Using AI to Design Inclusive Public Spaces

2024-09-11

Author: Rashid Mushkani, University of Montreal, Mila - Quebec AI Institute

Using AI to Design Inclusive Public Spaces Thumbnail

A conference talk about using AI not to flatten the city, but to reveal who it leaves out and how it might do better.

Lecture Overview

Presenter: Rashid Mushkani
Affiliation: Doctoral Candidate, University of Montreal
Conference: Innovate for Cities 2024

Watch the Lecture on YouTube

Why This Talk Mattered

Much of the public realm in contemporary cities was designed for a narrower public than the one that actually uses it today. Climate stress, migration, aging, disability, gendered experience, and cultural difference all shape how a space is read and whether it feels usable.

In this talk, I argued that AI can help surface those differences, but only if the technology is built around public judgment rather than imposed on it from above.

What I Showed

Predictive AI Model

I showed a predictive model designed to assess the qualities of public spaces. Trained on roughly 60 data points, it reached about 90% accuracy in predicting inclusivity from features such as sidewalk construction and surrounding buildings.

Generative AI Model

I also showed a generative model that produced conceptual redesigns for public spaces in a Montréal-specific context. The point was not to automate design taste, but to create visual scenarios that could support discussion and critique.

Community Workshops

The work was shaped through workshops with 20 participants from diverse backgrounds, including elderly, disabled, minority, women, and LGBTQ+ communities. Those sessions helped formulate prompts and ensured that the models reflected a wider set of perspectives than design software usually sees.

What It Suggested

The predictive model was able to distinguish between more and less inclusive spaces and turn those judgments into heat maps that highlighted where Montréal might need attention. The generative model, in turn, offered a way to visualize alternatives shaped by different public needs.

Together, the two models suggested a more useful role for AI in urban design: not to dictate decisions, but to help residents and planners see patterns, test alternatives, and make trade-offs more legible.

What Comes Next

I want to keep pushing this work toward a public-facing platform where residents can visualize, question, and contribute to the design of their neighborhoods. That means more data, better interfaces, and more deliberate collaboration across planning, AI, and civic institutions.