Artificial Intelligence Alignment for Inclusion (AIAI)

Using artificial intelligence (AI) to support citizen engagement in the co-creation of urban spaces

Montreal Belvedere at top of Mount Royal

Background

The field of urban design and landscape architecture can be a source of exclusion due to: i) the privileged background and individualistic culture that characterizes the industry; and, ii) the incentives that prioritize financial return rather than community benefit and support.

When citizen engagement in the urban design process does exist (because the budget is large enough, municipal buy-in is required, or there are political incentives to do so), a number of barriers prevent the process from being sufficiently engaging. These barriers are not only connected to challenges of accessibility in terms of workshop location and timing, but also to limitations of the tooling, in which citizens find it difficult to:

Convey the “feeling” of the spaces they would like to occupy;
Communicate their needs in physical space;
Iterate quickly amongst themselves and with organizers;
Have their ideas reported back to the architects in a way that authentically represents their expressed needs.

About the Project

AIAI is a machine learning research project designed to make urban architecture more inclusive, accessible and safe for marginalized groups.

The research involves building a dataset, in collaboration with vulnerable communities living in Montreal, of AI-generated images that have been labeled for their inclusiveness, accessibility and safety (among other metrics). Once the dataset is complete, it will be used to fine-tune a Stable Diffusion XL model to support the generation of more inclusive renderings of public spaces.

Should this research be successful, the fine-tuned, prompt-based AI image generator can be used by community members to help them visualize their needs and desires for public spaces in their neighborhoods. By providing this tool to third party facilitators during the community engagement process of an urban development project, it can help make the sessions more efficient and engaging, improving the likelihood that landscape architects incorporate community needs into their designs.

Learn more on how you can support this work

Objectives

The AIAI project is meant to be part of the solution. This research, which is being built for the Island of Montreal, will challenge the structures that prevent inclusion in the urban design process by helping participants to articulate how they would like to feel in their space, illustrating their needs and desires in physical space and allowing them to iterate quickly in conversation with others, summarizing this negotiation back to architects more accurately.

This project has been shaped by concepts like participatory planning, design justice, EDI, and intersectionality, and inter-sectoriality.

In the next six months, we aim to:

Create and publish an open source dataset. This dataset will consist of prompts and images containing notions of more inclusive, accessible, diverse, safe and comfortable public spaces.
Release benchmarks evaluating the model’s performance when fine tuned on our dataset.
Make available documentation detailing the process and best practices for community engagement.

Over the next year, we will determine whether this research can be made into a viable product. If so, we will identify funding opportunities, expertise and deployment partners to help bring this work to life.

Please get in touch if you’d like to support this work! You can contact the Senior Applied AI Projects Manager at allison.cohen@mila.quebec