-profile retractions are for exactly this reason.

Data privacy and responsible data handling

Core duties

1. Minimise: collect only what the question needs (not "just in case").
2. Anonymise / pseudonymise: strip or code identifiers.
3. Secure: encryption + access controls (only the research team).
4. Comply: GDPR (EU), HIPAA (US health), local data laws.
5. Plan retention & disposal: keep per policy (often 5–10 years), then destroy.

For our study

• Store PSS responses under a numeric code (no names); keep the name↔code key in a separate, encrypted file.
• Do not link to academic records unless consent explicitly covers it (it usually does not).

Responsible publishing

• Honest reporting: include non-significant results, limits, and methodological weaknesses.
• Authorship (ICMJE criteria): all of (1) substantial contributions, (2) drafting/ revising, (3) final approval, (4) accountability. Avoid gift/guest and ghost authorship.
• Reproducibility: share code/data where ethical (repositories: Zenodo, OSF).
• Pre-registration: publicly register hypotheses/methods before data collection (clinicaltrials.gov, OSF) to stop HARKing (Hypothesizing After Results are Known).
• Conflict of interest disclosure: funding, consultancy, patents.

Research misconduct: the FFP trio

Misconduct	What	Harm
Fabrication	Making up data/results	False knowledge
Falsification	Manipulating data/analysis	False knowledge
Plagiarism	Using others' ideas/work without credit	Stolen credit

Other questionable practices (not all are "misconduct" but are harmful):

• HARKing (Hypothesizing After Results are Known).
• p-hacking (trying many analyses until p < 0.05).
• Salami slicing (one study → many thin papers).
• Selective reporting (publishing only favorable outcomes).

Case study: research misconduct

The case of a high-profile psychology paper retracted for falsified data: the author recycled the same spreadsheet across experiments and doctored images. Detection came from a reader who noticed identical background noise in supposedly independent photographs. Lesson: transparent data and code make fraud harder to hide.

Case studies in research misconduct (brief survey)

1. Cold fusion announcement (1989): premature, non-reproducible claim → damaged credibility.
2. Stapel affair (2011): wide-scale fabrication → 58 retracted papers, reforms in Dutch psychology.
3. Macchiarini trachea implants: faked results + unapproved devices → patient deaths, retractions.

Common thread: lack of raw-data sharing and weak oversight enabled large-scale fabrication to persist.

Balancing openness with risk

Some data can't be shared (health data, vulnerable subjects). Balance:

• Maximise openness (code, synthetic data, descriptive stats),
• Minimise risk (redact identifiers, use data-use agreements),
• Be transparent about exactly what is and isn't shared and why.

Exam angle

For "ethical considerations in research":

1. State the three Belmont principles and connect each to a practice (consent, risk/benefit, fair selection).
2. List the four things informed consent must cover.
3. Name the three FFP forms of misconduct and one additional questionable practice (HARKing or p-hacking).
4. Mention data minimisation, anonymisation, and GDPR/ HIPAA by name.

Common mistakes

• Treating ethics as paperwork (ethics is integral, not a hurdle).
• Believing anonymity = automatic consent (still need justification).
• Collecting more data than required "for later use."
• Ghost/gift authorship (credit without contribution / contribution without credit).

Memory rules

• Belmont: Respect, Beneficence, Justice.
• Consent: purpose + procedures + risks/benefits + rights.
• FFP: Fabrication, Falsification, Plagiarism.
• Privacy: minimise, anonymise, secure, comply, dispose.
• Publish honestly, share code/data, disclose conflicts.

Check your understanding

1. Name the three Belmont principles.
2. What is the single most common form of scientific misconduct (worldwide)?
3. Give one reason to pre-register a study.
4. What does "minimisation" mean in data privacy?

Answers

Reveal answers after attempting the questions

1. Respect for persons, beneficence, justice.
2. Questionable: plagiarism is most commonly reported (fabrication and falsification are less common but more serious). (Accept "self-plagiarism / duplicate publication" too.)
3. To prevent HARKing and p-hacking by locking in hypotheses/methods.
4. Collect only the minimum data needed to answer the research question.

See also

• Module 9 for the writing/publishing side of research integrity.
• Glossary.md (human subjects, IRB, informed consent, conflict of interest).