Open Problems in Differentiable Social Choice: Learning Mechanisms, Decisions, and Alignment

TL;DR

This paper reviews differentiable social choice, a paradigm where voting and decision mechanisms are learned models, highlighting its theoretical foundations, practical applications, and open research challenges at the intersection of machine learning and social choice.

Contribution

It synthesizes existing work on differentiable social choice, connecting classical axioms with modern machine learning approaches, and outlines 18 open problems for future research.

Findings

Differentiable social choice models incorporate classical axioms as objectives.

The paradigm enables data-driven learning of voting mechanisms.

Identifies key open problems in the field.

Abstract

Social choice has become a foundational component of modern machine learning systems. From auctions and resource allocation to the alignment of large generative models, machine learning pipelines increasingly aggregate heterogeneous preferences and incentives into collective decisions. In effect, many contemporary machine learning systems already implement social choice mechanisms, often implicitly and without explicit normative scrutiny. This Review surveys differentiable social choice: an emerging paradigm that formulates voting rules, mechanisms, and aggregation procedures as learnable, differentiable models optimized from data. We synthesize work across auctions, decision aggregation, and preference learning, showing how classical axioms and impossibility results reappear as objectives, constraints, and optimization trade-offs. We conclude by identifying 18 open problems defining a new research agenda at the intersection of machine learning and social choice theory.