Questions: Research Ethics, IRB Oversight, and Research Integrity

Question 1

A researcher conducts an online survey about consumer preferences and argues that since it's 'minimal risk,' they don't need IRB approval. Which response best identifies the flaw in this reasoning?

• A. Online surveys never qualify as minimal risk — all digital data requires full board review
• B. Minimal risk studies still require IRB review; they may qualify for expedited rather than full review, but oversight is not waived
• C. The researcher is correct — minimal risk studies are specifically exempt from all oversight requirements
• D. The IRB only reviews biomedical research; behavioral surveys are not covered

Answer: (B)

Minimal risk means risks are no greater than those of daily life — it lowers the level of review required (to expedited), not the requirement for review itself. IRBs oversee all human subjects research, including behavioral and online studies. Full board review is required for more-than-minimal risk or vulnerable populations, but no category of human subjects research is simply exempt from institutional oversight.

Question 2

A researcher runs 20 slightly different analyses on the same dataset, reports only the one with p < .05, and publishes that result. This is an example of:

• A. Rigorous exploratory analysis — running many tests is standard statistical practice
• B. P-hacking — selectively reporting results to achieve statistical significance
• C. Falsification — the researcher altered the data to produce a desired result
• D. HARKing — the researcher changed the hypothesis after seeing the results

Answer: (B)

P-hacking (or 'fishing') involves running multiple analyses and reporting only the significant one, inflating the false-positive rate. If you run 20 independent tests at α = .05, you expect one false positive by chance alone. Reporting only that one creates a misleading impression of a real effect. This is distinct from falsification (manipulating data) or HARKing (reframing the hypothesis post-hoc) — though all three distort the literature. Exploratory analysis is legitimate, but all comparisons must be reported, or the analysis must be pre-registered.

Question 3

Research misconduct refers exclusively to fabrication, falsification, and plagiarism — practices involving outright dishonesty about data or authorship.

• T. True
• F. False

Answer: False

The FFP triad is the most severe category, but the modern conception of research integrity is broader. P-hacking, HARKing (Hypothesizing After Results are Known), selective outcome reporting, and failure to share data or methods are now equally scrutinized as integrity violations — even without outright lying. These 'questionable research practices' distort the literature and undermine replication just as surely as fabrication, only more subtly.

Question 4

Ethics functions as an external constraint on the research process — a set of rules imposed by institutions to limit what scientists would otherwise do freely.

• T. True
• F. False

Answer: False

Ethics is internal to the scientific process, not external to it. A study that harms participants, obtains coerced consent, or falsifies data doesn't just violate rules — it produces corrupted knowledge. The integrity of scientific findings depends on the integrity of the processes that generate them. Ethical violations don't just harm participants; they invalidate the research itself. Ethics and rigor are therefore the same thing viewed from different angles, not competing concerns.

Question 5

Why is transparency — pre-registration of hypotheses, open data, open materials — described as a structural response to problems of research integrity rather than merely a courtesy to other researchers?

Model answer: Transparency makes the research process legible to external scrutiny, which is the institutional mechanism for catching errors and deterring misconduct before they propagate through the literature. Pre-registration prevents HARKing by locking in hypotheses before data is collected. Open data allows others to detect analytical errors or p-hacking. Open materials enable replication. Without these structural checks, problematic practices are invisible to reviewers and readers who only see the polished final report.

The argument is that individual integrity, while necessary, is insufficient. Well-documented cognitive biases cause researchers to unconsciously favor results that confirm their hypotheses — not through malice but through motivated reasoning. Structural transparency removes the opportunity for these biases to operate silently: when the analysis plan is pre-registered and the data is public, researchers must confront disconfirming results rather than quietly setting them aside. Transparency converts the research process from a private activity into a public one subject to collective scrutiny.