Sidewalk Toronto Goes Sideways: Five Lessons for Digital Governance
During his push to erect Google/Alphabet’s first smart city on Toronto’s eastern waterfront, Sidewalk Labs CEO Dan Doctoroff admitted, “I don’t think I fully appreciated what I was getting myself into.” He had not been prepared, it seems, for public resistance to the vision of a big-tech city “built from the internet up.” And perhaps the resistance would not have been so vigorous but for the history of tech companies exploiting personal data and crashing into guardrails meant to protect the public interest. It was in large part because of this resistance that, some three years into the project, on May 7, 2020, Sidewalk Labs folded what it had branded “Sidewalk Toronto,” although the company blamed the COVID-19 pandemic. This was a hugely ambitious effort to re-imagine housing, work space, transit, sanitation, sustainability, and the public realm, all aided by massive data collection and analytics. But the ambition was inextricably bound up in a big-tech appropriation of space and data that lacked public legitimacy.
As tech companies build cities elsewhere, as Toronto itself looks forward to new deals, and Sidewalk engages with new urban partners, it is worth considering what lessons about digital governance (my focus) can be drawn from the Sidewalk Toronto experience. Seldom are so many superb planning ideas assembled for one place — new methods of building, new modes of transit, new possibilities for organizing housing, recreation, sanitation, and energy use, with the goal of a more sustainable, inclusive, and beautiful city.
Sadly, these visions outpaced public governance — not just about data collection and use, but also more fundamentally about who has power in the city. The audacity of the Sidewalk Toronto project- with data running through the most mundane and necessary features of life, like trash disposal- kindled anxiety. Big tech companies have broken too many promises about digital stewardship to be trusted unsupervised, as I showed in an expert report, prepared in connection with a civil rights challenge to the project. Sidewalk Toronto required guarantees not only about data privacy, but that algorithmic governance would be subject to democratic process and that vendors would not usurp public power. There were no such guarantees. On the collapse of the project, Julie Di Lorenzo, a former member of Waterfront Toronto’s Board who had resigned earlier in protest, said that the project’s failure proved “there is a fundamentally important role for a democratically elected government on governance of issues of privacy, the environment and the person’s right to housing.” Jim Basille, former co-CEO of Research in Motion, characterized Alphabet’s retreat as a victory for Canadian civil and digital rights against surveillance capitalism.
The Toronto experience demonstrated that public resistance to a vendor-first smart city can work. I identify five points where the Sidewalk Toronto project failed in terms of digital governance — failures of public process and private strategy from which smart urbanism can learn.
1. Privatization. The project design privatized public functions from the start.
Waterfront Toronto, the public authority responsible for developing the city’s waterfront, invited a private vendor to create a master innovation plan for the new district. In a lawsuit against Waterfront Toronto, the Canadian Civil Liberties Association alleged that the scope of the partnership was a fundamental abrogation of public responsibility. “Powers were given away to a private enterprise that cannot legally be given away.” Specifically, the vendor was tasked with determining what services should be offered in the new Quayside district, from telecommunications to underground infrastructure to transit, curb use and zoning. It was left to the vendor to determine data policies, including how people living in or visiting Quayside would (or would not) be able to prevent data collection. It was even left to the vendor to collaborate on procurement policies that might benefit the vendor or its affiliates. Specifically, Sidewalk was put in a position to nix competitive bidding when a “purposeful solution” (perhaps even offered by an Alphabet company) was uniquely suited for a particular need — a need that Sidewalk itself had designed into the city. While this procurement conflict was later eliminated, the initial power grab sowed public distrust not easily dispelled.
The usurpation of the public was not usually explicit, but embedded in the structure of the technology. We can start with how rights to the streetside curb are allocated. Sidewalk proposed technology to gather data about transit, parking, deliveries, pedestrian and bike use, and all other demands on the curb. This data would run through a private interface and be made available to city government. It would likely turn to another affiliated company, Replica, to model urban movement. With the city turned into a client of a private black-box interface, these arrangements create the potential for dangerous lock-in, dependency, and privatization. In response to these threats, the city of Los Angeles built its own interface to take control of the curb, and then contributed it to the Open Mobility Foundation for use by cities to use in resistance to private interfaces. Saleta Reynolds, the LA transportation manager responsible for this project, said “This platform — built on open standards and software and designed by and for cities — helps us achieve important city goals like increasing safety, equity, and health outcomes, while lowering emissions, and reducing congestion. Existing tools or proprietary platforms would never allow cities to do that, so we’ve decided to roll out our own.” This kind of public or open interface was not even part of the Sidewalk Toronto discussion, and the same can be said for the other forms of control in the city — all mediated through data — such as sanitation, energy use, and building use.
