Late in the afternoon of Monday, October 15, 2018, Sidewalk Labs released a densely-packed slide-deck which outlined its new and emerging data governance plan for the Sidewalk Toronto smart city development. The plan was discussed by Waterfront Toronto’s Digital Strategy Advisory Panel at their meeting on Thursday, October 18. I am a member of that panel, and this post elaborates upon the comments I made at that meeting.
Sidewalk Labs’ new data governance proposal builds upon the Responsible Data Use Policy Framework (RDUPF) document which had been released by Sidewalk Labs in May 2018. It is, however, far more than an evolution of that document – it is a different approach reflecting a different smart city concept. It is so different that Ann Cavoukian, advisor to Sidewalk Labs on privacy issues, resigned on October 19. The RDUPF had made privacy by design its core focus and promised the anonymization of all sensor data. Cavoukian cited the fact that the new data governance framework contemplated that not all personal information would be deidentified as a reason for her resignation.
Neither privacy by design nor data anonymization are privacy panaceas, and the RDUPF document had a number of flaws. One of them was that by championing deidentification of personal information as the key to responsible data use, it very clearly only addressed privacy concerns relating to a subset of the data that would inevitably be collected in the proposed smart city. In addition, by focusing on privacy by design, it did little to address the many other data governance issues the project faced.
The new proposal embraces a broader concept of data governance. It is cognizant of privacy issues but also considers issues of data control, access, reuse, and localization. In approaching data governance, Sidewalk is also proposing using a ‘civic data trust’ as a governance model. Sidewalk has made it clear that this is a work in progress and that it is open to feedback and comment. It received some at the DSAP meeting on Thursday, and more is sure to come.
My comments at the DSAP focused on two broad issues. The first was data and the second was governance. I prefaced my discussion of these by warning that in my view it is a mistake to talk about data governance using either of the Sidewalk Labs documents as a departure point. This is because these documents embed assumptions that need to be examined rather than simply accepted. They propose a different starting point for the data governance conversation than I think is appropriate, and as a result they unduly shape and frame that discussion.
Data
Both the RDUPF and the current data governance proposal discuss how the data collected by the Sidewalk Toronto development will be governed. However, neither document actually presents a clear picture of what those data are. Instead, both documents discuss a subset of data. The RDUPF discussed only depersonalized data collected by sensors. The second discussed only what it defines as “urban data”:
Urban Data is data collected in a physical space in the city, which includes:
● Public spaces, such as streets, squares, plazas, parks, and open spaces
● Private spaces accessible to the public, such as building lobbies, courtyards, ground-floor markets, and retail stores
● Private spaces not controlled by those who occupy them (e.g. apartment tenants)
This is very clearly only a subset of smart cities data. (It is also a subset that raises a host of questions – but those will have to wait for another blog post.)
In my view, any discussion of data governance in the Sidewalk Toronto development should start with a mapping out of the different types of data that will be collected, by whom, for what purposes, and in what form. It is understood that this data landscape may change over time, but at least a mapping exercise may reveal the different categories of data, the issues they raise, and the different governance mechanisms that may be appropriate depending on the category. By focusing on deidentified sensor data, for example, the RDUPF did not address personal information collected in relation to the consumption of many services that will require identification – e.g., for billing or metering purposes. In the proposed development, what types of services will require individuals to identify themselves? Who will control such data? How will it be secured? What will policies be with respect to disclosure to law enforcement without a warrant? What transparency measures will be in place? Will service consumption data also be deidentified and made available for research? In what circumstances? I offer this as an example of a different category of data that still requires governance, and that still needs to be discussed in the context of a smart cities development. This type of data would also fall outside the category of “urban data” in the second governance plan, making that plan only a piece of the overall data governance required, as there are many other categories of data that are not captured by “urban data”. The first step in a data governance must be for all involved to understand what data is being collected, how, why, and by whom.
The importance of this is also made evident by the fact that between the RDUPF and the new governance plan, the very concept of the Sidewalk Toronto smart city seems to have changed. The RDUPF envisioned a city in which sensors were installed by Sidewalk and Sidewalk was committing to the anonymization of any collected personal information. In the new version, the model seems to be of the smart city as a technology platform on which any number of developers will be invited to build. As a result, the data governance model proposes an oversight body to provide approval for new data collection in public spaces, and to play some role in the sharing of the collected data if appropriate. This is partly behind the resignation of Ann Cavoukian. She objected to the fact that this model accepts that some new applications might require the collection of personal information and so deidentification could not be an upfront promise for all data collected.
The technology-platform model seems responsive to concerns that the smart city would effectively be subsumed by a single corporation. It allows other developers to build on the platform – and by extension to collect and process data. Yet from a governance perspective this is much messier. A single corporation can make bold commitments with respect to its own practices; it may be difficult or inappropriate to impose these on others. It also makes it much more difficult to predict what data will be collected and for what purposes. This does not mean that the data mapping exercise is not worthwhile – many kinds and categories of data are already foreseeable and mapping data can help to understand different governance needs. In fact, it is likely that a project this complex will require multiple data governance models.
Governance
The second point I tried to make in my 5 minutes at the Thursday meeting was about data governance. The new data governance plan raises more questions than it answers. One glaring issue seems to be the place for our already existing data governance frameworks. These include municipal and provincial Freedom of Information and Protection of Privacy Acts and PIPEDA. They may also include the City of Toronto’s open data policies and platforms. There are very real questions to be answered about which smart city data will be private sector data and which will be considered to be under the custody or control of a provincial or municipal government. Government has existing legal obligations about the management of data that are under its custody or control, and these obligations include the protection of privacy as well as transparency. A government that decides to implement a new data collection program (traffic cameras, GPS trackers on municipal vehicles, etc.) would be the custodian of this data, and it would be subject to relevant provincial laws. The role of Sidewalk Labs in this development challenges, at a very fundamental level, the understanding of who is ultimately responsible for the collection and governance of data about cities, their services and infrastructure. Open government data programs invite the private sector to innovate using public data. But what is being envisaged in this proposal seems to be a privatization of the collection of urban data – with some sort of ‘trust’ model put in place to soften the reality of that privatization.
The ‘civic data trust’ proposed by Sidewalk Labs is meant to be an innovation in data governance, and I am certainly not opposed to the development of innovative data governance solutions. However, the use of the word “trust” in this context feels wrong, since the model proposed is not a data trust in any real sense of the word. This view seems to be shared by civic data trust advocate Sean MacDonald in an article written in response to the proposal. It is also made clear in this post by the Open Data Institute which attempts to define the concept of a civic data trust. In fact, it is hard to imagine such an entity being created and structured without significant government involvement. This perhaps is at the core of the problem with the proposal – and at the root of some of the pushback the Sidewalk Toronto project has been experiencing. Sidewalk Labs is a corporation – an American one at that – and it is trying to develop a framework to govern vast amounts of data collected about every aspect of city life in a proposed development. But smart cities are still cities, and cities are public institutions created and structured by provincial legislation and with democratically elected councils. If data is to be collected about the city and its residents, it is important to ask why government is not, in fact, much more deeply implicated in any development of both the framework for deciding who gets to use city infrastructure and spaces for data collection, and what data governance model is appropriate for smart cities data.