From smart city to digital urban commons: Institutional considerations for governing shared mobility data

Urban mobility is facing its first large-scale revolution since the automobile became ubiquitous. Electric batteries modernize propulsion systems and shared if not autonomous vehicles change demand and mobility patterns (Fulton et al 2017). At the same time, grid-locked cities, high morbidity and mortality from air pollution, and the call for climate change mitigation lead to increasing political pressure and concurrent proposals to act on transport systems (Colville-Andersen 2018, Creutzig et al 2012, Eliasson 2008, Gehl 2010). Political action will require traditional political instruments, such as pricing, land use policies, street space allocation and infrastructure provision. These are proven tools that will at least partially maintain their effectiveness also in the 21st century. However, it is unclear which governance schemes and institutions are best designed to manage the public good of the urban transport system vice a via the novel mobility technologies—electric, shared and autonomous—, and specifically how to govern the big data and artificial intelligence used to propel the new mobility schemes.

Shared mobility is defined as the shared use of a vehicle, motorcycle, scooter, bicycle, or other travel mode; it provides users with short-term access to a travel mode on an as-needed basis (Cohen and Shaheen 2016 and Shared and Digital Mobility Committee 2018). Shared mobility emerges at the nexus of the global megatrends digitalization, urbanization, and climate change. Apps, big data and artificial intelligence enable the provision of the new mobility services, and the more flexible use of existing public transit, that are enthusiastically embraced in cities worldwide, especially among young professionals, who appreciate the flexibility of multi-modal transport around the clock. Options include car sharing, e-scooters, bike sharing, ride pooling, and seamless boarding of public transit, integrated into multi-modal use in a smart mobility data platform that allows for frictionless use of a plethora of modal choices depending on the trip purpose, time and location. Shared mobility is self-accelerating: as applications require location data, services provide ever more of these data. But shared mobility is also constrained by existing infrastructure, path-dependent private vehicle use, and existing regulation.

The environmental impact of smart and shared mobility is ambiguous at best (Creutzig 2021 and Hu and Creutzig 2021). Ride-hailing services and transport network companies (TNC) such as Uber and Lyft are causing additional congestion (Erhardt et al 2019), mostly because more than 40% of all travel in TNC vehicles is made without passengers (deadheading) (Henao and Marshall 2019). In addition, about half of all TNC trips substitute for environmental modes, including walking, cycling and public transit, or would not have been made at all, thus adding traffic to the road (Erhardt et al 2019 and Rayle et al 2016). Similarly, motorbike ride hailing, after including all modal shift effects, at best is neutral with respect to changes in CO₂ emissions in urban transport, with positive and negative effects canceling out (Suatmadi et al 2019). More positively, carpooling services effectively lower CO₂ emission and congestion (Union of Concerned Scientists 2020), and have potential to systematically reduce GHG emissions in cities if substituting for private vehicles (International Transport Forum 2017). Carpooling reduces consequential marginal CO₂ emissions, because it increases occupancy levels, a key condition for low-carbon transport (Schäfer and Yeh 2020). Free floating car sharing schemes also draw mostly from public transit users, but nonetheless show a net beneficial effect on resulting CO₂ emissions as a relevant number of new car purchases are suppressed (Fromm et al 2019). Bike sharing services draw new users mostly from other sustainable modes of transport, and not the car (Fishman et al 2013), but nonetheless reduce car vehicle miles in most cities (Fishman et al 2014).

Given the overall environmentally ambiguous effects of shared mobility, the question is whether there is any need for these new modes of mobility at all. Possibly, established strategies of redesigning streets for people, as done, for example, in Paris, Amsterdam, and Copenhagen, work well and serve their purpose. Even more, the suspicion is that smart mobility is just another instance of surveillance capitalism privatizing and monetizing user data as a resource (Zuboff 2019). The main danger posed by current and future data collection is not the digital exposure of individuals but the proliferation of algorithmic procedures for population management (Mühlhoff 2020). Nonetheless, there are two arguments why the public and administrations should actively interact with the design and governance of shared mobility schemes. First, bike sharing and pooled car sharing offers advantages, and—if well designed—serves as substitute for private motorized mobility, and as complement to public transit. Second, smart mobility schemes are underpinned by huge investments and corporative takeover. A pure rejection of schemes imposes the danger of missing out on the opportunity to co-design institutions and governance of smart mobility when it is possible. Human agency and collective action will decide whether we can shape technology in our own interests or are subjected as data sources and objects to corporate algorithmic governance. Hence, it is important to consider and develop governance schemes consistent with the public good (Creutzig et al 2019 and Docherty et al 2018), and managing big data and digitalization for the public interest (Löfgren and Webster 2020).

Israel serves as shared mobility is also uncritically heralded as a new solution for congested urban spaces in Israel. A survey among Israel-based stakeholders reveals that smart mobility entrepreneurs are mostly concerned about commercial opportunities and lack a deeper understanding of what is required to transition to sustainable mobility (Noy and Givoni 2018). Noy and Givoni state that 'the belief among those entrepreneurs, it emerges, is that technological developments alone, specifically with respect to autonomous and connected vehicles, can lead to sustainable transport. This should be a real concern if those same actors are the ones who lead and pave the way forward for transport planning' (Noy and Givoni 2018). Active government involvement is thus required to steer smart mobility for the public benefit.

