Introduction to Algorithmic Fairness: Principles, Methods and Regulatory Perspectives

Dr. Erasmo Purificato
European Commission, Joint Research Centre (JRC), Italy

Abstract

The course provides a comprehensive introduction to algorithmic fairness, exploring key concepts such as definitions, bias characterisation and the potential sources of unfairness in machine learning models. Initially, we will thoroughly examine fairness criteria, bias detection metrics, and the limitation of fairness evaluation in binary scenarios. Then, we will analyse the emerging multiclass and multigroup approaches, and cover bias mitigation techniques and their practical trade-offs. Finally, the course will examine legal and ethical frameworks governing algorithmic fairness, with a focus on EU regulations such as the GDPR, DSA, and AI Act, as well as global policies.

Outline

• Lecture 1: Foundation of Algorithmic Fairness
  • Why fairness matters in AI and ML
  • Defining fairness and bias
  • Potential causes of unfairness in ML
  • Fairness criteria
  • Conflicts between fairness goals
• Lecture 2 – Measuring Bias and Fairness
  • Bias detection metrics
  • Challenges in binary scenarios
  • Extending fairness metrics to multiclass and multigroup scenarios
• Lecture 3 – Mitigating Bias
  • Bias mitigation strategies
  • Choosing the right fairness intervention
  • Trade-offs and practical implementations
• Lecture 4 – Legal and Ethical Frameworks for Fairness in AI
  • Overview of EU Regulations affecting AI and ML (i.e., GDPR, DSA and AI Act)
  • Fairness principles in EU Regulations
  • Fairness principles in global regulations
  • The future of algorithmic fairness and open research challenges

Schedule

The course will have a total duration of 10 hours, scheduled as follows:

• May 21, 14:00-18:00 Aula II
• May 22, 10:30-12:30 Aula F
• May 22, 14:00-16:00 Aula F
• May 23, 10:00-12:00 Aula F

Exam

The final exam consists either in a seminar presentation focusing on a specific topic studied during the course or in a test held on the last day of the course. The definitive format will be announced when the schedule is finalized. The course will be held in person. Please contact me if you are interested in joining.

References

The content of the course is based (but not limited to) the following articles:

1. Simon Caton and Christian Haas. Fairness in Machine Learning: A Survey. ACM Comput. Surv. 56, 7, Article 166 (2024). https://dl.acm.org/doi/10.1145/3616865
2. Corbett-Davies, Sam, Johann D. Gaebler, Hamed Nilforoshan, Ravi Shroff, and Sharad Goel. The measure and mismeasure of fairness. Journal of Machine Learning Research 24, no. 312 (2023). https://jmlr.org/papers/v24/22-1511.html
3. Dana Pessach and Erez Shmueli. A Review on Fairness in Machine Learning. ACM Comput. Surv. 55, 3, Article 51 (2023). https://doi.org/10.1145/3494672
4. Sahil Verma and Julia Rubin. Fairness definitions explained. In Proceedings of the International Workshop on Software Fairness (FairWare 2018). https://doi.org/10.1145/3194770.3194776
5. Purificato, Erasmo, Ludovico Boratto, and Ernesto William De Luca. Toward a responsible fairness analysis: from binary to multiclass and multigroup assessment in graph neural network-based user modeling tasks. Minds and Machines 34, no. 3 (2024). https://doi.org/10.1007/s11023-024-09685-x
6. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation),
   2016, OJ L119/1. http://data.europa.eu/eli/reg/2016/679/oj
7. Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on a Single Market For Digital Services and amending Directive 2000/31/EC (Digital Services Act), 2022, OJ L277/1. http://data.europa.eu/eli/reg/2022/2065/oj
8. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and
   Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act), 2024, http://data.europa.eu/eli/reg/2024/1689/oj