AI research at Tilburg University

Blending AI with cognitive science

Whereas AI research at Tilburg University gained prominence in the past through computational linguistics, the university’s current AI research is strongly integrated with cognitive science. It focuses on topics such as deep learning architectures for multimodal data, human-AI interaction, and gaming and robotics.

At first glance, artificial intelligence (AI) appears to be a hardcore computer science subject. It is no coincidence that AI is part of the computer science faculty at most universities. But at Tilburg University, AI is integrated with cognitive science and part of the Tilburg School of Humanities and Digital Sciences.
‘Historically speaking, it is quite natural to combine AI and cognitive science’, says head of the Department Cognitive Science and Artificial Intelligence, Marie Postma. ‘When AI emerged as a discipline in the 1950s, traditional cognitive science themes inspired the goals of AI: machine translation, decision making and problem solving, modelling human intelligence in a machine. The development of neural networks, which are nowadays very successful, grew out of the idea of creating an electronic brain.’
Another reason why AI and cognitive science are blended in one department at Tilburg University is the university’s focus on the social sciences, says Postma. ‘As far as AI is concerned, we are focusing on research and applications of AI where the human user plays an active role or where the data come from humans. The combination bears fruit in attracting new researchers. When we ask new group members what attracted them to us, they most often mention the combination of AI and cognitive science.’

Coaching-oriented culture

Research in the department focuses on four main themes, with a lot of collaboration and overlap between them: computational linguistics, deep learning for perception, human-AI interaction, and autonomous agents, robotics and gaming.
Of these four, computational linguistics has the longest history in Tilburg, dating back to the 1980s. Postma: ‘In the meantime, research within computational linguistics has become multimodal. Many of our projects no longer work with just speech and text data, but also with video data. For example, in the Horizon2020 project SignOn, we are investigating how to use machine translation to translate between sign and non-sign languages, facilitating communication between the deaf and hard-of-hearing and the hearing community.’
The department was established in 2017 and so the research group is still very much in the development phase. ‘We have grown rapidly’, says Postma. ‘The majority of our researchers are between 30 and 45 years old. My ambition is for young scientists to be able to develop their own research line, despite the fact that they have substantial teaching obligations. To achieve this goal, we have deliberately developed a non-hierarchical coaching-oriented culture. Through MindLabs, a partnership between knowledge institutions and other partners in the region, our young researchers can also become involved in a variety of projects such as the ICAI MasterMinds.’

VR training for pilots

The department is of an international nature, with about seventy percent of the people coming from abroad. Associate professor Maryam Alimardani, working in the Human-AI interaction group, is one of them. She spent thirteen years in Japan before coming to the Netherlands.
About the international flavour of the department, she says: ‘We have our very own, diverse culture, which is open, inclusive and respectful. In Japan, I always remained a foreigner in a group dominated by Japanese, whereas I came from Iran. And although Dutch culture overall is very mixed and diverse, I hear from expats in other groups that the higher the number of Dutch in the group, the more rigid the group culture. In our group, we feel a sense of belonging from day one because the majority is from somewhere else.’
Alimardani works in the field of brain-computer interfaces. ‘I collect human brain signals and translate them into information about user experience during interaction with a social robot, a computer or virtual reality. In one of our projects, we collaborate with the Royal Netherlands Airforce. Via a headset, we collect brain activity from pilots who are trained in a virtual reality simulation. The goal is to monitor their workload and give personalised feedback during the training sessions.’

AI to help radiologists

Sharon Ong is an assistant professor in the same department, working in the Deep Learning for Perception group. Ong specialises in medical image analysis. She was educated in Australia and has been a research scientist in Singapore. ‘I very much like the fact that the department is still under development. It gives more new opportunities compared to joining an established department.’
Ong is working on two projects that are a collaboration between the Elisabeth-TweeSteden Ziekenhuis (ETZ) in Tilburg and Tilburg University. The aim is for the results and conclusions from the various projects to be applied in practice. Ong: ‘In one project, we are creating an AI tool to detect bone tumours and monitor their progress. In a second project, we use AI to predict a patient’s cognitive functioning after a brain tumour has been removed. The second project, in particular, is a strong combination of cognitive science and AI. As an AI person, I work alongside cognitive science people. I want to develop and implement new AI tools that help radiologists and other medical practitioners to improve patient care.’

Group passport

Research fields:
  • computational linguistics, deep learning for perception, human-AI interaction, autonomous agents, robotics and gaming.
Institution
  • The Department Cognitive Science and Artificial Intelligence is part of the Tilburg School of Humanities and Digital Sciences at Tilburg University.
Labs
  • Virtual, mixed & augmented reality lab, Natural language and data processing lab, Robotics & avatars lab, Serious games & learning lab, DAF Technology Lab
Employees as of September 2022
  • 3 full professors, 38 assistant/associate professors, 4 lecturers, 4 postdocs, 19 (internal) PhD students.
Websites

Published in I/O Magazine #4 2022
Text Bennie Mols
Images Ivar Pel

November 28, 2022