Copyright notice

Produced on 25 October 2026.

Abstract

Sortition, the random selection of individuals to form deliberative panels, has seen a renaissance in recent years, as institutions realise the value of direct community and stakeholder input in deliberation and policymaking. These panels improve the quality of outcomes, increase public trust and engagement, and create more inclusive processes. Yet little attention has been paid to whether panels preserve a community’s intersectional structure – how demographic traits combine – rather than only its marginal proportions. This thesis uses a simulated sortition pipeline (population, survey sample, and panel selection via the LexiMin algorithm) to test whether quota-constrained selection preserves the joint two-variable association present in a survey sample. Across all simulated conditions, effectively every feasible panel showed a two-variable relationship significantly different from its sample, at a rate far exceeding simple random sampling. These findings motivated Intersection Inspector, a diagnostic Shiny app helping practitioners assess a panel’s representativeness before deliberation begins.

Keywords: sortition; citizens’ assemblies; representativeness; intersectionality; panel selection algorithms

Declaration

This thesis is an original work of my research and contains no material which has been accepted for the award of any other degree or diploma at any university or equivalent institution and that, to the best of my knowledge and belief, this thesis contains no material previously published or written by another person, except where due reference is made in the text of the thesis.

Student name: Daniel Leo Loza

Student signature: Daniel Leo Loza

Date: 25/10/2026

Reproducibility statement

This thesis is written using Quarto with renv (Ushey and Wickham 2025) to create a reproducible environment. All materials (including the data sets and source files) required to reproduce this document can be found at the Github repository github.com/Pian0Dan/Honours. Analysis and simulation were carried out in R (R Core Team 2026), making use of the ggplot2 (Wickham et al. 2026a), ggmosaic2 (Friendly et al. 2026), patchwork (Pedersen 2025), dplyr (Wickham et al. 2026b), tidyr (Wickham et al. 2025), tidyverse (Wickham 2023), knitr (Xie 2025), kableExtra (Zhu 2026), and shiny (Chang et al. 2026) packages.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

The Intersection Inspector app code (scripts/intersection_inspector/) is licensed separately under the GNU Affero General Public License v3.0, which requires anyone who modifies and runs the app as a network service to make their modified source available to its users.

Acknowledgements

I would like to thank my supervisors Dr Jessica Leung and Dr Dianne Cook for sticking by me throughout the development of this thesis. It was not an easy time, scopes changed greatly and often, and I was stubborn and slow at times. But I am proud of what we achieved, and grateful that they stood by me regardless.

I would also like to thank my parents, family and friends for making sure I remained alive between thinking, developing, investigating and writing this thesis, as well as through my general running around and complaining. Oh, and for being there while I ranted about ideas and repeatedly showed off bits and pieces I found, made, or thought were interesting.

I do not think this research would ever have happened, had I not been bitten by the sortition bug after local council engagements. So I would like to extend a thank you to Mel Jacobsen and Naomi Oosting from MosaicLab Co for showing me the power a voice can have in community. And another thank you to Mel for thinking of me and roping me in to work with MosaicLab Co myself. The experience was invaluable and formed much of my context and approach to the problems I wanted to demonstrate and work to solve.

I also believe thanks are in order for Bailey Flanigan and her team. Without your work, I would not have ended up on a community panel, getting the job I did, or where I am today as a whole, and would have had nothing to write about.

And finally, I would like to thank Muse for their song Unnatural Selection. It kept me grounded and sane within this project many a time, where I may otherwise have faltered.

AI Declaration:

I acknowledge the use of Posit Assistant (using an API form of Claude-sonnet) to assist me in the preparation of code, LaTex renders and initial drafting of section skeletons. The output from these was used to support my workflow in producing visualisations and clearly-written explanations.