back to list

Research Project: Fairness-aware AI

Description

FairML or Fairness-aware AI has emerged into a large interdisciplinary area. in DAI, we approach the challenges of fairness and non-discrimination from several perspectives, focusing on the technical ML aspects intersections of FairML and XAI, and interdisciplinary aspects including philosophical, legal and computer science. Several subprojects include:

• Fairness and unwanted biases in recommender systems (See research output of Masoud Mansoury co-supervised together with Bamshad Mobasher, Robin Burke, and recent collaboration with Bol.com)
• Fairness in Social Network Analytics
• FairML in banking and insurance (in collaboration with Rabobank, DLL, ING, Floryn)
• Moral and legal justification of FairML 
• VACE: Value alignment for counterfactual explanations in AI (co-led together with Emily Sullivan)
  
• Fairness in AutoML 
• Fairness in Reinforcement Learning scenarios
• Fairness as predictive modeling with independency constraints (pioneered in 2008 by Toon Calders and his then PhD candidate Faisal Kamiran)

We contribute to Fairlearn: an open-source, community-driven project that aims to help data scientists improve fairness of machine learning models.

We have co-organized a series of events, including BIAS @ ECMLPKDD, Fair ADM Lorentz Workshop Fairness in Algorithmic Decision Making: A Domain-Concrete Approach, SIGKDD and DAMI special issues on fairness and bias in AI.

Details

Principal Investigator

Mykola Pechenizkiy

Involved members:

George Fletcher

Akrati Saxena

Pratik Gajane

Hilde Weerts

Cassio de Campos

Maryam Tavakol