Participation matters: The social construction of digital twins for cities

Participation in smart cities

UDTs are seen by many as the latest wave of smart city innovation, integrating smart technologies and techniques such as ICTs, sensors, IoT, big data, data science, AI, Building and City Information Modelling and automation (Calzati and Van Loenen, 2023; Dembski et al., 2020). In smart cities, participation is often discussed in terms of stakeholder collaboration within double, triple, quadruple and quintuple helix models following developments in theorising innovation (Paskaleva et al., 2021). A key focus in all of these is on ‘who participates’, with a goal of bringing different types of stakeholders to the table – that is, into the ‘smart city ecosystem’. Stakeholders are viewed as representatives of archetypal societal domains of government, business / industry, academia / research and civil society, each representing largely homogeneous views from their respective groups. Interactions between stakeholders are most often described uncritically as collaborative (cf. Mills et al., 2021).

The double-helix model represents a collaborative arrangement between two stakeholder types, with configurations of business-academia and business-government considered in the literature (Grundel and Dahlström, 2016; Mora et al., 2023). A combination of these models evolved into the influential ‘triple helix’ model in which academia generates intellectual capital, governments represent the public sphere and finance innovation, and businesses bring innovations to market to create wealth (Borkowska and Osborne, 2018). Against this largely top-down conceptualisation of governance and innovation, more recent quadruple and quintuple helix models emphasise an additional role for bottom-up processes (see Figure 1). They explicitly include citizens and community organisations, and the ecological and environmental context, respectively, as part of the helices (Crumpton et al., 2021).OPEN IN VIEWER

Double and triple helix models have been criticised for promoting smart city innovation that is overly technology-focused, instrumental and technocratic (Grossi and Pianezzi, 2017; Nochta et al., 2018). The shift towards quadruple and quintuple helix models aimed to address this issue, however the role of civil society often remains confined to narrow ‘user’ and ‘consumer’ interpretations – even in smart cities that claim to adopt a ‘citizen centric’ ethos (Cardullo and Kitchin, 2019). Urban Living Labs – designated areas in cities to accommodate smart experiments and demonstrators – have become popular to facilitate bottom-up involvement and civic innovation, but evidence on their real-world impact and contribution to changing the dominant n-helix stakeholder conceptualisations is, at best, mixed (Engels et al., 2019; Paskaleva et al., 2021). Further, even those advocating for integrating bottom-up citizen and community involvement often seem to cluster civil society into a homogeneous ‘unit’ with similar views and interests (Borkowska and Osborne, 2018). Other stakeholders with potentially differing views and roles are also bundled together to represent the public and private sectors, and research and education, with their roles assigned based on this simple, high-level categorisation.

Few publications analyse participation patterns in smart city governance and innovation looking beyond the somewhat simplistic ‘who participates’ lens. There is a lack of focus on how participation takes place, and how more nuanced stakeholder ‘categories’ could or should be developed to better understand participation patterns and their impact. One exception is provided by Cardullo and Kitchin (2019) who focus on the process aspect of citizen participation in Dublin’s smart city projects. Beyond examining the forms, levels and impacts of citizen participation, their analysis also highlights the diversity of views and experiences of different individuals and communities vis-à-vis the smart city. Another interesting example is Soutar et al.’s (2022) analysis which develops a nuanced account of processes, stakeholder types and roles in smart local energy projects. They investigate how ‘project partners’ engage with different types of stakeholders which they categorise as ‘users’ and ‘communities’, highlighting differences and identifying several different types within each category (Soutar et al., 2022). This categorisation enables the authors to reveal complex patterns of engagement and context-dependent project development trajectories, going beyond the explanatory power of the n-helix models.

