Color is a subject which has always fascinated scientists and artists alike and it is therefore an excellent vehicle for forging relationships and deepening the dialog between the two cultures. The artist presents the results of his or her investigations by works that exercise or shake up the observer's routine visual experiences. They thus provide deep insights into the visual and cognitive processing spontaneously performed by a viewer: Naturalistic painters used sophisticated interplay of light and darkness to recreate the appearance of three dimensional objects. Impressionists like Seurat attempted to paint the raw perception falling on the eye. Fauvists destabilised natural expectations by using colors dissociated from their sources: faces are orange, water purple, trees pink.

The natural scientist, in the footsteps of Newton, sees color as a physically observable abstract dimension of reality which can be measured and investigated objectively by careful experimentation and mathematical modeling. The chemical synthesis of color pigments, the digitisation and display of color on computer screens, the optics of prisms and rainbows, are among the towering achievements of the natural science approach to color.

But we know since Goethe, and through the work of artists, that the human experience of color is quite different from its physical properties. Many beautiful psychological experiments show that color perception is influenced by prior expectations, habituation, and world knowledge, proving that the same physical colors can give rise to very different color perceptions and emotional states. The languages of the world appear to organise and categorise the color space in surprisingly different ways, illustrating that color categorisation and naming are at least partly cultural constructions as opposed to objective measures. Recent research in artificial intelligence and neuroscience has shown that the cognitive processing of color is a vastly complex affair which brings many brain areas into play. These various insights in human color experience run parallel to the insights of artists obtained by introspection and experimentation with paintings, sculptures and other media.

But all this raises many puzzles: What is the interaction between the physical and the cultural? How does the eye and the brain manage to produce before our consciousness the stark colored interpretation of the world that we experience? Are there any intermediary steps in processing? How can a culturally shared set of color perceptions, color categories, and names arise without a central coordinator or without color being innate? If color is not a pre-given abstract dimension how does color emerge?

In my own work, I use novel methods to explore these issues. In particular, I use autonomous robots endowed with sensors and computing power as experimental apparatus. The robots engage in language games and are instructed to develop jointly and from scratch a language and its underlying conceptualisation. The language should enable them to talk about objects in their world. Color is not given as an a priori dimension but emerges through cultural construction and transmission. Our experiments use many concepts from biology: self-organisation, co-evolution, structural coupling, synergy. They show how cultural evolution might happen concretely and how the complexities of color can become gradually understood in a way that links the physical and biological world to the cognitive and cultural world.


Luc Steels studied philosophy and linguistics at the University of Antwerp, Belgium and computer science at the Massachusetts Institute of Technology. In 1983 he founded the Artificial Intelligence Laboratory at the University of Brussels (VUB) where he currently serves as director of the laboratory and Professor of Artificial Intelligence. He has been a leading researcher in the area of behavior-based approaches to artificial intelligence through the experimental use of physical robots. He also serves as director of the Sony Computer Science Laboratory in Paris where his focus is on the origins of language and the emergence of grammar in autonomous robotic agents. His recent published contributions to the field artificial intelligence include The Artificial Life Route to Artificial Intelligence: Building Situated Embodied Agent, Lawrence-Erlbaum Association, New Haven, 1994 and The Biology and Technology of Intelligent Autonomous Agents, Springer Verlag, Berlin, 1995.