This paper asks how big data and artificial intelligence methods in smart mobility can become part of (urban) governance for not only smart but also sustainable mobility, here defined as mobility system that are low or zero carbon, low in congestion, and high in active and inclusive mobility. Given the hype about artificial intelligence and big data, and their potential to shape the technosphere and society, it is surprising how little attention is paid to the active role of governance. By directing technological innovation, vast potentials for improving the mobility of stressed urbanites and commuters can be leveraged and 'smart mobility' can possibly contribute to 'sustainable mobility' and meet the unfulfilled demand for better urban and sub-urban living.

In this contribution we hypothesize that an integrated data platform (IDP) for smart mobility is central to the governance architecture of smart mobility. This paper proceeds in three steps. First, it will delineate the potential and opportunity of an IDP. Second, it will discuss possible design options for the IDP. Third, it will suggest the next steps policy can take to advance the IDP. We will take the case of Israel for illustration.

The literature on digital platforms, digital urban governance, and surveillance capitalism is rapidly developing and relying on a plethora of overlapping but distinct concepts. A useful starting point is the concept of smart cities (Lazaroiu and Roscia 2012). Technically, it is based on several basic capabilities, including sensing, processing and decision making, control, and communicating (Akhras 2000), requiring sensors, command and control units, and actuators. Applied to urban transport systems, a benchmarking of cities measuring their degree of 'smartness' has been developed (Debnath et al 2014). The underlying parameters (e.g. automatic detection of parking facilities) make clear that smart cities contribute to more efficient service provision, even though their aggregate contribution to sustainability goals remains unknown.

Smart cities are a relatively new concept but must be understood in the history of the political economy of cities (Heaphy and Pétercsák 2018). The concept builds on similar but older framings, such as 'wired cities', 'digital cities', and 'cyber cities' and others (Kitchin 2014). But in contrast to older concepts, 'smart cities' gained traction inter alia because of the support of a well-organized epistemic community and of an advocacy coalition that realized an opportunity to advance their interest (Kitchin et al 2017). The notion of the smart city is polarized across two communities (Kitchin et al 2019). Computer scientists, industry, and parts of government aim to implement smart city technologies often within a relatively narrow technological focus reproducing and amplifying pre-existing political economy, arguing that they produce what consumers and the market request. Social scientists, in contrast, critique the notion of a smart city based on its surveillance or exploitative features. They argue that the smart city concept produces instrumental, technocratic, top-down forms of governance and government (Vanolo 2014), solutionist approaches uninterested in the structural causes of urban challenges, such as exclusion, pollution, and climate change, and surveillance technologies, such as ubiquitous face recognition that changes the meaning and phenomenology of the face, leading to a new step change in biopolitics (Smith 2020). Another critique maintains that artificial intelligence concepts, some of related to the smart city, promise solutions to issues like climate change, but instead are contributing, e.g., with the massive energy and resource demand of storage centers (Brevini 2020) (but see (Masanet et al 2020) for a recalibrated perspective). However, beyond critique only little pragmatic and constructive feedback emerged from the critical social science literature. As a result there are increasing calls to move beyond 'hope and fear' analytics (Leszczynski 2020).

An important development is captured by the notion of the smart city 2.0. In contrast to the vendor-based smart city 1.0, trying to selling smart sensors and application without regard of wider social issues and encompassing governance, the smart city 2.0 draws on urban platform governance via dashboards and shared public sector information (PSI; big data sets related to public urban assets) that encourage citizens, software developers, or academics to co-design government digital services, in turn instigating wider digital innovations that urban administration could never develop or even imagine themselves, but that are nonetheless extremely useful (Barns 2018). An example is FixMyBerlin, a platform that facilitates online feedback by citizens on bicycle infrastructure, information transport planners on desirable locations for interventions (Friedrichshain-Kreuzberg 2019). However, acknowledging this developments, it remains important to scrutinize the content behind the label: many 'citizen-centric' smart city initiatives remain disinterested in civil, social and political rights and the common good, and instead further civic paternalism and market-led solutions singularly focused on consumption choices and individual autonomy (Cardullo and Kitchin 2019).

Enter IDPs. Platforms are the essence of 21st century platform capitalism constituting digital infrastructures that mediate between different groups, such as users, advertisers, and content providers (Srnicek 2017). In essence, labels such as the sharing economy, uberization, and the internet of things all can be subsumed as instances of platform capitalism, drawing their success from organizing the intermediate structure required to accumulate data, and thus capital, as quasi-monopolies, including for example Google for search, Uber for taxis, and Facebook for social networks (Langley and Leyshon 2017). Platforms, however, have gained an additional role supporting, at least in principle, the public good, in particular as urban data-based governance, as dashboards, data market places, and as compliance instruments, elements of the alluded smart city 2.0 (Barns 2018), or as showcase and performance of data-driven city services (Currie 2020).

