Representative Social Choice: From Learning Theory to AI Alignment
Tianyi Qiu
TL;DR
This work introduces representative social choice, a framework that models democratic representation when both the number of issues and the population are large by sampling individual-issue pairs and restricting outputs to a finite candidate space. It leverages statistical learning theory to analyze generalization, deriving bounds for binary and non-binary settings, and extends classic impossibility results to the representative setting via Arrow-like theorems and novel combinatorial tools (privileged orderings and privilege graphs). The paper shows that simple mechanisms (e.g., majority voting) generalize well under low-complexity candidate spaces, while strong impossibility results reveal fundamental trade-offs between fairness and utility when issues interdepend. Implications for AI alignment and governance highlight how representative social choice can guide robust, scalable alignment and decision processes, while acknowledging limitations and directions for future research (incentive compatibility, computation, and handling noise).
Abstract
Social choice theory is the study of preference aggregation across a population, used both in mechanism design for human agents and in the democratic alignment of language models. In this study, we propose the representative social choice framework for the modeling of democratic representation in collective decisions, where the number of issues and individuals are too large for mechanisms to consider all preferences directly. These scenarios are widespread in real-world decision-making processes, such as jury trials, legislation, corporate governance, and, more recently, language model alignment. In representative social choice, the population is represented by a finite sample of individual-issue pairs based on which social choice decisions are made. We show that many of the deepest questions in representative social choice can be formulated as statistical learning problems, and prove the generalization properties of social choice mechanisms using the theory of machine learning. We further formulate axioms for representative social choice, and prove Arrow-like impossibility theorems with new combinatorial tools of analysis. Our framework introduces the representative approach to social choice, opening up research directions at the intersection of social choice, learning theory, and AI alignment.