On October 8, 2025, the School of Cities hosted Alexandra Flynn (Peter A. Allard School of Law, UBC), Jeff Allen (School of Cities), and Shion Guha (Faculty of Information and Department of Computer Science) in a multidisciplinary seminar on recent research in urban homelessness, with a focus on data and policy.
Flynn and Allen presented their project “Sheltering in parks: mapping the implications of local bylaws on unhoused populations”. To introduce the problem, Flynn presented an overview of how municipal bylaws in Canadian cities can restrict where people experiencing homelessness could shelter in parks. Such bylaws can be challenged by peoples’ constitutional rights to sleep in public spaces when they are not provided access to adequate shelter.
In 2019, the federal government introduced the National Housing Strategy Act that recognizes the International Covenant on Economic, Social and Cultural Rights. This Act obligates the federal government to recognize the progressive right to adequate housing as a fundamental human right. However, in this act and in international public law, and despite some municipalities having introduced housing charters, there are no mechanisms nor coordination at any level of government to enforce the right and address the need for adequate housing. Consequently, it falls to the courts to uphold charter rights for people who have been removed from sleeping in public spaces and to demonstrate that the bylaws preventing that use is unconstitutional.
As an example, here is an excerpt from Appendix A to the City of Hamilton Encampment protocol in 2023 of where encampments and temporary shelters are prohibited in Hamilton parks.
Flynn and Allen analyzed these bylaws in four cities; Vancouver, Hamilton, Prince George, and Kingston to answer questions of what proportion of parks can be used for encampments, tents, or temporary shelters? And where are these areas located relative to critical services? To do so, they collected, processed, and analyzed data on bylaw restrictions (e.g. 100m from a daycare) with geographic data to map bylaw restrictions for each of these four cities. Taken together, they were able to assess bylaw impacts and the development of legal arguments related to the constitutionality of the bylaw or conduct of officials.
Given the restrictions that the bylaws place on the usage of park space (e.g. prohibited areas for erecting tents or temporary shelters within a certain distance from playgrounds or schools), they found that the permissible amount of park space that is actually usable for individuals seeking a place to sleep is often limited.
For example, the map below shows areas where encampments and temporary shelters are allowed in the City of Hamilton after taking into account areas where encampments were prohibited based on an August 2023 protocol. The map shows that spaces where temporary shelters are allowed are limited, and much of it is a distance from where essential and support services are located. (Note that as of 2025, this protocol is no longer in place, and no encampments are allowed at all in Hamilton parks.)
Additionally, bylaws change over time, so mapping these bylaws over the years demonstrates how increasingly limited spaces where temporary shelters have become, particularly notable in Hamilton and Vancouver between 2020 and 2024.
Flynn and Allen’s work show that while spaces where camping and erecting temporary shelters become more restricted, people who have been forcibly moved from these spaces can challenge these bylaws based on charter and constitutional rights.
The ensuing discussion explored how case law has not been shown to favour municipalities’ bylaws on banning sleeping in park areas outright. Whether suitable shelter has been offered and provided depends largely on individual needs: the definition of “suitable” can be dynamic when applied to individual cases.
Guha presented his study led by PhD student Erina Moon, “Towards Designing Human-Centred Algorithms in Homelessness Services”, a research agenda to design human-centred practices for municipalities that
- Reframe outcomes to shift the focus of homelessness services from “who is risky?” to “who needs help, and what works?”, to optimize to serve people directly, rather than depending on a representation of a general client
- Audit in the wild to study how data and algorithms work in actual settings where frontline staff are working, to see what works, what doesn’t, and who is affected
- Create practice and policy to create toolkits, protocols, and programs for classrooms, public agenda, and policymakers
Guha noted that many of the systems for homelessness services that currently exist are designed and driven by people who often don’t engage with that system and embed their own values and design choices in these systems, leading to ineffective and incomplete service delivery. Systems that are focused on allocating resources efficiently build on data that are based on current client cases. As these clients are a diverse group who are discriminated against in many ways, these algorithms replicate those discriminations. Furthermore, algorithms are written to provide optimal solutions, but these solutions are not optimized for human client needs: “optimal solutions” tend to serve the systems of resource allocation, rather than solving for specific client needs.
The research for this project has found that algorithms focus on priorities via client profiling without fully considering what services clients need and are available in reality. Researchers can design systems using human-centred practices by conducting ethnographic studies, workshops, and interviews that centre the stakeholders most impacted by these systems. With this approach, scholars can begin to address the inequities that exist in the current service system.
Systems-builders can re-imagine algorithms for homelessness services by focusing on client needs, rather than client profiles. However, even when systems work well, capturing data accurately to better inform the system can be imprecise and messy. When clients navigating the system face constant and repetitious questioning as they move through different stages of service provision, and when frontline workers prioritize care for clients over accurate data collection, data quality suffers. Researchers can develop strategies to mitigate data quality issues, for example by collaborating with shelter service workers, so they can better understand how the data they collect can and will be used to support their clients. Frontline workers should also have access to robust feedback systems so that they can challenge data-driven decisions that negatively impact their clients.
The closing discussion focused on data safety and using AI conscientiously – to ease bottlenecks, for example, and use it in non-alienating ways. While cities have started to invest in AI systems in their processes, it’s also unclear how they define AI models. As such it is important to understand the risks of using AI, since the underlying architecture of the models are largely unknown and proprietary – and issues are often blamed on data rather than the underlying algorithm. Further, AI predicates on means-based statistics, but these social questions cannot be answered by the mean.
Ultimately, as Flynn notes, the objective for cities should not be to maintain temporary shelters; rather the goal should be safe and secure housing for everybody. Researchers should use data and AI systems to identify the challenges people experience with secure housing, rather than continue to use it to maintain the status quo of temporary, insecure shelter, and move away from maintaining a system that fundamentally doesn’t serve the people who are using it.
More information on Guha’s work on this topic can be found at https://tinyurl.com/soc-homelessness. Flynn will publish her report in Winter 2026.
