Computing Reference Classes
This work addresses a foundational issue in AI and statistics for researchers and practitioners dealing with uncertainty and evidence combination, but it is incremental as it builds on existing philosophical frameworks.
The paper tackles the problem of determining appropriate reference classes for evidential reasoning in systems with limited statistical knowledge, and it provides an analysis of the computational feasibility and necessary components of Kyburg's rules for reference classes.
For any system with limited statistical knowledge, the combination of evidence and the interpretation of sampling information require the determination of the right reference class (or of an adequate one). The present note (1) discusses the use of reference classes in evidential reasoning, and (2) discusses implementations of Kyburg's rules for reference classes. This paper contributes the first frank discussion of how much of Kyburg's system is needed to be powerful, how much can be computed effectively, and how much is philosophical fat.