In summary, whilst the evolution of n-helix models of smart city innovation towards increasingly broader participation is a welcome move, their analytical and practical potential has some limitations. As the above examples show, exclusive focus on n-helices is likely to result in downplaying within-type differences whilst also over-emphasising differences between stakeholder types and over-simplifying roles in smart city innovation. Analytically, this represents a problem for unpacking place-based and context-dependent complex stakeholder constellations and roles, and by extension, smart city outcomes. In practice, the n-helix models offer little guidance on assembling and facilitating participation beyond cross-domain collaboration. Nuanced approaches that integrate insights from various theoretical perspectives to address the aforementioned diversity-, context- and process-related discrepancies could contribute to advancing our existing knowledge on smart city innovation, with important implications for the development and implementation of UDTs.

Participation in urban planning and analytics

Another perspective sees UDTs as the next generation urban analytics tools for urban planning and management. In urban planning and management, the term ‘participation’ is primarily used to refer to citizen engagement (rather than multi-stakeholder collaboration). It is generally seen in a positive light, giving voice to citizens and communities through processes of consultation, collaboration and co-production to shape decisions that impact the environments in which they live (Healey, 2020; Tippett and How, 2020). This is against a conceptualisation of decision-making dominated by interactions between the ‘planning establishment’ (planning committees and officers, developers, investors, etc.) and the ‘urban growth regime’ (elected leaders, businesses, trade associations, universities, etc.) (Laskey and Nicholls, 2019).

The interests of the planning establishment and the growth regime converge around generating economic growth, tax revenue and profits, with local citizens and communities expected to benefit but also borne arising externalities. In parts of the planning literature, citizen participation is intended to negotiate, or secure acceptance for, proposals for urban developments and futures (cf. Wilson et al., 2019; Xu and Zhu, 2021). In others, it is seen as a necessary means to bring diverse local knowledges and experiences – ‘situated’ in different social-cultural settings – into decision-making that complement and/or challenge prevailing understandings in the planning establishment on matters of shared concern (Healey, 2020). Studies of participation patterns (inclusion and exclusion) have a focus on process, but tend to concentrate on often conflictual interactions at the boundary between the regime and the establishment on one side, and citizens and communities on the other. Differences, interactions and conflict within the two sides, and their implications, are less often discussed.

A core concern in both research and practice is process-related and extends to designing appropriate participatory mechanisms for different planning problems and goals. Arnstein’s ‘ladder of participation’ has been an important inspiration in such studies (Arnstein, 1969). It describes 8 levels that represent degrees of participation and transfer of decision-making powers to citizens and communities, placed on a spectrum from ‘tokenism’ to complete ‘citizen control’ (Arnstein, 1969). It has become extremely influential since its publication, promoting increased citizen and community engagement and participation in urban planning (Ianniello et al., 2019; Sharifi et al., 2023). At the same time, some critics have pointed to the risks of simplifying Arnstein’s message by (implicitly or explicitly) assuming that deeper forms of participation are more desirable in any given situation or context (Maier, 2001). Others have also questioned the real-world potential of institutionally designed participation to facilitate genuine and impactful bottom-up processes (Laskey and Nicholls, 2019).

Against this backdrop, urban analytics tools are often seen as another means to uphold the status quo, given the technical and professional knowledge required to navigate, apply and audit them (Williams, 2020; Wilson and Tewdwr-Jones, 2021). Nevertheless, advancements in the digital sphere have also opened up new opportunities for participation in planning decision-making. Examples include, but are not limited to, the use of participatory or voluntary geographical information systems (Daniel and Pettit, 2022), citizen science (Williams, 2020) and decision theatres (Wolf et al., 2023). These examples showcase the opportunities that digitalisation may offer for integrating broad participation processes into planning decision-making. However, in the mainstream, the function of digital tools remains largely focused on one-directional information provision or gathering (preferences, feedback) as opposed to more active forms of participation – such as community or co-design including of co-design of the tools themselves (Potts and Webb, 2023; White et al., 2021; Wilson and Tewdwr-Jones, 2021). So, whilst the urban planning literature has investigated the participation of citizens and communities in (digitally enabled) planning decision-making, and provided evidence on the impact of a variety of types of interactions beyond collaboration, there has been less focus on embedding participation in the design and implementation of urban analytics tools such as UDTs.