Here, we suggest to further advance urban governance and data platform solutions to make better use for the public good. While acknowledging that many opportunities for low-carbon and inclusive mobility exist independent of data platform governance, there are nonetheless two reasons to advance this agenda. The first is that platform solutions are promoted by vendors and monopolists in any case, and it is hence better to channel the existing ideas and energy into formats consistent with governance that protects privacy and prevents monopoly rents, but instead instigates public sector innovation. Data platforms may also be a the core of scaling smart transport solutions across cities (for scalability of smart transport solutions, compare with (Mingrone et al 2015)). The second is that shared mobility offers the opportunity to mitigate mobility injustices realized by a century of private automobility and foster low-carbon transport systems. We will discuss the case for such an IDP, and explore ownership and design options, which make a substantial difference, either resulting in a smart city 3.0 model or in an urban digital commons, achieved by recursive engagement with users (Teli et al 2015).

Multiple stakeholders shape smart mobility. Governments may want to improve the public good, at least in principle. Businesses seek new opportunities for innovation and profits. And citizens would like to improve their well-being and be active participants in decision-making and safeguarding their immediate environments. The relationship between these spheres is sketched out in figure 1. Each perspective is discussed in turn.

Zoom In Zoom Out Reset image size

Governments are interested in providing an IDP for three reasons. First, it facilitates the effective provision of public goods and services, such as uncongested roads, urban planning for good quality of life, and improved air quality (specified in box 1 for the Israeli case). Smart transport solutions have been suggested to contribute in improving services and reducing environmental burden (Mingrone et al 2015). Second, governments should also be interested in governing data to restrict data rentiership (Birch 2020) and surveillance capitalism (Zuboff 2019). Rentiership here refers to the appropriation of value through ownership and intellectual property, monopoly conditions, and regulatory or market devices and practices (Birch 2020), and has specific ramification in data monopolies that exert value extraction from personal data, a condition that mixes capitalist dominance with surveillance capacities (Zuboff 2019). Third, economically liberal governments may also want to invest in an open-access IDP to facilitate new business opportunities. In this function, an IDP would in itself constitute a public good by fulfilling two characteristics: every interested party can get access to the data; and every interested party can make use of data without compromising the ability of others to also make use of the data.

The key question is how an IDP can support the transformation of the transport system in order to reduce GHG emissions and to improve the daily life of commuters and residents in a notable way. A short look at urban areas that boast a high quality of life, as evidenced by multiple rankings, such as Vienna, Zurich or Copenhagen, reveals that an IDP is not necessary to bring about the envisaged benefits. These cities are built on a dense and high-capacity public transit system, urban planning as transit-oriented development, and bicycle highways, with restricted parking in inner cities. The Tel-Aviv lightrail will improve the situation, but it will not suffice to have notable impact in a growing economy. The real question is thus whether an IDP can support the reduction of congestion, air pollution, and accidents when high investments in public transit are politically not feasible, or when suburbanization makes efficient public transit difficult to realize. Here the argument is that an IDP achieves this in two steps: (1) an IDP supports a modal shift to shared mobility and low-carbon modes; and (2) this shift enables notable reduction in congestion, GHG emissions, and other environmental burdens.

An IDP facilitates shared mobility in the following ways. First, integrated data enable optimal provision of bike sharing schemes, considering both demand and inclusiveness. Considering that cities like Tel Aviv are hosting several bike sharing and e-scooter providers, coordinated and integrated solutions to locationing two-wheeled vehicles will outperform the competitive solution where all providers can draw only on their own data. In addition, integrated data management facilitates the entry of new mobility providers. Second, improved data on demand for and supply of shared pooled mobility can capture new market segments outside the inner cities, and thus more effectively replace private motorized traffic with low occupancy. Third, and most importantly, an IDP enables municipal and national governments to accurately regulate the mobility markets such as to incentivize shared mobility regimes with maximal benefit (low emissions, reduced congestion).

An important example, going beyond shared mobility, is the 'alternative' (formely 'Going Green') program of the Israeli government, providing financial incentives for car users to avoid congestion. If users agree, an synthetic version of the data could become part of the IDP, providing valuable information also for new shared mobility providers on geographically specific market potentials.

Resulting shared mobility schemes, if properly regulated, can effective provide public goods. Arguably, on of the keys to envisaging a congestion-free, environmentally friendly urban transit system with limited public transit capacity (although some capacity must be there) is shared mobility. Radical shared mobility scenarios demonstrate that congestion-free travel is possible even with cars. The key to make shared mobility also sustainable (reduced congestion and low carbon) is to increase occupancy levels (Schäfer and Yeh 2020), i.e. in this case to rely on shared pooled mobility instead of ride hailing and uberization. The International Transport Forum conducted two key studies that modeled shared mobility scenarios for Lisbon and Helsinki (International Transport Forum 2017). The detailed models show that replacing private car traffic with new shared mobility services in urban areas dramatically reduces the number of cars needed, cuts CO₂ emissions, and frees large swathes of public land for uses other than parking—without making it more difficult for users to get from door to door. With these shared services, all of today's car journeys in Helsinki's Metropolitan area could be provided by just 4% of the current number of private vehicles. The best results in terms of reducing emissions and congestion are achieved when all private car trips are replaced by shared rides (International Transport Forum 2017):

• CO₂ emissions from cars would fall by 34%;
• Congestion would be reduced by 37%;
• A lot of public parking space could be used for other purposes.

