EDPS0007 Chap.10 Research Ethics, Fraud and Evidence in Policy
Research Ethics, Fraud and Evidence in Policy
Research Ethics, Fraud and Evidence in Policy closes the module by asking how far research can be trusted and how far policy uses it.
Lectures 8 and 9 present cases of misconduct: suspect images across 132 papers by an influential Alzheimer's researcher, a university president resigning over flawed laboratory practice, an honesty researcher found to have altered data, and ghostwritten journal articles promoting a pharmaceutical firm's drugs.
Paper mills selling authorship of fabricated studies accounted for an estimated 0.7% of about 3 million articles a year in 2020. Short scenarios test ethical judgement, from deceptive experiments and withheld treatment to citation trading, biased module evaluations and journals favouring surprising findings.
The final week asks when policymakers act on research they commission, linking back to engineering funding, humanities places and China's Double Reduction Policy, and a study of AI assistance warns that better tools can weaken the motivation excellent work requires.
None of this makes research worthless; it makes checking sources, funding and incentives part of reading any study, which is the habit the module has been building since its first week.
What this chapter covers
- 01
Research fraud cases and their scale
- 02
Publish or perish and paper mills
- 03
Consent, harm, honesty and power
- 04
Randomised trials and withheld treatment
- 05
Publication bias towards surprising findings
- 06
How policymakers use commissioned research
- 07
AI assistance and the good enough problem
Worked example · free
Estimate how many paper-mill articles appear each year
- 10.7% of 3,000,000 = 0.007 × 3,000,000 = 21,000, which the lecture rounds to about 20,000 articles a year.
- 1The share is small, so most research is not fabricated, but the absolute number is large.
- 1For citing, prefer module readings and established sources, and check who produced a surprising study before relying on it.
Key terms
- Paper mill
- A company that sells authorship of fabricated scientific papers, often produced with AI.
- Publication bias
- The tendency for journals to publish striking or surprising results more readily than routine ones.
- Informed consent
- A participant's free agreement to take part based on adequate information about a study.
Research Ethics, Fraud and Evidence in Policy FAQ
Why would a researcher falsify data?
The lecture poses this as a question after its fraud cases. Pressure to publish, career advancement and funding are obvious candidates, but they should be treated as hypotheses about motive rather than established facts in any particular case.
Is a randomised trial of a new drug ethical?
It can be, if participants consent and the comparison group receives an existing treatment rather than none. The ethical problem in the lecture's withheld-treatment case is denying available care, not randomisation itself.
Do policymakers follow the research they commission?
Week 10 asks how often they act on it, and in which conditions. Module examples suggest research competes with money, values and politics, as in tuition fees, where evidence on access sits beside judgements about fairness and taxation.
What is a paper mill?
It is a business that sells authorship of fabricated research papers, often produced with help from AI. The lecture cites an estimate that such papers made up 0.7% of all scientific articles in 2020, which is roughly 20,000 papers each year.
Exam move
Use this chapter to sharpen critical reading across all three exam questions rather than as a separate topic. Learn two or three fraud cases well enough to describe them briefly and draw a lesson, such as checking images, funding and authorship. Practise the four ethics questions on each lecture scenario until you can say in one sentence what makes a design ethical or unethical.
For the research-policy question, collect two module examples where evidence and policy diverge and explain why, distinguishing missing evidence from competing values. Finally, prepare one paragraph on AI that uses the illustrators study carefully, noting it concerns professional work rather than students, and connect it to the module's assistive AI rules and to the risk of submitting drafts that are not your own.
For each fraud case, write the single warning sign a careful reader could have looked for, such as unusual images, undisclosed funding or a result that looks too neat, and add it to your evidence-checking routine.
Working through Research Ethics, Fraud and Evidence in Policy in EDPS0007? Sia is AskSia’s AI Education tutor — ask any EDPS0007 Research Ethics, Fraud and Evidence in Policy question and get a clear, step-by-step explanation grounded in how EDPS0007 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.