Blackboard architectures in computer aided engineering

Blackboards (BB) are global memory-resident databases, accessible through a uniform protocol and tailored for expert source cooperation. They have been developed for continuous speech understanding (Hearsay-II~), and successfully used in signal processing (HASP-SIAPZ), civil engineering design (Destiny3), and general expert systems shells (BB14). In a way, any expert system working memory is a kind of blackboard: rules can be considered as small expert sources, and they all look at the working memory simultaneously. However, the intelligence and activity of the working memory is very small, and it does not fully qualify for the 'Blackboard label'. For that reason, we claim that one should focus on blackboard architectures more than blackboards per se. What appears essential in a 'real' blackboard architecture is that it is especially tailored to manage complex cooperation of actors. This in turn requires some distinctive features such as uniformity of representation of the data, powerful structuring possibilities and a rich agenda-based monitoring of problem solving. In this paper, we shall further investigate the characteristic features of blackboards, expose some of t he reasons why this architecture will be very useful for Computer Aided Engineering (CAE) expert systems, and report on some preliminary results we have already obtained in the domain. In particular, we show what can be gained by designing a reflexive blackboard architecture where the explanation component is a fullscale knowledge source.