2. Participation. Sidewalk Toronto was presented as a fait accompli, dazzling and secret.
David Skok, editor of The Logic, writes that Sidewalk Labs staff told him the project’s “original sin” was the photo op of Oct. 17, 2017: Prime Minister Justin Trudeau, then-Ontario premier Kathleen Wynne, and Toronto Mayor John Tory stood alongside then-Waterfront Toronto CEO Will Fleissig and Sidewalk Labs CEO Dan Doctoroff to announce a plan for Toronto’s Quayside neighborhood. “The visual of the five of them standing shoulder to shoulder sent a powerful message that the deal was done, when it really hadn’t even begun.” Indeed, Ontario’s Inspector General in 2018 found that the initial award of the planning contract to Sidewalk Labs was problematic — the company had the inside track to become the chosen vendor from the start and government officials were not sufficiently consulted.
Even after Waterfront Toronto and Sidewalk Labs had inked their initial agreement in 2017, details of the deal remained secret and hidden even from most of Toronto City Council. It was not clear, and indeed never really became clear, what was in it for Sidewalk. They had agreed to invest $50 million just to socialize the plan with the public, but the company’s role — real estate developer, scenario planner, technology company — was ambiguous. As I wrote in a review of the first phase of the project, “No one could figure out what Sidewalk’s business plan was, how it would make money, how it would pay for the ambitious innovations it proposed, and what long-term commitments it was prepared to make to the city.” One way Sidewalk and Waterfront Toronto avoided real scrutiny about the nature of their partnership was by having Sidewalk manage public consultation. There were a fair number of roundtables nominally co-hosted by Sidewalk and Waterfront Toronto, but really company-run with all the hallmarks of a tech expo. Sidewalk set the agenda and conducted what amounted to a simulacrum of a public process performed on a stage set, without the accountability of real public hearings. The CCLA called this process “governance by mercenary.”
3. Transparency. Sidewalk Labs was cagy about its plans.
It is inevitable in a project the size of Sidewalk Toronto that plans will evolve over time. No one can expect all the answers to be there from the start, especially when the endeavor centers on innovation and new ideas. But it is necessary to have some non-negotiables in place and for high level values to be agreed up front. Will children be surveilled in school? Will public services be privatized? That did not happen with Sidewalk Toronto. There was a huge gap between Waterfront Toronto’s initial project scope and what Sidewalk Labs ultimately presented to the public. It was so huge that much of what Sidewalk proposed in the massive 1500 page Master Innovation and Development Plan in June 2019 — including matters of geographic scope, permitting, and data — had to be walked back under pressure only a few months later. Such overreach and subsequent retreat may fly in a business negotiation, but when done as part of public administration, it resulted in a calamitous loss of credibility.
There was a slipperiness to the entire project from the RFP stage to the Master Plan negotiations. What was supposed to be the development of a 12 acre parcel metastasized into a proposal that spanned 190 acres. There were leaks that Sidewalk had in mind the development of something like a social credit score based on massive urban surveillance. It seemed also that Sidewalk was hoping to get an ongoing share of tax revenues in return for building mass transit. Had Waterfront Toronto or other public authorities been clearer about what kinds of projects or data collection were simply out of scope, these leaks might have been less damaging. But the public entities seemed outmatched with the company on the march.
4. Governance. The vendor took charge of digital governance.
It’s obvious that there should be rules and accountability around data capture, use, and retention before a city or other government embarks on surveillant activity. It is also clear that Canadian and most other data privacy laws organized around consent are not well-suited to smart city use cases, where sensor networks are ubiquitous and consent impractical. Sidewalk itself in its Digital Appendix to the Master Plan reviewed the growing body of work on bespoke digital governance in the smart city context, tacitly recognizing that there were significant governance gaps. Waterfront Toronto admitted the same, adding hopefully the issues could be worked out in “commercial negotiations.” Waterfront Toronto’s blue-ribbon Digital Strategic Advisory Panel, which had provided excellent critique of the project throughout, concluded that the amalgam of private commitments and possible future policies does not amount to “a fully realized digital governance framework.”
Tech companies frequently say “we need a law,” but meanwhile proceed quickly without. This is the “move fast and break things” approach. So here, the planning for Sidewalk Toronto advanced while Toronto was just getting started on its Digital Infrastructure Principles. Even though Sidewalk promised to adhere to later-adopted policies, law does not retrofit well to developments already on the ground. See Uber. While it’s also not always possible for slow-moving law to anticipate fast tech, the answer to this pacing problem can’t be to simply privatize governance and hope for the best.
