About Human-centered Artificial Intelligence

Interdisciplinarity

Since it has human intelligence as its main object of study, Artificial Intelligence research is inherently interdisciplinary. Traditionally, AI has built on knowledge, models and tools from computer science, philosophy, psychology and linguistics. But with the success and growing applicability of AI come new close relations with many other disciplines such as data-science, neuroscience, biology (bio-informatics), physics (complex systems), educational science, legal science, social science, and economics (game theory).

Societal relevance

Artificial intelligence is a fast-growing scientific discipline​ that has already changed our daily lives in fundamental ways. Tasks traditionally performed by humans are executed cheaper, more reliably and more accurately by AIs. Therefore, decisions in the stock markets, by the cars on our roads, and in diagnosing diseases in our hospitals, will increasingly be made by algorithms. Some even envision that AI is going to play an important role in education, where it may, for instance, provide student specific learning methods. All these developments require us to think much more closely about design principles for responsible AI, the role of AI in society, and the moral and legal implications of the use of AI.

All these developments require us to think much more closely about design principles for responsible AI, the role of AI in society, and the moral and legal implications of the use of AI.

Impact on other scientific disciplines

Less known is that AI also has a deep impact on other scientific disciplines. The capabilities and technologies that are provided by artificial intelligence, such as machine learning, data mining, agent-based simulation, optimization, planning, automated reasoning, vision, natural language processing and robotics, have initiated and accelerated investigations on a wide variety of phenomena in the exact sciences, medical sciences, the humanities, and the social and behavioral sciences.

Finally there is the economic impact of AI. Future growth is estimated to be largely driven by new technologies involving AI. AI is seen by many as the technology that determines a country’s future economic competitive edge.

Within the Dutch research landscape determined by the ‘Dutch science agenda’, the above forms of impact point to the direct relevance for the following routes: Smart Industry; Smart liveable cities; Neurolabnl; Energy transition; Logistics and transport in an energetic, innovative and sustainable society; Value creation through responsible access to and use of big data; Living past.

The European union also recognizes the societal and strategic importance of AI, and has earmarked very significant amounts for financing future AI research. Through European initiatives like 'CLAIRE' and the 'European AI Alliance' the Utrecht AI community stays in close connection to these developments.

Utrecht’s position in the AI landscape

AI in Utrecht has a unique interdisciplinary profile that pervades various departments including computer science, philosophy, linguistics and psychology.

Symbolic approach

Characteristic for AI research in Utrecht is its mainly symbolic approach. Symbolic artificial intelligence consists in a collection of top-down model-driven methods and approaches that are based on symbolic, often human-readable, representation systems. This branch of artificial intelligence assumes that symbolic systems have necessary and sufficient means for intelligent decisions and actions. Symbolic artificial intelligence exploits upfront domain knowledge, allows explicit symbolic search and reasoning, and is essential for the development of explainable and verifiable intelligent systems. As the new European initiative for AI funding puts it, ‘explainability’ is one of the most pressing challenges. Explainability cannot do without symbolic models. The same is true for getting a grip on accountability and responsibility in the context of intelligent autonomous systems.

Explainability profits from a systematic integration between symbolic and sub-symbolic models

The emergence of big data, availability of computational resources, and the advance in machine learning have recently transformed research in this branch of artificial intelligence towards what is now called data-driven AI. Utrecht is currently expanding its AI research towards explainable data-driven AI approaches and models. We accordingly believe that the Utrecht AI profile is uniquely positioned to address the newly upcoming questions in data-driven AI.

AI Research excellence in Utrecht

Utrecht University hosts several outstanding researchers in AI. This excellence is reflected in the high number of NWO and ERC grants that members of the KI/AI teaching programmes have received over the last ten years. These include three ERC grants (consolidator and advanced), two VICI grants, four VIDI grants, as well as other significant grants like the Marie Curie reintegration grant, and NWO Free Competition and large investment grants.