Situated cognition in the semantic web era

The challenge of situated cognition mounted by Clancey and others 20 years ago seems to have had little impact on the technical development of AI systems. However, the hopes for the Semantic Web also seem far from being realised in much the same way as too much was expected of expert systems, and again this seems to be because of the situated nature of knowledge. In this paper we claim that a possible way forward is to always ground the use of concepts in real data in particular contexts. We base this claim on experience with Ripple-Down Rule systems.