Shared mobility also means fewer transfers, less waiting, and shorter travel times compared to traditional public transport. This could attract car users that do not currently use public transport and encourage a shift away from individual car travel.

Bike sharing is already a proven model of shared mobility in reducing CO₂ emissions. In New York, Citi Bike sharing saved 0.7 MtCO₂ from avoided gasoline combustion in 2015 (Sobolevsky et al 2018). Similarly, on an annual basis, Mobike bike sharing reduced 2.3 MtCO₂ in Chinaw.

A central requirement for achieving frictionless radical sharing scenarios is the seamless integration of different transport modes with an IDP. One advantage of this scenario is that it does not require massive infrastructure investments and may be faster to implement and generate tangible results compared to expanding the railway infrastructure. It may also be preferable because of the higher system speed (travel time from door to door) it facilitates. It will, however, still depend on mass rapid transit along key arteries and corridors that would otherwise still be congested if used only by shared mobility. Finally, such a smart shared mobility scenario may be particularly fitting for Israel as a tech-affine nation.

The second key argument to be made for fostering an IDP in Israel: access to integrated mobility data and the provision of a vast opportunity space for shared mobility can benefit the Israel start-up economy. Values generated in mobility are rapidly shifting from hardware (cars) to software (digitized mobility services). Business and start-ups have significant interests in integrated data. Machine learning algorithms require considerable data of high quality to be trained and used for generalized learning and new applications. Companies that leverage customer behavioral insights outperform their peers by 25% in gross margin (McKinsey 2017). Israel has already numerous companies in this field, and an IDP could serve as the infrastructure that allow start-ups and business to thrive. Mobility data is required to train and implement any sort of algorithmic mobility service, and to provide ample innovation space to try out new solutions. For example, e-scooter providers can optimize the location where devices are deployed. And urban planners can optimally plan high-quality bike and two-wheeler networks.

Start-ups may have the most direct interest in an IDP. A plethora of imagined and unforeseen business models and innovations could be established with the help of big data and machine learning tools. The most obvious case is the provision of multi-modal integrated trips that are flexibly designed according to time, space, and needs. With app-integrated routing and ticketing, the transaction costs of travel can be drastically reduced. Access to IDPs can serve as an innovation booster and generate profits, possibly distributed across many parties.

Some established data rentiers, such as Uber, Tesla, and in particular Google, may have few interests to share their data. As data monopolies these companies may consider that proprietary control of quasi-financial resources is optimal. However, this is not-only incompatible with a competitive economy, stifling newcomers, but also inappropriate for an entrepreneurial state who understands innovation as a public good (Mazzucato 2011). We will discuss solutions to this conundrum below.

Within the automotive industry in Israel, new start-ups in smart mobility spring up faster (13% annually) than those in autonomous mobility (10%) or electric mobility (7%) (Bernhart and Ernst 2018). However, the smart mobility start-ups are more likely to rely on seed funding and have so far been less likely to make it to the revenue growth stage (Bernhart and Ernst 2018). Access to an IDP may offer the potential to scale existing business ideas. Smart mobility start-ups would profit from an IDP in the conventional sense of being able to make profits, and in the wider sense of making use of a powerful resource to explore new solutions, thus enabling innovation. The widest sharing of benefits, for both customers and companies, is achieved when returns are obtained as profits, not as data rents.

Individuals have a two-fold interest in an IDP for smart mobility. First, they would stand to benefit from most of the mobility services that make use of the IDP. Seamless integrated transport facilitates the daily commute and, perhaps more importantly, trip chaining, possibly also addressing gender discrimination in today's transport system. Arguably, current design of smart mobility realizes many of these consumer benefits, more than benefits via the public good provision route. Second, citizens may themselves make use of the IDP, for example by computing relevant metrics at neighborhood scale and subsequently visualizing this information to lobby for modifications in street design or transport infrastructures. Well-designed access to high-quality data can (re-)empower individuals, turning them from pure consumers into active societal agents—citizens.

This positive view must be contrasted with insights from mobility justice, and particular the argument that sustainable mobility and mobility justice need to go together (Sheller 2011). A system of automobility excludes those without cars, and automobile infrastructure often disadvantage poorer neighborhoods, and hinder mobility of those most vulnerable, including children, pregnant women, seniors, and those without cars (Sheller 2020). A shared mobility system and associated ITP must hence take justice concerns into account. For example, e-scooters and other shared mobility offers are often restricted to well-off neighborhoods with residents able to pay high marginal charges. In Tel Aviv, poorer South Tel Aviv neighborhoods were excluded from e-scooter provision until the municipality mandated that micro-vehicles must also be placed in this disadvantaged communities to equally share mobility opportunity. This example demonstrates hence that novel IDP based governance requires reflection of mobility justice at street and neighborhood level, and explicit political regulation. A shared mobility system that replaces private vehicles, and thus also improves accessibility and mobility by low-cost environmentally friendly modes of transport, such as cycling and public transit, has however considerable potential to not only meet sustainability but also justice concerns.

