Previous: Roles of Knowledge Level Modelling Up: Issues in Knowledge Level Modelling Next: References

Conclusion

The purpose of this paper was to bring together views and ideas on knowledge level modelling for knowledge systems. The main points it has been making are:

The implications of the above points for computational architectures of knowledge based behaviour were briefly explored. It is argued that explicit problem solving methods aim at making explicit the control in problem solving and that this leads to a form of brittleness because of lack of flexibility. On the other hand the notion of case model is not fully exploited. It is suggested that newer architectures will make explicit the goal of problem solving as modelling, making control emergent from the operationalisation of the method's principle of rationality.

With respect to the role of knowledge level modelling in knowledge systems development three possibilities have been listed that typically go together in methodologies. However, the combination of these roles hides an implicit assumption which is not necessarily valid. As an alternative I have suggested the use of knowledge level modelling as a way to find out relevant characteristics (task features) of a problem situation, as a pre-requisite for the construction of a solution for that (organizational) problem situation. This can be done in a variety of ways, including the introduction of a knowledge system.

Viewing knowledge as a means of modelling behaviour (not necessarily of generating it) calls for a re-orientation of knowledge engineering. Knowledge systems are rarely problem solvers themselves. They support a process of agents interacting with the environment and building their case model of the problems and situations that they are faced with. Understanding how to support this process requires more than knowledge acquisition alone. Moreover it can be supported in other and maybe better ways than by building a knowledge system. Knowledge engineering must be concerned with analysing and shaping knowledge based processes. As this paper has tried to show, the knowledge perspective remains an important tool to do this, but it seems that some of the issues that it raises (e.g. the role and nature of the 'grounding' relation in Fig. ) must be taken more seriously.

walter@arti