Tasked with freelancing digital governance, Sidewalk Toronto came out early with a “responsible data use framework” making good general points about the need for transparency, security, access, privacy, and control. It proposed a data trust to park “urban data” collected in the district — an interesting concept, but half-baked in implementation, apparently contrary to Canadian law, and requiring strict anonymization protocols that Sidewalk was not offering. The company dropped the idea. Sidewalk relied heavily on data de-identification techniques. These have been effectively critiqued as insufficiently privacy protective and were apparently so lacking that prominent privacy expert Ann Cavoukian resigned as a project consultant. The Master Plan contained bromides about “digital restraint” and not selling data to third parties. But “restraint” is relative and data can be squeezed for harmful “data derivatives” without ever being sold.
Beyond its specific proposals, the way that Sidewalk Toronto framed digital governance was problematic. It advanced 160 technologies along with techniques to mitigate potentially harmful data flows. This approach put the technology “solution” first, and then dealt with possible harms rather than putting values first. As Bianca Wylie – one of the most influential tech critics of the project – put it: “[w]hen did we as a society say that however we move around in public space—that this is something we want to share and commodify?” Public — and public regarding — digital governance requires prior commitments. Where are the spaces that should be free from surveillance? Who should benefit from data and who should bear responsibility for misuse? What is the surveillance load a person should have to bear? If we are serious about notice and consent, are there realistic alternatives for people who opt-out?
5. Power. Digital governance is about power, not just privacy.
By now, the discourse around smart cities has elevated the privacy and security harms of sensor networks and algorithmic decision systems. This focus naturally leads to promised correctives around minimization, anonymization, encryption. What I hope emerges from the discussion above, especially about privatization and governance, is that the data flows implicate power as much as privacy. It’s not easy to get purchase on this question of power or even to describe it, but fears of powerlessness were palpable among engaged citizens who saw the future Sidewalk Toronto as beyond their control. It seemed their city was sleepwalking into tech captivity – a case of what Langdon Winner calls technological somnambulism. Two ways the power issues play out is in tech solutionism and the mirage of optionality.
The Toronto project proceeded from aspirations about resilience and inclusion to a set of “solutions.” Because questions about data governance were not addressed early, the data-related costs of these solutions were hidden, while the benefits leapt from screens of beautiful renderings. This is just what a sales pitch should accomplish. And in theory, the public bodies that ultimately issue the permits to break ground can assess costs later. In practice, however, the “solutions” have a first mover advantage that propels them into reality. The authorities competent to deal with digging and building are not data experts and there is a good chance that the digital costs will be tallied too late. An alternative to solutionism is a real public process that weighs the necessity and proportionality of the tech deployments, as compared to other approaches. An alternative would take seriously the pre-solution question of whether platform urbanism is in the public interest. Even if particular technologies handle data flows well, the cumulative effect of surveillance-based service provision may modify human behavior, labor, and the provision of goods and services in ways that are unwelcome and will be hard to change once actualized.
When the risks of “solutions” are framed principally as privacy risks, the mitigation strategies turn to notice and consent. Sidewalk launched an initiative to make data collection and use more transparency in a smart city using iconography.
While cities do need to communicate better about data flows, the implied promise of optionality — that residents and visitors have a choice about data collection – is a mirage without real alternatives. Sidewalk Toronto proposed to collect unit-level data about energy usage and disposal, to collect personal information in order to control package delivery, and to run parking and electric vehicle charging through a “dynamic curb” data interface. Because it was not Sidewalk’s business to offer alternatives, and because Sidewalk was in charge of planning, alternatives to this “voluntary” private platform were not included. Visitors and residents of Hudson Yards, a crude prototype for Sidewalk Toronto, have no data optionality. Questions of optionality are now central to how governments are implementing COVID-related surveillance. Ultimately, whether a technology is or in any sense could be voluntary, especially for vulnerable populations, is a question of power that should be surfaced upfront. Real distributed power will require civic innovation that won’t be “rolled out” by tech companies and indeed may be “meaningfully inefficient” in ways that Eric Gordon and Gabriel Mugar discuss in their book on civic design.
In the end, it may be that Sidewalk Labs, as an affiliate of Google, could not have succeeded in Toronto no matter how the process had been run. And maybe that is one of the lessons: concentrated power in the biggest tech is incompatible with democratic governance of cities. At a more granular level, we may be able to learn from the mistakes of both Waterfront Toronto and Sidewalk — arrogance, evasion, distraction, obfuscation, and a failure to lead with public digital governance.
Ellen P. Goodman is a Professor at Rutgers Law School. Follow her on Twitter @ellgood.