How would you describe your work in 90 seconds?

I investigate formal models of normative reasoning. This is a type of reasoning involving normative concepts. Think of laws, rules, values, rights, obligations, and permissions. These concepts influence our day-to-day reasoning and decision-making. The investigation of this type of reasoning from a formal perspective is motivated by the fact that we now have formal systems, like AI systems, making decisions and drawing conclusions that directly or indirectly have an impact on our lives. We want to make sure that these systems can account for certain norms and values in the reasoning process. This is part of AI alignment, which is the quest for aligning AI behavior with societal norms and values. There are three immediate questions that come up in this context: what would these norms and values be that we are imposing on AI? How can we represent these norms and values in formal systems? And lastly, how can we assess whether an AI system is correctly aligned? In my research, I address these questions. For instance, I investigate how conflicts between and violations of norms influence reasoning and how to use artificial explainability methods to explain such reasoning processes. Additionally, I also work on methodology for AI ethics audits, which are methods for investigating whether an AI system is aligned with certain values and ethical principles.

How did you get in touch with informatics?

I started out with a bachelor’s in economics, where I took a couple of courses on business ethics and philosophy of science. There, I got quite excited about questions concerning the ethical impact and ethical aspects of business, which is why I decided to switch and study philosophy instead. I particularly focused on practical philosophy, which is the philosophy of action and agency, and ethics is an essential part of this. Through my studies I got to know about logic, and there was a professor at the University of Amsterdam who was interested in building mathematical models of metaphysical theories. I was quite intrigued by this and wondered, is it really possible to bring the worlds of ethics and logic together? That is how I found out about this research field that applies logical methods to normative reasoning (it’s called deontic logic) and decided to do my master’s degree in logic. I ended up applying for a PhD in computer science (at TU Wien) where there is quite an interest in technical ethical questions and formal normative reasoning, which requires philosophy and logic.

Where do you see the connection between your research and everyday life?

T he connection lies in the observation that norms and values are an essential part of what it means to be a human being. We, humans, have ideas about how we would like the world to be and develop values and principles, which are enforced socially or legally, to pursue these ideals and ultimately improve the quality of life. They have a profound impact on the world and shape it. We are now giving AI systems and formal systems similar rights by outsourcing decision-making that directly impacts our lives to such systems. The investigation of how we can formally integrate norms and values into these systems to ensure that they behave in accordance with our value systems is an essential question that directly ties to everyday life. This integration is, however, not trivial, and the biggest challenge that we are facing is how to transform general ethical principles and values into concrete, implementable solutions and testing methods. Of course, we humans often have quite different ideas of what these principles must be. In any case, we need to have a good technical understanding of principles and values and how they should behave in formal decision-making contexts.

What makes you happy in your work?

The first thing is that the type of research I do is very interdisciplinary. It involves policymakers, law experts, philosophers, mathematicians, computer scientists, and social scientists. We have this shared problem, and we need to come together to solve it. It’s really a collaborative effort. First of all, this means that we must overcome a language barrier: each of us is looking at the same phenomenon from a completely different angle, having our own vocabulary, methodology, and scientific background. A big part of this type of research is trying to understand each other and trying to find a common ground from which we can start addressing the problem. I find this very exciting. It’s quite time-consuming but once this common ground is reached, it’s highly rewarding to see what can be done together that none of us can do individually.

I also very much enjoy teaching. Especially teaching where there is a chance for interaction, exchange, and reflection on topics from different disciplinary angles, like philosophy and computer science. For instance, when teaching AI ethics, it is inspiring to see students improve their discussion and argumentation skills and present their own perspectives on ethical questions in AI convincingly. For me this is also great food for thought.

Why do you think there are still so few women in computer science?

This is an interesting question. I believe it is for a part due to certain gender expectations still being implicit in our culture, and these include expectations about what somebody’s life or career should and could look like. The main question for us is, how can we create an openly inviting and inclusive educational environment? A good starting point is the presence of role models explicitly showing that things can be different and that things are changing. In teaching, I believe, it is important to discuss a diverse and inclusive range of authors when introducing students to a field and its community. I believe it’s important to make clear, also as a university, that anybody who is interested in a certain field and who has the right qualifications is, by definition, a good candidate for studying and working in that field. Breaking biased expectation patterns is a big challenge for society in general, but as a university, we have an important role to play here, too.

Kees van Berkel is Assistant Professor at the Research Unit Theory and Logic at TU Wien Informatics.

Discover the whole #5QW series.

Curious about our other news? Subscribe to our news feed, calendar, or newsletter, or follow us on social media.