🤖 AI Summary
Classical social choice theory assumes a finite set of discrete alternatives, rendering it inadequate for open-ended textual decision-making—such as collectively generating policy positions from unstructured input.
Method: We propose *generative social choice*, a novel paradigm that extends proportional representation theory to free-text generation by integrating the formal rigor of social choice with large language models’ (LLMs) capabilities in preference inference, text generation, and clustering. Our approach comprises oracle-guided preference modeling, extrapolation-based preference completion, LLM-assisted stance generation, and proportionality-constrained sentence selection.
Contribution/Results: Evaluated on empirical data from 100 U.S. residents regarding abortion policy, our method generated five stance statements rated “excellent” or “outstanding” by 84 participants as representative. This work delivers the first theoretically falsifiable framework for proportionally representative textual aggregation, establishing a verifiable foundation for AI-augmented open democratic deliberation.
📝 Abstract
The mathematical study of voting, social choice theory, has traditionally only been applicable to choices among a few predetermined alternatives, but not to open-ended decisions such as collectively selecting a textual statement. We introduce generative social choice, a design methodology for open-ended democratic processes that combines the rigor of social choice theory with the capability of large language models to generate text and extrapolate preferences. Our framework divides the design of AI-augmented democratic processes into two components: first, proving that the process satisfies representation guarantees when given access to oracle queries; second, empirically validating that these queries can be approximately implemented using a large language model. We apply this framework to the problem of summarizing free-form opinions into a proportionally representative slate of opinion statements; specifically, we develop a democratic process with representation guarantees and use this process to portray the opinions of participants in a survey about abortion policy. In a trial with 100 representative US residents, we find that 84 out of 100 participants feel"excellently"or"exceptionally"represented by the slate of five statements we extracted.
