SUPERVISOR: PROF.DR. FRANK VAN DER VELDE; 25/35EC
Concepts and their relations play a crucial role in human cognition. In particular, they are the building blocks of our semantic cognition, with which we understanding our environment. Concepts can vary from concrete, as given by the concept "dog", to abstract, such as the concept "honesty". Learning concepts can be based on learning perceptual classifications, such as learning the concept "dog" from classifying individual dogs, or by classifying or recognizing actions as performed by certain agents. But concepts can also be learned by combining other concepts and their relations. So, the concept "animal" could be learned from understanding the similarities between concepts such as "dogs" and "cats" and their differences with other concepts like "chair" or "house". In this way, we also learn relations between concepts, for example that a dog is an animal, but not every animal is a dog. Because concepts (such as actions) are typically learned in (certain) relations to each other, a 'conceptual space' (or knowledge base) can arise, which forms the basis for our semantic cognition.
How we learn concepts and conceptual spaces, and how they are represented in the brain, is a topic of very active research. Learning of concepts and relations is also an important theme in machine learning. The key issue in this project concerns the way in which concepts and their relations in a given domain are learned and how they are combined to form a conceptual space. The domain can be chosen one, such as the "sport domain" (with concepts like "player" or "game") or the "health domain" (with concepts like "virus" or "medicine"). Or it could be designed for the project to study how humans learn such a new domain. The chosen topic can be studied with experimental techniques such as card sorting or priming studies. Or the conceptual space in a chosen domain could be designed (e.g., for use in machines) and evaluated by humans, for example by using questionnaires. Aspects of concept learning and conceptual spaces can also be modeled with computer modeling, such as Deep Learning or other techniques.
This is a general topic that can be specified into a 25EC or a 35EC thesis project.
Rouder, Jeffrey; Ratcliff, Roger (2006). "Comparing Exemplar and Rule-Based Theories of Categorization". Current Directions in Psychological Science. 15: 9–13. doi:10.1111/j.0963-7214.2006.00397.x.
Lambon-Ralph, M. A., Jefferies, E., Patterson, K. and Timothy T. Rogers, T. T. (2017). The neural and computational bases of semantic cognition. Nature Reviews Neuroscience 18, 42–55. doi:10.1038/nrn.2016.150
Huth, A. G., de Heer, W. A., Griffiths, T. L., Theunissen, F. E., & Gallant, J. L. (2016). Natural speech reveals the semantic maps that tile human cerebral cortex. Nature, 532(7600), 453-458. doi:10.1038/nature17637