"PhD Position on Mathematical Foundations for Explainable AI"

Your project: Explainable Artificial Intelligence (XAI) – From Mathematics to Meaning

AI and machine learning models are becoming increasingly powerful, but their decision-making processes often remain opaque, relying on vast numbers of learned parameters that are difficult to interpret. For example, why does a model predict that a patient should remain in intensive care? Or why does it flag a certain transaction as potential money laundering?

This project aims to demystify such decisions by developing a mathematical framework for explainable AI (XAI) — one that generates understandable, trustworthy explanations of machine learning models. We will combine theoretical and experimental approaches: using tools from mathematical machine learning theory to derive guarantees about explanation quality, and programming and empirical testing to validate those methods.

During the PhD, you will begin by working on a pre-specified subproject, giving you the opportunity to build foundational expertise in the topic. As your experience grows, you will take on increasing responsibility for shaping your own research direction. You will collaborate closely with Dr Tim van Erven and other team members, and connect with national and international researchers in XAI.

Tasks and responsibilities

• Conduct research within the scope of the XAI project;
• Publish findings in peer-reviewed journals and conference proceedings;
• Present research at international conferences and seminars;
• Participate in the Faculty of Science PhD training program;
• Assist in teaching and mentoring undergraduate or Master’s students;
• Complete and defend a doctoral dissertation within four years.

Example references

(You are not expected to study these in detail prior to applying, but they offer insight into the project's scope and style.)

1. Fokkema, H., de Heide, R., & van Erven, T. (2023). Attribution-based Explanations that Provide Recourse Cannot be Robust. JMLR, 24(360), 1–37.
2. Fokkema, H., Garreau, D., & van Erven, T. (2024). The Risks of Recourse in Binary Classification. AISTATS, 238, 550–558.
3. Fokkema, H., van Erven, T., & Magliacane, S. (2025). Sample-efficient Learning of Concepts with Theoretical Guarantees. arXiv:2502.06536.

What do you have to offer?

Required qualifications

• A Master’s degree in mathematics, artificial intelligence, or computer science with a strong mathematical component (including formal proof writing), to be completed before the position starts;
• Strong interest in mathematical machine learning theory and explainable AI;
• Proficiency in Python (or similar language) for developing and testing new algorithms;
• Excellent verbal and written communication skills in English;
• Demonstrated ability to work independently and in teams.

Preferred qualifications

• Background in mathematical machine learning theory or mathematical statistics;
• Experience with or interest in explainable AI;
• Extracurricular engagement (e.g. in student associations, music, sports).

What do we offer you?

We offer a full-time PhD position (38 hours/week) for four years. The contract will start with an 18-month appointment and be extended upon positive evaluation.

• Start date: Preferably between 1 September 2025 and 1 January 2026;
• Salary: €2,901 gross/month in the first year, increasing to €3,707 in the fourth year (scale P);
• Benefits: 8% holiday allowance and 8.3% year-end bonus;
• Conditions: Governed by the Collective Labour Agreement of Dutch Universities (CAO-NU).

As part of the role, you will receive an educational training plan, attend international meetings, and contribute to the University’s educational programs.

Your place at the UvA

The University of Amsterdam (UvA) is the Netherlands’ largest university and a top international research institution. Home to over 42,000 students and 3,000 PhD candidates, UvA is an intellectually vibrant community with a culture of innovation and curiosity. You will join the Faculty of Science and be part of a collaborative, cutting-edge research environment.

Important to know: Application & Contact

If this opportunity excites you and fits your profile, we welcome your application by 2 June 2025.

Your application should include (in one single PDF, except for your CV):

• A CV (can be a separate file);
• A motivation letter (max. 1 page), including your estimated starting date;
• A transcript of BSc and MSc grades;
• A sample of a mathematical proof you have written (e.g., coursework or thesis);
• (Optional) A thesis or project report demonstrating your academic writing;
• The name and email address of an academic referee (e.g., thesis supervisor) who is willing to provide a recommendation.

Send your application via the red application button on the UvA site. Only complete applications submitted via the official portal will be considered.

Contact for informal inquiries:

Dr Tim van Erven – tim@timvanerven.nl

Tentative application timeline

• Initial notification: by 10 June 2025;
• Round 1 (online interview): week of 23 June 2025;
• Round 2 (in-person interview): week of 7 July 2025.

Note: A knowledge security check may be part of the hiring process. For more information, see the Dutch national knowledge security guidelines.