Research at LUCS

Concept formation

In order to understand how humans form new ideas and how machines could be made to learn new concepts, one needs adequate models of concept formation. Within cognitive science, symbolic models have competed with concept learning in neural networks. At LUCS, the focus is instead on the dimensions that make up concepts. Using 'conceptual spaces', the aim is to develop models of concept formation that are more psychologically realistic than the symbolic, and that allow faster learning than neural networks.

Learning

Since our environment constantly changes we need to learn about it in a flexible way. Cognitive science studies learning in many areas: how we discover and learn regularities in the environment from the consequences of our actions, how we learn to find our way around in the world, why children can learn a language whilst this is near impossible for animals, and how computers and modern technology can best be used to support and enhance various learning processes.

Language

In a cognitive setting, language is not studied as an independent system, but in relation to perception, memory, visualisation, and concept formation. When investigating the use of language, semantic and pragmatic aspects become more important than syntactical. There is also an interest in evolutionary scenarios for the emergence of language. Implications of the cognitive approach are pertinent to natural language communication with computers and robots, and to how we construe the relation between words and pictures.

Vision and eye-movements

The research carried out on vision ranges from computer simulations of basic visual processes, such as depth perception and object recognition, to empirical studies utilising data from eye-tracking equipment. Recordings of eye-movements are currently being used to study visual attention to gestures in face-to-face communication, and the similarities in visual and spoken information processing. Ergonomic applications focus on design evaluation in the industry, and gaze control of computer interfaces.

Cognitive design

This research is aimed at understanding the cognition underlying human interactions with artifacts, and applying that knowledge to the improved design of both artifacts and work situations. Research is currently being carried out on a number of ways of adapting technology to the ways in which humans think and act. One project focuses on computer interfaces tailored to individual differences in cognitive style, other work is directed at the educatioal uses of technology, and design that minimises human error.

Evolution

Cognitive processes in humans and animals are largely the result of biological evolution. These are processes that have evolved to solve certain ecological and social problems. By investigating the adaptation of thinking to various constraints in the world, and by formulating plausible evolutionary scenarios, we are furnished with an additional tool for arbitrating between competing theories and architectures. Besides being an important constraint on our speculations and theorising, evolutionary considerations also help us at the start of theory construction.

Robotics

The research in robotics is aimed at solving the problems facing an autonomous robot acting in real time in a realistic environment. The problems encountered are many, ranging from navigation to eye-hand coordination, and research is conducted in near collaboration with the work being done in vision and learning. By implementing computational models of such things as object learning and recognition, the research on robotics also functions as a testing ground for the theories being modeled.

Computer simulation

Simulation is not itself a topic of research at LUCS, but a tool used for modelling cognitive processes. Some that have been modelled so far are: stereo vision, motion detection, learning, and the evolution of linguistic meaning in a simple lexicon. A number of these models have also been implemented in our robots. Working with computer simulations ensures that our theories are fully specified and explicit, and allows us to be confident of the real consequences of our theories.

Intentionality

Cognitive science has inherited a number of problems from philosophy concerning the relation between the mind and the physical world. There is the question of how electric activity in the neurons of the brain can constitute thoughts, plans and fantasies. Or, what it would take for a computer to have a will? These problems have been given new dimensions with the development of both technology and neuroscience. The main problems concern how intentions, goal, motivations and action should be modelled.

Everyday cognition

Interest in the cognition of everyday activities has increased as the limitations and biases of more traditional experimental studies has become apparent. Experimental results often generalise poorly to situations outside the laboratory setting, partly because factors of vital importance to cognition have been excluded. Interaction with the physical environment is one such factor, social interactions is another. An emerging perspective views cognitive processes as distributed between actors and artefacts rather than as a phenomenon of the isolated mind.