Researchers' 'Maximally Random Sortition' algorithm boosts assembly diversity with maximum entropy

Researchers Gabriel de Azevedo and Paul Gölz have published a paper titled 'Maximally Random Sortition,' introducing a novel algorithmic approach for selecting members of citizens' assemblies. The core innovation is designing algorithms that sample from maximum-entropy distributions over possible panels, which mathematically maximizes randomness. This approach can incorporate constraints on individual selection probabilities while ensuring the final panel is as unpredictable as possible, a key theoretical improvement aimed at enhancing fairness and resistance to strategic manipulation.

The team rigorously tested their algorithms by benchmarking them against a large set of real-world assembly lotteries. They evaluated performance on two critical measures: intersectional diversity (ensuring representation across multiple demographic dimensions) and the probability of satisfying unseen representation constraints. The results were favorable, demonstrating that maximum-entropy sortition can outperform traditional methods. Furthermore, the researchers have deployed one of their algorithms on a practical website for citizens' assembly practitioners, moving the theoretical work into a tool for real-world democratic innovation.

The paper, submitted to arXiv, falls under Computer Science and Game Theory (cs.GT). It provides a formal investigation into the properties of these algorithms, including their transparency and robustness. By shifting the goal from simple random selection to maximizing entropy, the work offers a more sophisticated mathematical foundation for a process central to deliberative democracy, potentially leading to panels that are more representative and trusted by the public.