Automated Reasoning in Social Choice Theory - Some Remarks

Our objective in this note is to comment briefly on the newly emerging literature on computer-aided proofs in Social Choice Theory. We shall specifically comment on two papers, one by Tang and Lin (2009) and another by Geist and Endriss (2011). We also provide statements and brief descriptions of the results discussed in this note.

💡 Research Summary

This paper provides a concise yet thorough commentary on the emerging literature that applies computer‑aided proof techniques to Social Choice Theory, focusing specifically on two influential works: Tang and Lin (2009) (hereafter TL) and Geist and Endriss (2011) (hereafter GE). Both papers share a methodological backbone: they reduce an arbitrary‑size impossibility problem to a “small” base case and then verify that base case using automated reasoning tools. The present note analyses how this reduction works, what new insights it yields, and where the approach meets its limits.

TL studies various versions of the classic Arrow aggregation problem. Using conventional induction, TL shows that if a non‑trivial Arrovian Social Welfare Function (ASWF) exists for any number of voters and alternatives, then a non‑trivial ASWF must also exist for the minimal instance with two voters and three alternatives. The authors then encode the properties of an ASWF (Independence of Irrelevant Alternatives – IIA, Weak Pareto – WP, Non‑Imposition – NI, etc.) as a Constraint Satisfaction Problem (CSP). Instead of enumerating all 36 × 6 possible rules for the base case, they employ a depth‑first search with back‑tracking that prunes any partial rule as soon as it violates an axiom. This CSP‑SAT approach makes the verification tractable: the whole search finishes in a second on a standard processor.

The most striking result of TL is Theorem 5, which proves an Arrow‑type impossibility using only IIA, without any Pareto‑type axiom. Among the 36 possible Arrovian aggregators for the 2‑voter, 3‑alternative case, only 9 × 4 satisfy IIA, and all of them are either constant, dictatorial, or inverse‑dictatorial. This demonstrates that IIA alone is powerful enough to force a dictatorship‑type conclusion, a fact that would have been hard to conjecture without exhaustive computer search. TL also reproduces Wilson’s theorem (IIA + NI ⇒ null, dictatorial, or anti‑dictatorial) and Sen’s “Impossibility of a Paretian Liberal,” and it generates a number of new impossibility theorems by exploring novel combinations of axioms.

GE builds directly on TL’s framework. It first proves a “Preservation Theorem” that guarantees the reduction from the general case to the base case for a wide class of axiom systems. Then, using the same CSP‑based search, GE verifies 84 results, including the Kanai‑Peleg theorem, several extensions of Wilson’s theorem, and many previously unknown impossibility statements derived from unconventional axiom bundles. The authors also provide manual proofs for a subset of the new theorems, illustrating how automated search can guide human insight.

Beyond the technical contributions, the note discusses the broader significance and the inherent constraints of this approach. Strengths: (1) The CSP formulation turns abstract social‑choice axioms into concrete logical clauses that modern SAT/SMT solvers can handle efficiently. (2) The reduction technique isolates the combinatorial core of the problem, making exhaustive verification feasible where traditional analytical proofs would be cumbersome. (3) Automated search can uncover surprising patterns—such as the decisive role of IIA in Theorem 5—that may escape human intuition.

Limitations: (i) Not every social‑choice problem can be expressed as a finite CSP. For instance, characterising the Borda social welfare function requires reasoning about real‑valued scores that cannot be captured by propositional variables alone. (ii) The tractability of the base case hinges on its size; even modest increases (e.g., three voters or four alternatives) cause an exponential blow‑up that overwhelms current back‑tracking algorithms. (iii) Domains involving randomisation, cardinal utilities, or divisible commodities (auction design) inherently lack finiteness and thus fall outside the reach of current SAT‑based methods.

The authors outline several promising research directions where automated reasoning could be fruitfully applied. Dictatorial domains: recent work shows that the class of domains forcing dictatorship is surprisingly large, yet a full characterisation remains open. Since reductions from the general case to two voters are known, a TL‑style CSP search could help map the landscape of such domains, though a more sophisticated reduction for the number of alternatives is needed. Single‑peaked preferences: median‑voter rules are more intricate than dictatorships, and proving a reduction on the number of alternatives is challenging. Automated exploration of small instances may suggest conjectures about median‑rule characterisations. Strategic voting models: the Gibbard‑Satterthwaite and Muller‑Satterthwaite theorems are already amenable to CSP encoding; extending this to richer strategy‑proofness concepts (e.g., participation, monotonicity) could yield new impossibility results. Finally, the note stresses that even when a full reduction is impossible, generating and analysing small counter‑examples computationally can provide valuable intuition for the general theory.

In conclusion, the TL and GE papers represent valuable milestones demonstrating that automated reasoning can become a standard tool in the social‑choice theorist’s toolbox. Their success hinges on (a) the ability to reduce a general impossibility problem to a finite, manageable base case, and (b) the availability of efficient CSP/SAT solvers. Future work should aim to broaden the class of social‑choice problems that admit such reductions, incorporate richer logical frameworks (e.g., SMT with arithmetic), and develop hybrid methods that combine automated search with human‑guided abstraction. By doing so, the community can push the frontier of what is provably possible—or impossible—in collective decision‑making.