Introduction: the explanation problem and reconstruction Reconstructability analysis [1,2] in systems theory has put emphasis on the issue of system reconstruction from subsystems (namely, a set of variables). Yet there is another kind of reconstruction problem which resides in the interests of systems studies, namely, explanation.Following the discussion given in[3], the explanation problem in an information processing system can be defined as "an information processing operation that takes the operation of an information processing system as input and generates a description of that information processing operation as output". Therefore, explanation provides a description of how the solution of a problem is achieved by the system, thus enhancing the user's confidence of the system.In terms of terminology in systems theory, we propose the notion of viewing explanation as solution reconstruction. Since such a kind of reconstruction is concerned with solutions rather than the systems, it is different from system reconstruction. Nevertheless, it still demonstrates the basic intuitive idea of "pasting parts (pieces) together to get the whole thing". Notice that the term reconstruction implies that the way of putting pieces together does not necessarily follow the way of splitting the original thing apart; rather it is reconstructed.Explanation is a complicated issue and there are many aspects of explanation. To make our study focused, we will examine the notion of explanation as a "solution reconstruction" in knowledge-based systems, explaining how knowledge-based systems reason involves presentation, user factors, and the way systems understand their own problem-solving knowledge and strategies. In this short article, two types of systems in the knowledgebased arena will be examined briefly, namely, causal networks and expert systems. In both fields, explanation itself is treated as a problem-solving process; the task of this process is to describe how the original problem is solved by the system.
Reasoning and explanation in causal networksThe concept of a causal network or Bayesian network is due to , and has found many applications [7,8]. In this short paper we will follow the discussion of [6]. After a brief description of the concept, we will take a look at the relationship between reasoning and explanation in causal networks.