Mobility patterns concern everyday life, and active engagement by citizens with data will help governments to find fitting solutions. With the appropriate specifications, an IDP can serve to establish community data ownership, thus bringing economic rights to primary data providers and citizen groups, which in turn can support democratic functioning (Singh and Vipra 2019). The key principle is that individuals should own economic rights over their own data, and collective data, e.g. those associated with communal mobility patterns, should be owned by the community. Similarly to the Nagoya protocol, which requires benefits of genetic resources to be fairly and equitable shared with communities in question (Protocol 2011). Below, we will discuss IDP models compatible with such requirements.

Figure 2 summarizes the benefits of an IDP for different stakeholders, and their interaction, also emphasizing the need to complement an IDP with traditional public pull and push policies, such as expanding bicycle infrastructures and public transit, and restricting harmful private car use.

Zoom In Zoom Out Reset image size

The IDP will be defined within two main design categories: (1) data quality and quantity; and (2) ownership and access rights; and by one key outcome category: (3) success indicators. Let us discuss each in turn, starting with the measurement of success, since the identification of a desirable outcome will also help to define the architecture of the IDP itself.

Outcomes and value generated will depend on the use value for various constituencies and purposes, as outlined above for business, consumers and citizens, and governments and administration. Specific metrics are summarized in table 2. Specific ex-post evaluation metrics could consider the additional value generated by applications building on the IDP data, and possibly even the number of start-ups emerging as a result of having access to an IDP. Second-order success metrics could register the export of products, since there is considerable demand for shared mobility applications worldwide.

Evaluation metrics for governments would center on macro-economic metrics, such as reduced defensive health expenditure due to improved mobility and better air quality, as well as softer metrics, such as satisfaction with administrations and restored reputation and trust in public institutions.

Evaluation metrics for customers could take account of effective time saved in traffic and subjective improvements in happiness. Examples of successful participation through utilizing an IDP could be evaluated by monitoring citizen empowerment and trust in the government.

Risks include the misuse of data, such as travel-route profiling of individuals and the commodification of data without consent; the development of an undesirable data monopoly and resulting antitrust concerns; and the abuse of integrated data for political purposes, especially, but not only, by authoritarian regimes. For example, if people feel and are constantly tracked in their movements, governments and business can easily socially control entire populations (Creemers 2018). The metric to maintain here is data privacy, i.e. the fundamental right of personal protection as related to data use. This is, however, not only about individual rights, but also about the potential for mass manipulation by algorithmic structures. Hence, the protection of autonomy may be the more encompassing notion to be aspired to.

With the advancement in the location tracking capabilities of mobile devices, data privacy and autonomy concerns are becoming more pressing. Location data is some of the most sensitive data being collected. A list of potentially sensitive professional and personal information includes the identity of the user, their home address, individual interests as well as significant events, such as participation in demonstrations or a visit to an abortion clinic or church. In fact, just four spatio-temporal points, approximate places and times, are enough to uniquely identify 95% of 1.5 million people in a mobility database—even when the resolution of the dataset is low (de Montjoye et al 2013). Therefore, even coarse or blurred datasets provide little anonymity. This emphasizes the fundamental risk that any management in control of the IDP turns into a new monopolistic data rentier. Specific ownership considerations are hence important (see section on → ownership and access rights).

The above analysis of costs and benefits points in two different directions, highlighting on the one hand the need (a) to provide high-quality data in large quantities that can be used by everyone; and on the other (b) to restrict and prohibit the use of personal data. The risks can be turned into opportunities if data management translates into ownership consistent with broad participation and democratic engagement (Singh and Vipra 2019), greater trust, and transparent and smooth access to relevant data, where agreed upon, for government agencies and business. We will see below how this apparently hard trade-off may be solved, at least partially, by design principles and technological solutions.

In each case, proper access to large quantities of high-quality data is desirable. This is true especially for innovation, where possible from various domains to open up a space for innovations whose potential may not yet have been identified. For example, a detailed online monitoring of general travel data, together with e-scooter-specific data, may enable e-scooter and other shared mobility data to optimally relocate vehicles, while also taking transaction costs into account.

Data for shared mobility include on the one hand data on infrastructures and transport networks, such as street maps, maps of street allocation for different modes, public transit stations and networks, settlement structure and population density, street connectivity, and the socio-economic characteristics of neighborhoods. On the other hand, they also include individual travel data, such as origin-destination matrices, individualized trip-chaining data, travel diaries, and use behavior of transport modes. They may even code implicitly or explicitly for purpose of travel. This is important for understanding the service quality underlying mobility and for developing new solutions. Table 3 provides an overview on data qualities and potential source, as specified for Israel.

The first kind of data—infrastructures—is mostly risk-free, insofar as personalized data are not involved. Exceptions are security considerations and the identification of vulnerable infrastructures. The second kind of data—personalized data—is subject to data protection laws and entails substantial risk of abuse, as well as the potential of self-enslavement to data technology. It is therefore crucial to maintain data privacy and autonomy.

The apparent trade-off between the use value of data and the maintenance of data autonomy is shown in figure 3. The figure also indicates how the trade-off can be dealt with effectively. First, infrastructure data usually maintain data autonomy and can already generate some use value. Second, not all personalized data has the same value. A focus on the most valuable data only may realize most of the use, while minimizing the loss of data privacy and data autonomy. Given that less may be more, it would make sense to collect only the most valuable personalized data in the first place. Additional data could be designated as 'dumb data' and ignored (figure 3). This has the added benefit of reserving valuable server space and man-hours spent on data cleaning for the most valuable data, thus saving money and personnel resources.

