Project
EF6-2
Deliberation Processes in Citizens’ Assemblies: An Optimal Experimental Design Perspective
Project Heads
Sven Wang, Jobst Heitzig, Claudia Schillings
Project Members
Dennis Nieman
Project Duration
01.10.2024 – 31.12.2025
Located at
HU Berlin
Description
Citizens’ assemblies are increasingly relevant for involving the broader civic population in political issues. A key question for those assemblies is that of optimal design of the multiple-round discussion process, which recently has been aided by algorithms aimed at maximizing ‘engagement’.
Roughly speaking, the process of organising a citizens’ assembly has two phases. In the first phase, a pool of volunteers is recruited, and the assembly is randomly selected from this pool. The assembly is required to be representative of the population in terms of different features, e.g. gender, age, ethnicity and education. In recent years, an algorithmic framework for the fair selection of such panels has been developed.
This project will focus on the second phase of such processes, where the selected panel discusses the issue at hand (deliberation), informing their policy decisions. To facilitate the exchange of information and to guarantee ‘fair shares’ of speaking time, discussions are structured and moderated. Specifically, there are typically multiple rounds of deliberation—in each round, participants are assigned to a different ‘discussion table’ consisting of 5–8 people. [This problem is closely related to the well-known “Oberwolfach problem”.]
We will be concerned with the key question: How should one best design the deliberation process? Already now, Sortition Foundation, one of the most prominent NGOs for running citizens’ assemblies, has made accessible an open-source app called GroupSelect which generates table assignments based on systematically maximising encounters between participants. This algorithm takes a somewhat ad-hoc, but intuitive approach: it aims to simply maximise the number of pairs of participants who encounter each other at least once.
This project takes a new perspective on this design problem, borrowing tools from the theory of optimal experimental design. Based on this viewpoint, we develop and test a new class of algorithms for the optimal design of deliberation processes. The main novelty of our approach is that we take into account the individual characteristics of the participants.
Project Webpages
Selected Publications
- Manipulation-robust selection of citizens’ assemblies.
B. Flanigan, A. Procaccia and S. Wang
Proceedings of AAAI (2024, selected for oral presentation) - Distortion under Public-Spirited Voting
B. Flanigan, A. Procaccia and S. Wang
Economics and Computation (EC) (2023) - Quasi-Monte Carlo for Bayesian design of experiment problems governed by parametric PDEs
Vesa Kaarnioja, Claudia Schillings
arXiv preprint (2024)
Selected Pictures
