Introduction
In collective decision-making, a group often faces the "discursive dilemma": a situation where every individual makes a perfectly rational choice, yet the aggregated group decision violates logical consistency. For instance, a committee might vote in favor of three separate policies that, when combined, violate a pre-agreed budget constraint. This paper addresses the fundamental problem of binary aggregation—where a group makes yes/no choices on multiple issues—and seeks to identify the exact conditions under which an aggregation procedure "lifts" rationality from the individual to the collective level.
The Framework: Rationality as Logic
The authors bridge the gap between social choice theory and formal logic by representing "rationality" through Integrity Constraints (ICs). These are formulas in propositional logic that define the set of feasible outcomes. For example, a constraint like $\neg(p_1 \wedge p_2)$ stipulates that it is irrational to accept both issue 1 and issue 2 simultaneously.
An aggregation procedure is considered "collectively rational" with respect to a specific logic fragment if, whenever every individual's ballot satisfies the constraints, the resulting group ballot is guaranteed to satisfy them as well.
Methodological Approach
The research employs an axiomatic approach to characterize these procedures. The authors adapt standard social choice axioms to the binary setting, including:
- Unanimity (U): If every voter agrees on an issue, the group decision must reflect that agreement.
- Anonymity (A): The identity of the voters does not affect the outcome; all voters are treated equally.
- Independence (I): The group decision on a specific issue depends solely on the individual votes for that issue, rather than on related issues.
- Neutrality (NI/ND): The procedure treats all issues (Issue-Neutrality) or all values (Domain-Neutrality) symmetrically.
By mapping specific fragments of propositional logic (e.g., conjunctions, bi-implications, or clauses) to these axioms, the authors prove mathematical equivalences that define the limits of collective rationality.
Key Insights: The Complexity Ladder
The core contribution of the paper is the discovery of a direct relationship between the complexity of logical constraints and the restrictiveness of the required aggregation rules. This "complexity ladder" reveals several critical thresholds:
1. Simple Constraints and Unanimity
If the integrity constraints are limited to "cubes" (conjunctions of literals, such as "Issue 1 must be YES"), the authors prove in Proposition 5 that the only requirement for collective rationality is Unanimity. As long as the group respects a consensus, the logical integrity of the outcome is preserved.
2. Symmetry and Neutrality
When constraints involve relationships like "Issue 1 must equal Issue 2" (bi-implications), the aggregator must satisfy Issue-Neutrality or Domain-Neutrality. These axioms ensure that the procedure does not have an inherent bias that would break the logical symmetry required by the constraints (Propositions 6 & 7).
3. The Robustness of Majority Rule
One of the most significant findings involves the Majority Rule. The authors demonstrate in Proposition 18 that majority voting is surprisingly robust: it remains collectively rational as long as the integrity constraints are restricted to "2-clauses" (formulas with at most two variables, such as $p_1 \to p_2$). This explains why majority voting is effective for simple dependencies but fails in complex scenarios like the Condorcet paradox, which involves 3-way logical cycles.
4. The Limit of Rationality: Generalized Dictatorship
To guarantee collective rationality across all possible propositional logic constraints, the authors prove in Proposition 10 that the procedure must be a Generalized Dictatorship. In this scenario, the group is forced to simply adopt the ballot of a single individual for every possible profile to ensure no logical contradictions occur.
Theoretical Limitations
The study identifies a fundamental challenge in characterizing axioms like Independence and Anonymity. Because collective rationality is a "pointwise" property—meaning it is evaluated on a single voting profile at a time—it cannot fully capture "inter-profile" axioms that describe how a procedure behaves across different voting scenarios. Furthermore, the current model is strictly binary; extending these findings to multi-valued choices or rankings remains a task for future research.
Conclusion
This work provides a rigorous logical foundation for understanding collective decision-making. By identifying the specific axioms required to preserve different types of logical consistency, the authors offer a roadmap for designing aggregation procedures that are both democratically fair and logically sound. The findings suggest that while simple dependencies can be handled by majority rule, more complex logical environments necessitate increasingly restrictive—and eventually dictatorial—rules to maintain group rationality.
Summary updated