Zoom In Zoom Out Reset image size

Data controllers can also draw on technological options to minimize the trade-off. First, data can be de-identified, for example by separating location data from a person's name and other unique person data. This could be taken as a first step, similar to the procedures adopted by TIMNA, the data science office of the Israeli ministry of health, to protect highly sensitive integrated health data. This is only a soft safeguard, however, as sophisticated programmers can often re-identify people.

Second, data statistics can be randomized and synthesized into new datasets with identical statistical information but synthetic identities. Such services are already delivered by a private company, MDClone, which offers a new healthcare data paradigm, enabling fast and direct access to healthcare data while fully protecting patients' privacy. Using original datasets, MDClone's Synthetic Data Engine creates anonymous data statistically identical to the original but with no actual patient information.

A third security mechanism is access via remote lock-in, not including data transfer. In this case, hybridizing with other data sources, which could help to re-identify personae, is blocked or requires an additional permission process. Additional options, such as multi-party computing, relate to the overall design of the data platform and are discussed along with particular ITP designs below.

All of these technological options expand the boundaries of the trade-off towards the most desirable outcomes of high use value and minimal data privacy inference (figure 3).

An IDP also needs to be defined in terms of the services it can or should provide, for example with regard to:

• The level of access granted at different levels;
• Whether it stores data or serves as intermediator between data controller and beneficiary;
• Any role in acquiring data;
• The amount of data preparation and cleaning;
• The provision of ancillary services, such as standard AI tools for interested parties without own capacity.

On a technical level, an IDP should conform to a number of desirable requirements, such as adequately coded metadata proving provenance (O'hara 2019). Independent of overall IDP governance and ownership, meeting these requirements is crucial to guarantee technical success of the IDP. A full list of requirements is provided in appendix A.

Ownership and access questions constitute the fundamental governance domain of an IDP. Here it is important to recognize that ownership is a priori an indeterminate term that does not specify the options that an owner has with respect to land/property. Different specifications are possible. Honoré identifies eleven possible characteristics of ownership, including (i) the right to possess (exclusive physical control), (ii) the right to use and access; (iii) the right to manage; (iv) the right to the income and others (Honoré 1961). Key design questions then revolve around who owns the data, who regulates it, and who has access and usage rights and to what degree. Technologies, such as a data renting and blockchain, make it possible to expand the scope of different options.

This insight opens up a space for highly differentiated considerations of ownership. For example, one party—the IDP—may manage the data, while a second party—e.g. users of the transport system sharing some data qualities—may hold the right to the income, while third parties—such as urban planning agencies or shared mobility providers seeking to improve their services—have limited or unlimited access rights, conditional on a fee. If the first party is also obliged to prevent harm but is excluded from the right to the income, an effective separation of regulatory power and commercial interests is ensured.

A more radical version would keep all personal data, and in the case of mobility in particular geolocation data, as personal property of individuals. Communal ownership could be established for the aggregate statistics of mobility patterns, revealing generic demand for travel across locations and time profiles at diminished resolution, sufficient to establish mobility services, but insufficient to reveal information about individuals. Interestingly enough, it is technologically feasible to use personal data for computation of aggregate statistics without having access to the raw data (Zyskind et al 2015a). This mechanism is based on a highly optimized version of secure multi-party computation, guaranteed by a verifiable secret-sharing scheme, where an external blockchain controls the network (Zyskind et al 2015b). Such a scheme enables the community of users to hold communal ownership rights on aggregated data. For example, in a block-chain based Uber alternative, Uber drivers themselves could own the data and capitalize on the services they offer (Tapscott and Tapscott 2016). Alternatively, mobility users of a city, providers of mobility data, could obtain the data-based surplus of more efficient mobility services.

Another question touches on the sourcing of data. Who brings the data in? Public institutions? Or private institutions as well? And what about data donations by interested citizens? And why should third parties share their data with an IDP in the first place?

Essentially, there are two data-sourcing paradigms. The first could be labeled: barter and negotiate. In this paradigm, the IDP negotiates with third parties on conditions and possibilities of data exchange and acquisition. The second could be labeled: regulate and govern. In this paradigm, an IDP is simultaneously an effective regulator and provides, for example, licenses for shared mobility conditional on data sharing and, possibly, sharing services that improve the larger public good in some to-be-defined sense.

Let us now break these considerations down into specific governance options and design solutions.

Governance options include market-based solutions (laissez-faire; monopoly), private-public interface (regulated market), an aggregating governance agency, and solutions in a third domain (data trust; decentralized data) (table 4). All of them could be realized but they differ in key ways.

