CARTER and Conflict Resolution

We have developed a method and prototype program called CARTER that helps two experts to agree on what knowledge should go into a single consensus knowledge base. We believe that the ideas in that prototype are important to any effort that aims to build large knowledge bases.

CARTER's knowledge is organized in a catalog currently containing 35 entries, each of which in turn consists of a discrepancy detection procedure and a corresponding resolution procedure. This simple detection-resolution organization of the catalog facilities adding new entries as we gain more experience with the task.

Among the types of discrepancies that CARTER can recognize and repair are:

1. Differences in the character of the result: one system is content to classify the problem (e.g., specify the nature of a defect in the data, like Heteroscedasticity), while the other both classifies the problem and then goes on to suggest what to do about it (e.g., do a Log-Transform).

2. Differences in vocabularies: one expert refers to the Defects of a regression model, while the other refers to its Problems, but they are referring to the same thing. Other forms of vocabulary discrepancy the system knows about include differences in representation choice (e.g., one expert represents a concept as an attribute, while the other represents it as a value) and missing terms (e.g., one KB contains values missing from the other).

3. Differences in pattern of inference: the experts agree on the overall vocabulary, but interconnect the terms differently, as in the case where one expert uses only an F-test statistic to judge the quality of a model, while the other relies on both an F-test and an S-squared statistic.

4. Differences in the rules: the experts agree on the vocabulary and pattern of interconnection between terms, but write different rules. For instance, one expert has a rule that an F-test result below a specified threshold indicates that the Quality of the Model is Poor, while the other reasons from the same evidence that the Quality of the Model is Fair. Both reason from the value of the F- test to the Quality of the Model, but they use different rules. Another form of rule discrepancy occurs when two otherwise identical rules have differing levels of certainty.