Evaluating Fairness and Accuracy of Sortition Algorithms for Citizen Assemblies

• In Event: Digital Practices in Democratic Innovations

Abstract

Accurate representation of the population is a core promise of citizens’ assemblies. This also means giving an equal chance of participation to everyone in the population. Therefore, it is important to also examine the accuracy of that promise using a more sophisticated approach, i.e., one that considers to what degree a final panel matches the population, not just the absolute quotas.

We contribute on the topic of investigating the promise of minipublic’s accurate random selection.
Current best practices for random selection algorithms at the moment merely recommend matching the externally given quotas. If the quota is met, the algorithm claims accuracy – but that does not evaluate the actual composition of population and panel. We give guidance for practitioners on which stratification criteria to select for an assembly in the first place and which criteria to prioritize in the algorithm. We then draw on an improved sortition algorithm called simulated annealing algorithm in comparison with other algorithms to assess the consequences of using different sets of criteria plus algorithms for the selection of an assembly and leaving others out.
Our findings imply broader repercussions for one of the most disputed questions in the existing literature on sortition, i.e., whether the assembly needs to be a real mirror image of the population or whether the diversity rather than an exact representation of each stratification criterium suffices to have an effective citizens’ assembly.

Authors

• Alexander Matthias Geisler, University of Geneva

• Paulina Pospieszna, Adam Mickiewicz University of Poznan