Laissez-faire and monopoly are two private sector-oriented data governance options. These are very important benchmark options as one of them will arise by default if there is no dedicated initiative to bring data governance into the public space or organize it as a commons. There are clear advantages to these options. They bring the private-sector expertise in delivering solutions for consumers to the forefront. It is not unlikely that laissez-faire will transform into the monopoly or an oligopoly option, as working with hybrid big data and machine learning requires substantial capital investment beyond the capacity of smaller companies, at least when delivered as a whole portfolio of solutions. The laissez-faire option has the added advantage of enabling further experimentation by start-ups and bigger companies alike. Both options are, however, deeply problematic. Private companies have few incentives to deliver solutions that benefit the public good, even when they claim to have green, smart, or sustainable credentials. Equally problematic is the fact that both options will keep data in a proprietary format, stymying the generation of new business opportunities and sustainability solutions by new market entrants or interested parties. In this case, companies will aggregate data, sell some of the data at their discretion, and gather high data rents, which in turn stifles innovation by other actors. In some cases, data will be withheld, and even government agencies will not be able to buy data that would be useful for designing public policies (for example, Google withholds data from the Ministry of Transport that would be useful for reducing congestion). Proprietary data will serve their holders, not the public.

A second array of options entails a higher degree of involvement by the government. This can happen via a regulated market or a central government agency that aggregates the data. In the 'regulation' option, the government could mandate data-sharing arrangements and also prohibit the formation of a data rent-seeking monopoly. A disadvantage is that the regulator is often years behind in terms of technological development and may be challenged to dynamically design the best regulation for rapidly changing contexts.

A central government agency can aggregate all relevant data and utilize it, in principle, for the public good. As part of a large bureaucracy it may, however, be less well positioned to develop solutions that serve the users of smart mobility. Depending on the specific institutional setting, it is also not clear whether the organizational mindset is appropriate to delivering solutions such as easy access to the IDP for start-ups or measures to reduce congestion and improve mobility. Another key concern is data security and the potential for abuse by an authoritarian surveillance apparatus.

A data trust or foundation could help to ensure that an IDP serves the wider public good and is used appropriately. Data trusts are 'proven and trusted frameworks and agreements' that will 'ensure exchanges [of data] are secure and mutually beneficial' by promoting trust in the use of data (Hall and Pesenti 2017). They are supposed to 'provide ethical, architectural and governance support for trustworthy data processing' (O'hara 2019). Data trusts are motivated by the understanding that data protection laws are formalistically narrow and insufficient to build trust between agents and institutions working with or being the subject of data. Data trusts are frameworks that are intended to overcome the lack of trust, but do not require any new legal action. So far, a data trust has not been considered for the case of smart mobility, even though it could deliver multiple benefits.

As its central feature, the data trust would operate as an independent not-for-profit entity—neither public nor private—with a clear mission. The mission would be decided by a governing board of relevant stakeholders and could include serving the specific goals of combatting congestion, planning for low-carbon and highly accessible urban infrastructures, and providing data access for start-ups. It could also provide technical (machine learning) expertise for public entities, such as municipal governments, that seek to make use of the data to enhance the provision of public goods. The data trust's mission could also be to specify limits to data aggregation and a stringent securitization of the data, e.g. by de-identifying data or resampling data to synthetic formats, and by data-renting formats. Ideally, it would be run like a private business but with the goals of a public agency. A data trust could also contribute to co-design with citizens and data-sharing formats, e.g. by enabling citizens from specific districts to help design their district based on existing data, and, possibly, by integrating new data provided by citizens.

A data trust does not have to be an IDP (see box 1). For example, a data trust could be an institution that manages metadata (who controls what quality of data, including its provenance) and data-sharing tools, thus acting as an intermediator between data controllers and beneficiaries (O'hara 2019).

A relevant example is POSMO (Positive Mobility), a Swiss-based collective mobility data platform that manages member mobility data (POSMO 2020). Mobility data can be used to generate mobility profiles for cities, which in turn are valuable for the design of shared mobility systems. The collective operates as custodian of its members who need to acquire participation certificates. In turn, member decide on how their data is used, e.g. for low-carbon urban planning.

A last design solution is decentralized and based on blockchain technologies. Blockchains may be the best solution in cases where either the capacity for or trust in an IDP is lacking, possibly in large parts of the world (Herko 2019). Via smart contracts, blockchains would guarantee trusted and safe interaction and payments for smart mobility. Via distributed ledgers—i.e. communication protocols that enable administratively decentralized, replicated databases—cryptographic security, auditability, and automation of business processes can all be guaranteed. A global blockchain infrastructure for shared mobility could be built on Linux open-source Hyperledger project (Herko 2019). Because of the resulting very low transaction costs and high degree of trust, a blockchain-based approach to shared mobility could unleash a wide variety of new business opportunities and solutions.

A key question is whether decentralized blockchain technologies can also deliver public goods and other IDP goals. In a direct way, the answer is negative, since data-based solutions based on data integration are not possible. However, there are two ways in which blockchains could still be compatible with the goals of reducing congestion and improving air quality. First, individual users could voluntarily share their personal data for specific purposes, also via blockchain technologies (data donations), or sell the data individually and in various formats to private enterprises. Second, distributed ledgers might also contain information on time, location, situation, and vehicle, thus coding for congestion, fuel quality, CO₂ emissions, and other issues of public interest. Payment transactions could consider these properties, and protocols could be designed to ensure that economic incentives align with the provision of public goods. This solution would require considerable development in blockchain technologies as well as considerable public policy in order to integrate all mobility users into the decentralized system.

