Chapter 6 — Peer critique and revision
Lesson map
Lesson 06 published. Map, example, EX033.
flowchart TD
N0["Learning outcomes"]
N1["6.1 The critiques"]
N2["6.2 Responding to the statistician"]
N3["6.3 Responding to the lawyer"]
N4["6.4 Responding to the planner"]
N5["Synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5
N5 --> EX["Worked example"]
EX --> Q["Exercise EX033"]
Logical drawing
flowchart LR
N0["Learning outcomes"]
N1["6.1 The critiques"]
N2["6.2 Responding to the statistician"]
N3["6.3 Responding to the lawyer"]
N4["6.4 Responding to the planner"]
N5["Synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> Q["EX033"]
[Learning outcomes]
|
v
[6.1 The critiques]
|
v
[6.2 Responding to the statistician]
|
v
[6.3 Responding to the lawyer]
|
v
[6.4 Responding to the planner]
|
v
[Synthesis]
|
v
[Worked example] --> [Exercise EX033]
Learning outcomes
You will (1) respond to substantive peer criticism; (2) narrow claims to what evidence supports; (3) distinguish legal, technical and ethical objections.
6.1 The critiques
Ines presents her draft at a doctoral workshop. Three critiques stand out. A statistician argues that synthetic populations may be less diverse than real ones, so uniqueness and linkage may be underestimated. A lawyer argues that the thesis treats epsilon as if it settled the legal question of anonymity under GDPR Recital 26, which it cannot do alone. A planner argues that rating tables on paper is not the same as using them in a real timetable meeting. Ines writes each critique down in full before responding, so that she answers what was actually said and not a weaker version.
6.2 Responding to the statistician
Ines accepts the point and tests it. She regenerates the synthetic population with a skewed distribution in which most riders live in a few dense zones, which is closer to real cities. Uniqueness falls slightly in dense zones and rises in sparse ones, but the overall conclusions do not change. She adds this sensitivity analysis to the thesis and narrows her claim: results hold for populations with similar structure, and a real-data replication inside the authority's secure environment, under its own legal basis and data protection impact assessment, is recommended as future work rather than assumed.
6.3 Responding to the lawyer
Ines agrees that the legal status of a release depends on all the means reasonably likely to be used to identify people, as Recital 26 puts it, including cost, time and available technology. Epsilon is evidence for that assessment, not a substitute for it. She rewrites her conclusion to say that differentially private counts with a documented budget give strong technical evidence that the authority can use in its anonymisation assessment, which its data protection officer and, where necessary, the supervisory authority must still judge. She also cites EDPB guidance on data protection by design as context for documenting such choices.
6.4 Responding to the planner
Ines cannot run real timetable meetings during the thesis, so she narrows H2 to "planners judge the tables usable in a structured review", and adds a limitation. She also proposes a follow-up in which one route review uses the released counts and planners record any decision that differs from the one they would make with exact data. The planner agrees that this is a fairer test. Ines notes in her revision log that two of the three critiques changed her claims, and that this is a normal and healthy outcome of peer review rather than a sign that the thesis was weak.
Worked example
Ines turns three critiques into a sensitivity analysis, a narrower legal conclusion and a narrower utility claim, recording each change in a revision log.
Synthetic case. Workshop participants are synthetic; GDPR Recital 26 and EDPB Guidelines 4/2019 are real.
Errors and limits
Narrowing claims can go too far and leave nothing useful; Ines keeps the core result and limits only its reach.
Exercises
- EX033 — Diversity critique. How does Ines test the statistician's critique? Rubric: 3 = regenerates a skewed population and reruns the analyses as a sensitivity check;; 2 = mentions rerunning;; 1 = dismisses it;; 0 = none.
- EX034 — Reasonably likely means. Why can epsilon not settle anonymity under Recital 26 alone? Rubric: 3 = Recital 26 asks about all means reasonably likely to identify, a legal judgement that uses technical evidence;; 2 = mentions the law;; 1 = says epsilon is enough;; 0 = none.
- EX035 — Narrowed H2. How does Ines narrow H2 after the planner's critique? Rubric: 3 = to usability judged in a structured review, with a limitation and a follow-up route test;; 2 = mentions narrowing;; 1 = keeps H2 unchanged;; 0 = none.
- EX036 — Recording critiques. Why write critiques down in full before responding? Rubric: 3 = to answer the actual argument, not a weaker version;; 2 = mentions accuracy;; 1 = omits;; 0 = none.
- EX037 — Healthy revision. Is it a weakness that two critiques changed her claims? Rubric: 3 = no; revising claims to match evidence is the purpose of review;; 2 = says it is fine;; 1 = says it is a failure;; 0 = none.
- EX038 — Real-data replication. What conditions does Ines attach to a future real-data replication? Rubric: 3 = inside the secure environment, with a legal basis and a DPIA;; 2 = names one;; 1 = omits;; 0 = none.
Synthesis
Chapter 6 shows critique sharpening claims. Chapter 7 assembles the thesis and a reproducibility archive without leaking personal data.
Continue studying — chapter 06 of 08
← Chapter 05 (draft) · Course chapter index · Chapter 07 (draft) →