On the logical and physical combination of evidence in intelligent machines

✍ Scribed by S. Mangiaracina; G. Beni

• Publisher: John Wiley and Sons
• Year: 1986
• Tongue: English
• Weight: 627 KB
• Volume: 1
• Category: Article
• ISSN: 0884-8173
• DOI: 10.1002/int.4550010205

✦ Synopsis

We give a rigorous interpretation of the ways of combining evidence in Mycin-type expert systems. We demonstrate the probabilistic inexactness of the Mycin model and its incompleteness for combining evidence from quantitative data extracted in well-defined physical systems. Finally, we present a new method of combining evidence including both logical and physicaf combinations of evidence. The method is applicable to intelligent machines capable of gathering evidence via sensors. managing uncertainty of data acquired by robots or other intelligent machines. '

Intelligent machines make decisions on the basis of sensory data and these decisions lead to well defined changes of their state. For this reason, the decision process of an intelligent machine must have a quantitatively output (e.g., the new position of a robot arm).

Actually, expert systems can output quantitative answers, but if they do so it is without justification. (The only consistent approach that manages uncertainty of meaning is the one that inputs "fuzzy" statements and outputs "fuzzy" statements.') In contrast, the great majority of expert systems input nonquantitative statements, manipulate them mathematically, and output a quantitative answer. However, it is not obvious that the mathematical operation used are applicable, nor that the quantitative results generated from qualitative inputs have any meaning. A new approach is required in order to manage the uncertainty of measurements that intelligent machines need in order to make decisions.