Semantic Ambiguity in Expert Systems: The Case of Deterministic Systems

Typically with expert systems, the user is asked to assess or categorize evidence (E), based on data (D). The system is then responsible for matching that evidence (E) to the appropriate hypothesis (H). Throughout, the user provides the characterization of data as evidence for the system. Unfortunately, different users may develop different categorizations of data. Based on different points ofview, different users may categorize D as E or E'. This is referred to as semantic ambiguity. Such potential differences can have a major impact on the results recommended by expert systems. This paper presents empirical results that indicate that there can be substantial ambiguity about the categorization of D as E or E'. In one situation, 50%of the subjects categorized data as E and 50% as E'. This paper also develops a probability model that is used to investigate semantic ambiguity in deterministic systems (no weights on the rules).