Fairness-aware machine learning and data mining

Nowadays, many decisions are made using predictive models built on historical data, for example, automated CV screening of job applicants, credit scoring for loans, or profiling of potential suspects by the police. Algorithms decide upon the prices we pay, the people we meet, or the medicine we take. 

Growing evidence in the media and research suggests that decision making by algorithms can potentially discriminate people. This may happen even if the computing process is fair and well-intentioned. This is because most data mining methods are based upon assumptions that the historical data is correct, and represents the population well, which is often not true in reality. Moreover, usually predictive models are optimized for performing well in the majority of the cases, not taking into account who is affected the worst by the remaining inaccuracies.

Fairness-aware machine learning and data mining studies in which circumstances algorithms may become discriminatory, and how to make predictive models free from discrimination, when data, on which they are built, may be biased, incomplete, or even contain past discriminatory decisions. 


Žliobaitė, I. and Custers, B. (2016).
Using sensitive personal data may be necessary for avoiding discrimination in data-driven decision models. 
Artificial Intelligence and Law 24(2), p. 183-201.  PDF

Žliobaitė, I. (2015). 
On the relation between accuracy and fairness in binary classification. The 2nd workshop on Fairness, Accountability, and Transparency in Machine Learning (FATML) at ICML'15.  PDF  arXiv

Calders, T. and Žliobaitė, I. (2013). 
Why Unbiased Computational Processes Can Lead to Discriminative Decision Procedures. Discrimination and Privacy in the Information Society, series: Studies in Applied Philosophy, Epistemology and Rational Ethics, Vol. 3., p. 43-57DOI  PDF

Kamiran, F., Žliobaitė, I. and Calders, T. (2013). 
Quantifying explainable discrimination and removing illegal discrimination in automated decision making. Knowledge and Information Systems 35(3), p. 613-644. DOI  PDF

Kamiran, F. and Žliobaitė, I. (2013). 
Explainable and Non-explainable Discrimination in Classification. Discrimination and Privacy in the Information Society, series: Studies in Applied Philosophy, Epistemology and Rational Ethics, Vol. 3., p. 155-170. DOI  PDF

Žliobaitė, I., Kamiran, F., Calders, T. (2011). 
Handling Conditional Discrimination. Proc. of the 11th IEEE Int. Conf. on Data Mining (ICDM'11), p. 992 - 1001. DOI  PDF  code


May 2016 » Ethical machines: data mining and fairness AID-FOrum, Helsinki summary
Jul 2015 » Can algorithms discriminate? at EU Fundamental Rights Agency, Vienna, Austria slides
May 2015 » Can machines discriminate? and how to avoid that, seminar talk at HIIT slides


Fall 2015 » I am teaching a seminar course on Non-discriminatory machine learning (T-61.6010) at Aalto University and University of Helsinki.


We organized a workshop on Discrimination and Privacy-Aware Data Mining at IEEE ICDM 2012 in Brussels.