UDTs at the Nexus

UDTs can be understood both as the latest smart city innovation and the next generation of urban analytics tools. To better understand the implications of inclusion and exclusion participation patterns for UDT development, it is therefore helpful to investigate how these perspectives discuss participation. Existing studies on smart city innovation and governance emphasise the participation of stakeholders from various societal domains in technology design and implementation. However, they assume that stakeholders are willing and able to participate – in ways pre-determined by their societal domain – and that interactions are collaborative in nature. In turn, the planning literature highlights the importance of power and resource inequalities, competition and conflict in stakeholder interactions, participation patterns and outcomes. Existing studies indicate that such factors may inhibit more active or meaningful forms of citizen and community engagement in planning decision-making, with participation via digital means often restricted to providing and/or collecting information. However, the analytical focus often remains on the broader planning system and decisions rather than on technology design and implementation. Despite the limitations, both strands of literature provide important insights for UDTs and participation.

A few studies have started to explore links between the concepts of UDTs and participation specifically. Among these, some build on the n-helix models (Hämäläinen, 2021; Nijkamp et al., 2023), whilst others explore the use of UDTs as tools to facilitate communication among decision-makers and civil society in urban planning and management (Adade and De Vries, 2023). Some have also started to discuss how participation patterns might shape (built environment) digital twin development. For example, Solman et al. (2022) conceptualise ‘digital twinning’ processes as ‘acts of governance’ to highlight the impacts of participation patterns on digital twins in wind energy sector. Korenhof et al.’s (2021) analysis foregrounds both the role(s) of designers, surrounding institutions, users, and other stakeholders in steering the content of digital twins as necessarily incomplete representations of reality, as well as the role of the twin models in controlling physical systems. Others have expressed critical views on the n-helix models’ potential to facilitate broad participation for UDTs, with Calzati and Van Loenen (2023, p.38) stating that ‘the quadruple helix is not enough; it should be better regarded as the baseline, rather than the optimum’. Expanding on these perspectives, in the next section we illustrate how the Social Construction of Technology (SCOT) lens can contribute to interpreting and analysing multi-actor participation in the design and implementation of UDTs.

Introducing the core concepts of SCOT

The SCOT approach is operationalized through its constructs of Interpretive Flexibility, Technological Frames; Relevant Social Groups; Closure and Stabilisation, and the Wider Context (see e.g. Oti-Sarpong and Leiringer, 2021). Interpretive flexibility refers to some form of ambiguity of meaning that surrounds technological artefacts (Pinch and Bijker, 1984). As interpretations of a technology change over time, interpretive flexibility can decrease through processes of stabilisation, or increase when new junctures emerge (Bijker et al., 2012).

A technological frame is a collective lens that emerges from converging interpretations of, and meanings attached to, a particular technological artefact by a social group. Whilst it results from actors’ interpretations, a technological frame in turn also guides, and constrains, such interpretations (Basu, 2023). A technological frame may include both material and non-material elements such as actors’ problem framings, goals, current theories, tacit knowledge, practice (e.g. design methods and criteria) and specialized testing procedures (Bijker, 2001, 2010). A technological frame is specific to an actor group, place and time. Hence, technological frames are context-dependent products of thoughts, past experiences and accumulated knowledge of similar technologies (Leonardi and Barley, 2010). The composition of technological frames is not fixed, and actors (un)wittingly mobilize these to (re)shape a technological artefact to reflect collective interests of relevant social groups. However, once a technological frame emerges it will shape understandings, behaviours and interactions in the social group that subscribes to this frame (Prell, 2009).

Relevant social groups are groups of actors who coalesce around particular technological frames that reflect their shared interpretations of a technical artefact (Bijker et al., 2012). In contrast with the helix models’ categorisation of actor groups based on societal domains, relevant social groups in SCOT are constructed based on the proximity of actors’ views about the technological artefact (Bijker et al., 2012). In another important difference with the n-helix models, social groups change over time in tandem with technological frames. Previously separate social groups may merge as their interpretations converge, reducing the artefact’s interpretive flexibility (Orlikowski, 1992). Interpretive flexibility can also increase if previously common interpretations are questioned by some actors in a social group, or new actors become interested in a technology and constructing new technological frames over time (Doherty et al., 2006).

