Fully-Funded PhD Position in Symbolic Explainability for Logic-Based AI System

Fully-Funded PhD Position in Symbolic Explainability for Logic-Based AI System

Category: PhD

1 - Working at the VUB

For more than 50 years, the Vrije Universiteit Brussel has stood for freedom, equality and solidarity, and this is very much alive on our campuses among students and staff alike.

At the VUB, you will find a diverse collection of personalities: innovators pur sang, but above all people who are 100% their authentic selves. With some 4,000 employees, we are the largest Dutch-speaking employer, in the private sector, in Brussels; an international city with which we are only too happy to connect and where (around) our 4 campuses are located.

Add to this our principle of free research - in which self-reflection, a critical attitude and an open, creative mind around scientific and social issues are central - and you have a university that is fundamentally groundbreaking and pioneering in education and research. In short: the VUB all over again.

Moreover, the VUB is a member of EUTOPIA, an alliance of like-minded European universities, all ready to reinvent themselves.

The Faculty of Sciences and Bioengineering Sciences, Department Computer Science, is looking for a PhD-student with a doctoral grant.

More concretely your work package, for the preparation of a doctorate, contains:

Artificial Neural Networks and Large Language Models offer state-of-the-art performance at numerous AI tasks, but being black boxes, they lack explainability which makes them difficult to trust. By contrast, symbolic approaches have reliability and explainability "built in", since they are based on logical reasoning.

This motivates the development of neurosymbolic approaches as a way of combining the best of both worlds. However, it turns out that even the explanations that are generated by symbolic systems are not always easy to understand for people. This is particularly the case when the space of possibilities grows large, as typically happens in neurosymbolic approaches.

Typical methods for explaining the conclusions of symbolic reasoners are based on notions such as unsatisfiable cores: minimal sets of assumptions that lead to a specific conclusion. A well-known downside is that these cores are typically quite big, which makes them hard to interpret for a human expert. Therefore, there is a need for methods that allow large explanations to be broken down into more understandable smaller pieces. Within the domain of causal reasoning, the concept of Actual Causation has been studied. Originating from the philosophy literature, this concept tries to identify the most relevant causes among all contributing factors for a given effect. Seminal work by Joseph Halpern and Judea Pearl has studied it in the context of AI systems.

The goal of this PhD is to investigate how concepts from the causality literature can improve the explainability of logical reasoning systems, as a possible component of trustworthy neurosymbolic AI. There is both a theoretic and practical component to this topic, since the ultimate goal is to implement efficient algorithms that allow reasoning engines to produce explanations that are in line with the philosophical insights about causality. This will be joint PhD with the nearby KU Leuven university, offering you the chance to work with experts from two highly ranked universities at the heart of the EU.

We offer:

• A stimulating and lively scientific environment
• The opportunity to carry out research at the highest international level, aiming for excellence in research with impact on society
• The opportunity to work in internationally renowned research groups at VUB and KU Leuven university

For this function, our Brussels Humanities, Sciences & Engineering Campus (Elsene) will serve as your home base.

3 - Profile

What do we expect from you?

You have obtained (or will obtain in 2026) a master’s degree in artificial intelligence, computer science, mathematics or a related domain. You have a solid academic track record, at least at the cum laude level.

• Additional background in philosophy or (cognitive) psychology may be considered an asset.
• You are interested in Knowledge Representation & Reasoning, Symbolic/Logic-based AI methods. You have good mathematical and programming skills.
• You strive for excellence and have a scientific mindset.
• You are a loyal team player, who can work autonomously and deliver solid scientific work.
• You have strong communication skills in English. Proficiency in (or willingness to learn) Dutch is considered an asset.
• You have not performed any works in the execution of a mandate as an assistant, paid from operating resources, over a total (cumulated) period of more than 12 months.
• As a (non-)EE