Information Refinement and Revision for Decision Making: Modeling for Diagnostics, Prognostics, and Prediction
Papers from the AAAI Spring Symposium
Kai Goebel and Piero Bonissone,Cochairs
Many companies have discovered the value of preserving and maintaining their corporate knowledge as they are collecting large amount of process data and business information. This collection is accelerated by the use of advanced and less expensive sensors, massive information storage, and internet-facilitated access. As a result, diagnostic decision makers are faced with the daunting task of extracting relevant morsels from this information hodge-podge, dealing with conflicting information, repudiating stale and outdated information, and evaluating the merits of a found solution. Automated decision-making systems also need to heed the effect of degrees of redundancy in the information considered, which may skew the decision pursued. In addition, temporal effects play a major role in the decision making process not only because information integrity fades over time but also because new information needs to be factored in. Although this new information does not exist at the time of the system design, one must provide a system maintenance plan to account for it. Ways to judge the relevance of this new information and optimization issues need to be discussed in this context. Finally, the quality and uncertainty of the newly found system and its resulting decisions need to be evaluated.