How Good Are Privacy Guarantees? Platform Architecture and Violation of User Privacy
Bibliographic record
Abstract
Many platforms deploy data collected from users for a multitude of purposes.While some are beneficial to users, others are costly to their privacy.The presence of these privacy costs means that platforms may need to provide guarantees about how and to what extent user data will be harvested for activities such as targeted ads, individualized pricing, and sales to third parties.In this paper, we build a multi-stage model in which users decide whether to share their data based on privacy guarantees.We first introduce a novel mask-shuffle mechanism and prove it is Pareto optimal-meaning that it leaks the least about the users' data for any given leakage about the underlying common parameter.We then show that under any mask-shuffle mechanism, there exists a unique equilibrium in which privacy guarantees balance privacy costs and utility gains from the pooling of user data for purposes such as assessment of health risks or product development.Paradoxically, we show that as users' value of pooled data increases, the equilibrium of the game leads to lower user welfare.This is because platforms take advantage of this change to reduce privacy guarantees so much that user utility declines (whereas it would have increased with a given mechanism).Even more strikingly, we show that platforms have incentives to choose data architectures that systematically differ from those that are optimal from the user's point of view.In particular, we identify a class of pivot mechanisms, linking individual privacy to choices by others, which platforms prefer to implement and which make users significantly worse off.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.018 | 0.068 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.006 |
| Scholarly communication | 0.008 | 0.019 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.005 | 0.005 |
| Insufficient payload (model declined to judge) | 0.005 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".