Abstract:Measures of voting power have been the subject of extensive research since the mid 1940s. More recently, similar measures of relative importance have been studied in other domains that include inconsistent knowledge bases, intensity of attacks in argumentation, different problems in the analysis of database management, and explainability. This paper demonstrates that all these examples are instantiations of computing measures of importance for a rather more general problem domain. The paper then shows that the best-known measures of importance can be computed for any reference set whenever one is given a monotonically increasing predicate that partitions the subsets of that reference set. As a consequence, the paper also proves that measures of importance can be devised in several domains, for some of which such measures have not yet been studied nor proposed. Furthermore, the paper highlights several research directions related with computing measures of importance.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is to propose a general method for calculating the relative importance (or "importance measure") of elements in different fields. Specifically, the paper shows how to calculate the best - known importance measure for any reference set by given a monotonically increasing predicate. This method is not only applicable to existing application fields, but also can be extended to new fields where such measures have not been studied or proposed yet.
### Main contributions of the paper
1. **General framework**: The paper proposes a general framework for calculating the relative importance of elements in various application fields. This framework is based on monotonically increasing predicates and can be applied to multiple fields such as weighted voting games, inconsistent knowledge bases, argument strength, database management, and interpretable artificial intelligence.
2. **Exploration of new application fields**: The paper points out that importance measures can be envisaged in many practical application scenarios, but the measurement methods in these application scenarios have not been studied or proposed yet. For example, in model - based diagnosis, components can be ranked in importance according to the system's fault behavior; in an inconsistent system of linear inequalities, relative importance can be assigned to inequalities.
3. **Unification of existing measures**: The paper unifies several existing importance measures (such as Shapley - Shubik index, Banzhaf index, and Deegan - Packel index) under one framework and shows how these measures can be applied to broader scenarios.
4. **Algorithms and implementation**: The paper discusses algorithms for calculating minimal sets and points out that these algorithms can be used to calculate and enumerate minimal sets without relying on specific application fields. In addition, the paper also explores the exact calculation of these measures and their approximate calculation methods.
### Definition of importance measure
For a monotonically increasing predicate \(P: 2^N\rightarrow B\), where \(N = \{1, 2,\ldots,m\}\) is the set of elements and \(B=\{0, 1\}\) represents the result of the predicate, the paper defines the following concepts:
- **Characteristic function**:
\[
\upsilon(S):=\text{ITE}(P(S), 1, 0)
\]
where \(\text{ITE}\) is a conditional operator, which means that if \(P(S)\) holds, it returns 1, otherwise it returns 0.
- **Relative influence**:
\[
\Delta_i(S):=\upsilon(S)-\upsilon(S\setminus\{i\})
\]
This value measures the importance of element \(i\) to set \(S\).
- **Importance measure**:
\[
\text{Sc}(i):=\sum_{S\subseteq N}\varsigma(S)\times\Delta_i(S)
\]
where \(\varsigma(S)\) is a weight factor, depending on the type of measure considered (such as Shapley - Shubik, Banzhaf, etc.).
### Case analysis
The paper uses two specific cases to illustrate how to apply these measures:
1. **Dominating set problem**: In graph theory, calculate the relative importance of each vertex in the dominating set.
2. **Sadaka training problem**: In the Dune universe, calculate which training items have the greatest impact on soldiers being eliminated.
### Conclusion
By introducing a general framework, the paper shows how to calculate the relative importance of elements in multiple application scenarios. This framework not only unifies existing importance measure methods but also provides new directions for future research, especially in those unexplored importance measure fields.