Solution Concept, Rule, and Correspondence
For a fixed population $N$ and alternative space $A$, three closely related shapes describe ways of selecting alternatives from a social choice instance $I = (F, P)$ ([[social_choice.instance]]).
-
A solution concept is a predicate $\sigma : \mathrm{Instance}\, N\, A \to A \to \mathrm{Prop}$ marking the alternatives that the concept deems acceptable in $I$. In Lean:
SolutionConcept N A := Instance N A → A → Prop. -
A rule is a deterministic single-valued choice that, for each instance, returns a feasible alternative: $\rho(I) \in \{a \in A \mid I.\mathrm{feasible}\,a\}.$ In Lean:
Rule N A := (I : Instance N A) → {a : A // I.feasible a}— the subtype carries the feasibility witness. -
A correspondence is a set-valued choice: $\Phi : \mathrm{Instance}\, N\, A \to \mathcal{P}(A).$ In Lean:
Correspondence N A := Instance N A → Set A.
These three shapes cover the standard styles of social choice output: a predicate (e.g. "is a stable outcome"), a function (a deterministic rule with a feasibility guarantee), and a set-valued map (a Pareto correspondence, a choice correspondence, and so on).
Specializations appear at later layers: the voting layer instantiates solution concepts and rules into social welfare functions and social choice functions ([[social_choice.voting.swf]], [[social_choice.voting.scf]]); fair division supplies rules returning a feasible allocation.
References
- [MSZ, Chapter 21] Maschler, Solan, and Zamir, Game Theory. Choice rules and correspondences.