Master Thesis Project: Evaluating Fairness Interventions in Machine Learning

Details

This thesis will investigate and compare approaches for mitigating unfairness across the machine learning lifecycle, including data preparation, model development, and prediction adjustment stages.

The research will assess how different fairness interventions affect model performance and fairness metrics using representative datasets and practical use cases. The objective is to provide insights into the strengths, limitations, and trade-offs of different techniques, and to develop recommendations for the responsible design of machine learning systems.

At the ING Analytics department, model predictions are made about retail and corporate clients, where fairness is an important consideration. The Wholesale Banking Advanced Analytics department is a large team of data scientists, data engineers, software developers, and other specialists focused on applying data, machine learning, and statistical modeling to products for clients and internal users. The team works closely with master's students on academic and practical topics and has extensive experience with student supervision.

Responsibilities

• Investigate fairness interventions across the machine learning lifecycle.
• Compare the effects of different techniques on model performance and fairness metrics.
• Work with representative datasets and practical use cases.
• Develop recommendations for the responsible design of machine learning systems.
• Aim to contribute to a publication.
• Complete the thesis project over a period of at least six months.

Requirements

• Master's student interested in machine learning, responsible AI, algorithmic fairness, or applied data science.
• Solid experience with Python.
• Experience with machine learning.
• Strong skills in statistics and linear algebra, including matrix rank, singular values, and matrix decomposition.
• Availability for at least six months to complete the thesis project.
• Enrollment at a Dutch university, or at an EU university for EU passport holders, throughout the internship.
• A curious, collaborative attitude.

Benefits

• Compensation of EUR 700 per month.
• Internship allowance based on a 36-hour work week.
• Close supervision and interaction with a community of data scientists.
• Own work laptop.
• Hybrid working, combining home working with office working.
• Personal growth and challenging work.
• Informal working environment with innovative colleagues.