Interactions within and among different groups and technological frames that lead to collectively accepted interpretations are described as processes of stabilisation and closure in SCOT. Stabilisation may occur within a social group. Individual actors’ views in a social group may be more or less aligned with the common technological frame. The more homogenous actors’ interpretations of an artefact are, the higher degree of stabilisation is achieved within this social group (Bijker, 2010). An example of processes of stabilisation could be the gradual simplification of the design or name of a technological artefact (cf. Prell, 2009). Stabilisation however does not automatically lead to closure given that different interpretations can ‘stabilise’ in different groups. Closure occurs between relevant social groups, and leads to particular meanings becoming, often irreversibly, attached to a technological artefact. It can develop in diverse ways: First, different groups may reach a consensus on the meaning of an artefact. Second, rhetorical closure occurs when relevant social groups perceive the problem as ‘resolved’ – this is often achieved through advertising. Third, the redefinition of the problem that the technology is seen to be addressing can also lead to closure (Bijker et al., 2012).

The wider context comprises a wide array of factors, including geopolitical and socio-cultural elements, as well as legal regulations, conventions and norms; as well as sectoral, national and international factors. Beyond shaping socio-technical interactions, the factors also provide insights about the contextual elements that influence technological developments. The wider context is therefore not merely a geographical locus within which socio-technical interactions take place (Aibar and Bijker, 1997). Rather, an active component in shaping how a technology develops in a place (Oti-Sarpong and Leiringer, 2021).

Despite the novel insights it provides on how technological development unfolds in a place, SCOT has some shortcomings (Basu, 2023). Critics have pointed out the strong emphasis on agency and relative neglect of structural forces that constrain agency and technological development (Klein and Kleinman, 2002). Others have also noted analytical issues which may limit the explanatory power of the theory, including the continuously changing composition of relevant social groups, and the apparent endless nature of closure and stabilisation processes (Humphreys, 2005). However, in our view, SCOT can offer an option to improve the conceptualisation of participation in UDT design and implementation, in the context of existing prominence of structural accounts in both the smart cities and urban planning literatures.

Applying the SCOT concepts in UDT development

Using the SCOT lens, UDTs can be characterised as a technological artefact that carries multiple interpretations and design possibilities (cf. Bijker et al., 2012 p.34). This recognises that no fixed UDT concept exists (yet), providing space for interpretive flexibility. For some, digital twins are sensing systems designed to monitor the behaviour of complex physical systems to identify, and mitigate and/or eliminate undesirable emergent behaviour (Grieves and Vickers, 2017; Lu et al., 2020). Others describe digital twins for cities as cyber–physical–social eco-systems that couple not only the physical and digital system components but also people (Tomko and Winter, 2018). Yet another perspective places the emphasis on the importance of predictive capability in UDTs, and specifically the possibility experimenting with ‘what if’ scenarios (Mohammadi and Taylor, 2021). Although these different interpretations suggest certain commonalities across different visions of UDTs, they also highlight differences and varied options for designing UDTs in (to) practice.

Interpretive flexibility also points to the (co-)existence of multiple technological frames that attach different meanings to UDTs, promoted by different social groups. The UDT concept is thus subject to alternation and variation depending on context, place, envisioned use and actors involved. UDTs are often positioned as innovative data products that keep pace with, and make use of, cutting-edge technological developments (Callcut et al., 2021). This framing tends to exaggerate notions of novelty, expected benefits and universal applicability. Whilst it could be expected based on (assumed) stakeholder interests, this view is not confined to certain companies or research teams acting as product developers (Austin et al., 2020). In practice, in contrast, those working on UDTs of subterranean assets developed different interpretations due to case and place-specific constraints and use requirements. Given the financial resources and effort required to secure access to good-quality data, ‘fitness-for-purpose’, standardisation for data sharing and economic viability have become key considerations. This led to diverging from the focus on cutting-edge technology in general, and led to significantly different trajectories for technology development and stakeholder involvement across different contexts, for example, in the UK (Kessler et al., 2023), Flanders (Rombouts, 2023) or Singapore (Schrotter and Van Son, 2019). Moving beyond the focus on technical content, these examples also highlight the importance of varied interactions (including negotiations or conflict resolution) across a range of stakeholders, and the wider context, in shaping UDT conceptualisations in different places.

