10 Ethics and Privacy

In this block we cover:

• Ethical data science
  • The law: Human Rights Act, GDPR and the Data Protection Act 2018
  • GDPR and its implications for data science
• Statistical disclosure
  • Anonymity and re-identification (Governor Weld, the Netflix prize)
  • Statistical disclosure attacks
  • Differential privacy, the Laplace mechanism and randomized response
  • Practice for sharing anonymised data
• Interpretability and fairness
  • Interpretable algorithms vs interpreting black box algorithms
  • Algorithmic fairness: counterfactuals and proxy data
• Discussions
  • Opting out of NHS data
  • Privacy in the census
  • Interpretable models or black boxes?

Lectures:

Ethics and Privacy Overview:

• 10.1 Ethics in Data Science (Part 1, Overview)
• 10.1.1 Opting Out of NHS Data Discussion

Statistical Disclosure:

• 10.2 Ethics in Data Science (Part 2, Disclosure)
• 10.2.1 Privacy in the Census Discussion

Fairness and Interpretability:

• 10.3 Ethics in Data Science (Part 3, Fairness and Interpretability)
• 10.3.1 Interpretable or Black Box? Discussion

Assessments:

• The main long form portfolio Portfolio 10.
• Block10 on Noteable via Blackboard (2026-27 link)

References

The law

• Human Rights Act (HRA 1998)
• EU General Data Protection Regulation (GDPR 2018)
• Data Protection Act 2018 (DPA 2018)
• KDNuggets has a very good description of how GDPR affects Data Science.

Statistical disclosure and privacy

• Dwork and Roth 2014 The Algorithmic Foundations of Differential Privacy.
• ONS policy on disclosure control.
• Sweeney 1997. Weaving technology and policy together to maintain confidentiality. Journal of Law, Medicine & Ethics, 25:98–110.
• Reidentification becomes a matter of statistics, on the Governor Weld case.
• Narayanan and Shmatikov 2008. Robust de-anonymization of large sparse datasets (how to break anonymity of the Netflix prize dataset). IEEE Sec. and Priv.
• George Danezis Statistical Disclosure Attacks.

Fairness and interpretability

• Christoph Molnar 2019 Interpretable Machine Learning: A Guide for Making Black Box Models Explainable.
• Francesco Bonchi Algorithmic Bias Tutorial with Slides from KDD 2016.
• Hardt, Price and Srebro 2016 Equality of Opportunity in Supervised Learning, explored in Adrian Colyer’s blog.

Navigation:

Previous: Block 09. Next: Block 11.