In the case of Israel, a tender for the provision of an IDP is envisaged, with potential realization as a public-private partnership, combining specific data from government agencies and digitalization expertise of specific companies. This construct has the advantage that an IDP will be underpinned by reliable capacity. It however is also at risk of realizing monopoly power or abuse and of insufficient participation of citizens. Hence, even if realized as public-private partnership, it should be governed by strict principles and possible realized in form of a data trust.

The various options for an IDP may appeal to different audiences. To avoid gross misuse and enable free market entry for start-ups, some public action should be implemented to provide an IDP as a service to various constituencies. This could happen under a regulatory regime, a government-controlled IPD, or via a data trust.

The experience of Alphabet's Sidewalk Labs is instructive here (Carr and Hesse 2020). Sidewalk Labs is developing a waterfront city district in Toronto, Canada, engineering the new district with all kinds of smart data. The main obstacles to advancing the project are that residents do not trust the company, especially concerning the issue of data residency (Leszczynski 2020), and the municipality does not have the capacity to govern the data interface itself. A data trust, not under control, would be a step towards solving this conundrum. As a non-interested institution, trusted with the data of the new district's residents and infrastructure, it could serve as an interface between the company, the municipality, users and citizens.

The ideal design of an IDP clashes with the political economy of data accumulation and path dependency in institutional development. Data incumbents, such as TNC, have in many cases already accumulated numerous data. There is little incentive to forfeit this position of power. In fact, in many cases data-based start-ups and companies are explicitly motivated to gather and monopolize data and if only to become the next 'unicorn', start-up company to be valued above a $1 billion. Arguably, data rentiership and financialization of data became already the main (and problematic) driver of innovation (Birch et al 2020). The same may be true for government agencies who start to gather data. Realizing the power they gain, administrators and personnel have little incentives to give up on this role. Instead, prospective data rentiers employ strategies of deflection and concealment to avoid any compromise to their data monopolies.

An important question is whether already-monopolists such as Google and Uber would be willing to share their data. The likelihood is small as data rentiership and expropriation is their business model. However, for mobility providers, at least, there is a way forward even for municipal administrations: licensing of shared mobility modes and services can be made conditional on data residency, data privacy and data sharing arrangements. Especially smaller companies, such as Tier, Lime and Bubble may have few issues here, also given that they will profit from an IDP that allows them to optimally (and given appropriate regulation also inclusively) offer their services. A similar arrangement for globally dominant data monopolists, like Google, may be elusive at municipal level. Instead, it would be the task of larger, possibly supranational entities, such as the EU, to mandate data sharing arrangements for products such as Google Maps.

In the specific Israeli case, data have been accumulated by several shared mobility providers, including e-scooter services, like Lime and Tier, and the shared pooled provider Bubble, most of the active in Tel Aviv. However, each controls only a small segment of mobility data. A much more comprehensive data set is controlled by Ayalon Highway (tellingly highlighting 'innovative' as their main attribute), a private subsidiary of the Transport Ministry, which is in charge of managing the data of 'alternative', and thus tracking mobility patterns of car users. It would be important to convince stakeholders in Ayalon Highways and among other incumbents that an IDP, operated as a common good, is in the wider public interests and deserves their active support.

In this paper, we have discussed the opportunity to advance the notion of a smart city that remains mostly vendor-based to make use of more valuable urban data platforms and go the next step by designing new institutions that use integrated data management in support of goals such as user-retained data ownership and control, mobility justice, and low-carbon transport systems. We propose further systematic evaluations of success conditions for shared pooled mobility and data governance. While systematic reviews have already captured well the contribution of urban transport for sustainability goals (Javaid et al 2020), the question of urban data governance for sustainability is still new and deserves comprehensive evaluation. The analysis suggests that realization of an IDP that supports sustainability goals requires strong political leadership and a civic-minded active communities of mobility users, transport system providers, and data engineers.

FC acknowledges support by the Heinrich-Böll-Foundation, the Israel Public Policy Institute (IPPI), and hosting by Israel's Prime Minister's Office.

No new data were created or analysed in this study.

The corresponding author states that there is no conflict of interest.

Altogether, an IDP may have the following desirable properties (following (O'hara 2019)):

• (a)  
  Transparent interface. Users need to be able to ascertain the existence, properties, and quality of the data in the first place.
• (b)  
  Provenance. Potential users need to be able to assess the quality of data by getting access to metadata about its provenance and other properties.
• (c)  
  Access controls. Data controllers—managers of the IDP or third-party data controllers relying on the IDP as an interface—need to be able to retain control over who gets access. This can follow transparent rules, or individual negotiations, especially in a more decentralized data trust format. The liability for data protection breaches remains with the IDP or individual data controllers.
• (d)  
  Access. There must be mechanisms and tools for obtaining access to data. Access may be partial, and limited to synthetic data, or limited to data-renting formats. Access may also be conditional on, for example, contributing to the overarching goals of the IDP.
• (e)  
  Identity management. The identity of those gaining access to data is tracked.
• (f)  
  Audit. A record of uses of data needs is stored subject to audit for compliance with legal requirements, ethical principles, and the mission of the IDP.
• (g)  
  Accountability. Data controllers and data users must both be held accountable in the case of data misuse.
• (h)  
  Impact analysis and learning. Use and value generated, but also misuse, must be recorded and evaluated to constantly improve the design of the IDP through iterative learning.