28 April 2025

On March 24, Dominik Drexler at the Department of Computer and Information Science (IDA) successfully defended his thesis about integrating learning and artificial intelligence (AI) planning. "Pursuing a PhD at IDA has been one of the most rewarding experiences of my life", he says.

Dominik Drexler headshot
Dominik Drexler

Congratulations on your PhD! What was your background when you started your PhD?

"Thanks a lot! I completed both my Bachelor's and Master's degrees in Computer Science with a specialisation in Artificial Intelligence at the University of Freiburg in Germany."

What is it like to be a PhD student at IDA?

"Pursuing a PhD at IDA has been one of the most rewarding experiences of my life. While the journey was certainly challenging, it was also incredibly enjoyable. One aspect I really appreciated was the minimal administrative burden on PhD students. Thanks to the excellent support from the administrative and technical staff, I could fully focus on my research and teaching. Being part of the Wallenberg AI, Autonomous Systems and Software Program (WASP) was another major benefit. It allowed me to visit universities across Sweden and connect with PhD students working in related areas. Overall, IDA offers a great environment for both personal and academic growth."

What is your next step?

"I recently started as a postdoc at LiU in the same research group. I really enjoy the research area I explored during my PhD and plan to continue developing it further, while also exploring new approaches and taking on more teaching and supervision responsibilities."

Dominik Drexler's summary of his thesis:

An objective in artificial intelligence is the development of agents that act intelligently to achieve a given goal. There are two approaches for addressing this task, often called System 1 and System 2, each with its strengths and limitations. System 1 learns from experience to derive the actions to achieve a goal. System 2 is based on a world model that enables reasoning to find the actions needed to achieve a given goal. System 1 is often considered intuitive and fast, while System 2 is considered analytical and slow. My thesis addresses the integration of System 1 and System 2 by learning subgoal structures from experience that split problems into subproblems that are efficiently solvable by System 2. For example, many problems that we frequently solve in our household, such as doing the laundry, watering the plants, and so on, could be addressed by our general method in an automated way without prior knowledge.

Contact

Organisation

Latest news from LiU

Northern lake

Higher methane emissions from warmer lakes and reservoirs may exacerbate worst-case climate scenario

Emissions of methane from lakes and reservoirs risk doubling by the end of the century due to climate change according to a new study from LiU and NASA. This in turn could raise Earth’s temperature more than suggested by current worst-case scenario.

A man on stage is addressing the audience

Strengths and challenges in research revealed

For the first time, LiU has conducted an evaluation of all research – its quality, the culture, and the conditions for further development. The evaluation has been carried out through self-evaluation and external review by experts.

Two women discussing in the lab.

Sperm molecules can predict IVF success

The sperm is not a passive supplier of genetic material to the egg. A study shows that certain molecules that come with the sperm, so-called micro-RNA, contribute to the development of the embryo several days after conception.