Algorithmic fairness

Algorithmic Fairness

Course Leader

Yara Abu Awad (Swiss Federal Statistical Office)

Official statisticians wishing to learn about algorithmic fairness, bias mitigation strategies and explainability in the context of AI.

• Sound command of English. Participants should be able to actively participate in discussions and to make brief presentations.

The main objectives of the course are:

• To introduce the concept of algorithmic fairness, how bias is measured and how it is mitigated.

• To introduce the concept of explainability and how it can contribute to algorithmic fairness.

• What is algorithmic fairness

• Steps in training an AI model

• Measuring fairness / bias

• Bias mitigation strategies

• Legal considerations around AI discrimination

• Introduction to explainability

• What if? Causal inference for explainability

Participants should be able to:

• provide examples of biased algorithms and explain the sources of bias;

• list different bias mitigation strategies;

• explain the concept of explainability and how it can contribute to algorithmic fairness;

• explain the concept of causal inference and how it can be used for algorithmic fairness and explainability.

• Presentations and lectures.

• Group discussions and group presentations.

None

Yara Abu Awad (Federal Statistical Office)

Christopher Sulkowski (Federal Statistical Office)

Kerstin Johansson Baker (Federal Statistical Office)