How to Develop Fair Algorithms?
Organized by ZHAW and ethix
The 21st century is shaped by the ever-increasing use of data for getting new insights and making better decisions. The center of such applications are data-based prediction models. More often than not, these systems do produce unintended discrimination and social injustice, a phenomenon which has been called “algorithmic bias” or “algorithmic fairness”. Newly built tools and the ones already in place today are both affected. This has triggered European Union lawmakers to develop and publish a proposal for a Regulation on Artificial Intelligence (AI) earlier this year, which requires algorithmic systems to avoid such biases. In addition to that, there is an increasing awareness of society and customers regarding the social implications of data-based systems. More and more, these systems are expected to be fair, non-biased, and non-discriminatory.
However, in practice, it is not clear how to create fair algorithms and how to ensure that data-based prediction and decision models fulfill clearly defined fairness requirements. In this workshop for practitioners and data scientists, you will learn how to build fair prediction-based algorithms, and how fairness issues are to be included in predictive modeling.
Lessons to be learned:
In this hands-on workshop:
- You will learn how to combine data-based prediction models with fairness requirements.
- You will learn how algorithmic (un)fairness is defined and measured in a practical context.
- You will learn how to construct fair decision algorithms while still harvesting the benefit of a good prediction model. This is based on a recently developed methodology.
- You will apply the methodology to concrete use cases and examples.
Our workshop is mainly intended for data scientists. However, deep technical knowledge is not required. Therefore, we are happy to welcome a broad audience with different backgrounds from both industry and academia.
Time Frame: June 22, 13:00 – 17:00
The workshop consists of three parts:
Part 1: Introduction to Algorithmic Fairness
- How do we measure discrimination or unfairness in practice?
- What is the reason that prediction-based decision systems naturally tend to be racist and sexist?
- How can we ensure that a data-based decision algorithm achieves maximum performance while satisfying a fairness constraint?
Part 2: How to build fair models and algorithms?
- How to analyze the fairness requirements of a specific application?
- How to transfer these requirements into a fairness metric?
- How to combine business requirements and fairness requirements in prediction-based decision making?
Part 3: Hands-on activities
- You will apply the insights from the first two parts to various use cases.
- You will enforce different fairness requirements on machine learning algorithms.
- You will understand the impact that fairness requirements have on the performance of these algorithms.
- You will gain experience in developing fair prediction-based decision systems. This enables you to harvest the benefits of such systems while still making fair decisions.
University of Zurich
Digital Society Initiative
Room: Eventroom SOC-E-010
Max. participants: 30