The case of the subterranean asset UDTs indicates that relevant social groups, coming together around particular place-specific technological frames, may be more diverse than what the n-helix models can capture. Many of these projects would fall into the double-helix category, with typically only public sector bodies and private sector utility companies involved. However, the examples above showcase the diversity of views and requirements among stakeholders within each type, and of the resulting digital twins. Nevertheless, researchers often tend to stick with the n-helix categories. For example, Batty et al. (2023) argue that different social groups’ interpretations led to three different UDTs for a single location in London, representing different urban scales (Quantitative Urban Analytics (QUANT), Virtual London (ViLo) and Here East (HE)). In comparing QUANT and ViLo, Batty et al. (2023, p.73) conclude that ‘although both these models can be used in public participation, QUANT is more likely to be useful to technical experts, whereas ViLo is more suited to a non-expert audience’. The HE model’s content is driven by university researchers’ interest, with sensors monitoring key performance indicators of energy use in university buildings. Different purposes, and the participation of different groups of stakeholders, culminated in three distinct models for the same place being developed largely in separation (Batty et al., 2023). In the Cambridge CDT / UDT case, Nochta et al. (2021a) report on high-level UDT requirements according to stakeholder background types (local authority, citizens). Subsequent more detailed accounts however show that certain local authority officers and ‘expert’ citizens shared an interpretation of the UDT as a cutting-edge data product – despite not having a ‘stake’ in product development (Nochta et al., 2021b). Other mixed groups downplayed the importance of technical features and considered it as ‘just another’ planning support tool which, however, could contribute to communication across policymakers and civil society (Nochta et al., 2021b).

Despite the existence of multiple technological frames, we can also identify processes that, over time, may lead to some degree of stabilisation or closure. One might recognise an attempt at facilitating rhetorical closure by certain industry stakeholders who view UDTs as smart city innovation, advertising ready-made digital twin products assumed to be applicable to both engines and cities without much modification or reflection (Korenhof et al., 2021). Another interesting example is a gradual move away from ‘real-time’ to ‘near-real-time’ to ‘right time’ in how the frequency of data collection is discussed in relation to UDTs, and digital twins more broadly (Callcut et al., 2021). For example, the status of the Cambridge CDT /UDT as digital twin has been questioned by some who consider it rather as a static model (Jeddoub et al., 2023; Nochta et al., 2021a). It could also be questioned whether subterranean asset digital representations can be considered UDTs. At the same time, there has been a growing recognition that digital twins must be purpose driven. The appropriate ‘refresh rate of data’ is, in turn, determined by this purpose and there are ‘some use cases where using real-time data would not be possible or even necessary’ (Callcut et al., 2021, p.15; see also Austin et al., 2020; Pregnolato et al., 2022; Wang et al., 2023).

As the above discussion demonstrates, a combination of the SCOT concepts of technological frames, relevant social groups and stabilisation and closure can help to analyse and reflect on the current landscape of UDT developments. SCOT concepts appear well-suited to understand how different versions of UDTs may emerge and get implemented in different places and settings, through uncovering inclusion and exclusion participation patterns and their impact, and highlighting the strategies that actors employ to influence UDT iterations. At the same time, SCOT-informed analysis can also reveal structural constraints from the wider context that may facilitate similar UDT constructs in different places or settings – as opposed viewing these simply as being a product of inevitability and technological determinism.