The urban right to AI: Pluralistic co-design and governance of public space
This thesis develops a civic 'Right to AI' framework and two empirical tools tested in Montreal to govern public space algorithms. The Street Review tool combines participatory research with computer vision to map resident evaluations of streets using 45,000 street-view images, while the LIVS dataset compiles 37,710 pairwise comparisons across 13,462 images with 30 community organizations to fine-tune a Stable Diffusion XL model. The resulting framework translates these pluralistic evaluations into municipal governance practices, including procurement rules, oversight, and recourse mechanisms.
Why it matters — It establishes a method for municipal AI governance that treats public disagreement and value differences as valid, persistent evidence rather than as annotation noise or measurement error to be averaged out.
Caveat: The empirical applications and participatory datasets are grounded in the specific local context of Montreal, which may limit how directly the specific model alignments transfer to other cities.