How relevant is the digital twin to the urban world?

Abstract

Digital twins are computer models that provide different but connected simulations of some particular system such as a city. They are used to explore how the system can be controlled, managed, and designed in ways that enable us to develop a better understanding of different city systems.

The term ‘digital twin’ was first coined over 20 years ago by Michael Grieves (Grieves and Vickers, 2023). He applied it to digital simulations or models of well-defined physical systems, often focussed on mechanical-aerodynamical structures from which one could mount experiments to figure out how such models could be used to understand, predict and control the form of the real system. The real system itself is now increasingly referred to as the ‘physical twin’ – an analogue to the digital models of the system. As the name suggests, digital twins are generally computer-based in their representation. As the term has matured and found increasing use in the wider field of simulation, what is now called a digital twin has broadened in scope with digital twins being extended to embrace related models coupled together with the ‘human in the loop’ as an essential component of the integration (Grieves and Vickers, 2017). In short, digital twins increasingly link different models (or twins) together, as coupled or integrated systems, federations of twins, which are continually under development as new information technologies evolve. For example, contemporary computing methods based on big data and AI are fast becoming part of a constellation of tools including digital twins as models and methods that define all computing platforms and environments.

The earliest articulation of the digital twin was predicated on the basis that the twin be as close as possible to its physical counterpart. It cannot, however, be equivalent to the physical twin, otherwise it would be the twin. This is a contradiction in terms because a physical model is not a digital model either. In short, a digital twin interacts with its physical twin through the passage of information between the physical and digital and vice versa. This functionality drives the twin and its use in simulating interactions and controls between the real and digital worlds. Both digital and analogue are models, abstractions, and simulations, that interact with one another in diverse ways, and in this sense, we are only at the start of exploring the wider meaning of digital twins in science. In my commentary here on Rose’s (2025) article, she introduces many different perspectives on the concept of digital twins – reflecting particularly on cultural mores, on visualisations, on games and movies, on the role of the individual scientist, policy-maker and on how individuals as groups mesh with one another in the wider context of building and using digital twins in various types of public participation. These are all critical to elaborating their role for understanding, predicting, designing and more generally organising the way such twins might be used in analysis of various kinds.

Digital twins can be used in many ways and to this end, they have very different roles in any area of theory and application, particularly in the domain of geographical and urban analysis, the focus of our interest here. In the narrowest sense, physical and digital models are directly useful for explaining the way the physical twin fits together in interfacing with the digital twin. This can be developed in standardised form building on theories and data that define the way the city or the urban system functions as a sequence of theoretical propositions defining the components of the city system that interact with each other, simulating the flow of materials, energy, information, and capital between their various parts, This is the traditional way of defining digital twins that pertain to cities, what Rose (2025) calls City Digital Twins (CDT) but there are many other ways of modelling such systems. In fact, the idea of the CDT can be applied to any and every way of defining the city system in digital terms (Batty, 2024). She defines two specific approaches: first elaborating the visual development of the city digital twin in terms of the way human behaviours interact with the three-dimensional forms; and second, defining ways in which energy is associated with building forms as transport and other types of flow system through visual media brings the idea alive in different applications. All these approaches weave into one another, thus providing a very rich kaleidoscope of forms and materials that twist the notion of a geometric digital twin into different pathways. These focus the twin on a variety of ways in which we can view the city in its physical form in digital but visual ways, drawing out our ideas from a wide array of digital forms.

City digital twins do not emerge phoenix-like from the physical twin but they are constructed usually in computable modules from the bottom up. The basic building blocks of course are digital components of the city. These may be buildings themselves or street segments and other geometric features that can be positioned all over the terrain. Sometimes they include the location of activities such as populations reflecting different degrees of spatial abstraction and resolution. They can be assembled as modules which are usually organised as hierarchies. Many spatial software systems which are extensions of GIS using remote sensing and 3D LIDAR data, are being collapsed into systems such as ESRC’s City Engine (2024) and various developments in geodesign technologies (Steinitz, 2016) are using these technologies supporting this quest. In fact, over the last 40 or more years since three-dimensional computer graphics reached the point where buildings and city blocks could be modelled and visualised, digital twins have continued to evolve as computer graphics technologies have developed. Early visualisations appeared as soon as the personal computer and workstation technologies were invented where the dominant interface was graphical. If you look at the way such models developed in visual terms – and of course they were not called digital twins until much later – then it is easy to see that there is not one digital twin but an infinity of such twins as they develop with the user continually evolving the digital twin as more and more realistic and abstract conceptions are added (Batty, 1987).

In fact, the earliest building blocks of digital twins are components of maps that are assembled to generate computer cartographies mainly on the two-dimensional flat plane. Because the map can be projected rather easily onto a flat plane, they are somewhat easier to relate to hand-drawn images and pictures and it is not surprising that maps often display a wide array of pictograms, a good deal more extensive than the assemblage of components that usually make up a city digital twin. As soon as we open up the map and the model to the world of fiction and speculation, CDTs can reveal a wide range of possible forms and imagery. In fact, although CDTs are largely used as models of real physical systems, they are being developed to simulate fictional cities, that inform our thinking about possible futures in terms of their imagery. Indeed, developing a digital twin provides a framework for generating many different alternative forms which can be tested ‘to destruction’ so-to-speak. Indeed, a wide range of city forms can be generated in this way, and as soon as we broach these possibilities, our digital twin can be parameterised and tweaked to generate literally thousands of alternative forms which bound the solution space within which we might be searching for optimal patterns. Once we enter the world of future scenarios, we are always dealing with more than one twin, an infinity in fact as part of our speculations in an open-ended future.

I will pose several questions that spin off from Gillian's essay which are somewhat independent of the technical process of constructing digital twins. We can identify these questions as follows:

who builds digital twins and what theory and applications do their builders bring to bear on their forms and functions?

what kinds of scientists build such twins and what are the main scientific questions that are defined?

what kinds of users, different from the narrow professional field of their builders, are associated with digital twins and how do these users relate to typical and thence atypical scientific questions generated from CDTs?

what kind of social scientists with little or no training in policy applications and/or in the scientific basis of digital twins as well as users who are concerned with bias, uncertainly and privacy in system, use or at least have the potential to use and adapt such twins to their own concerns?

To an extent, my own critique here which relates to Gillian's focus on the types of individuals who are interested in the qualitative aspects of CDTs, is part of this wider question involving who are ‘the human(s) in the loop’? The question of who builds the CDT brings to the fore the notion of what the dominant skills and ideologies are of those who construct such artefacts. Computing is a singularly male-dominated occupation. This has been known since computers first appeared and it relates to the nature of male domination in science. This may be due to the bias in early and continuing education, but it is also possible to relate it to intrinsic differences between men and women in how they think about the world and its construction. There are many statistics revealing this although these questions involve the most human of properties relating to the construction of the twin and even the construction of the physical twin to which it is matched. A reasonably authoritative survey from many estimates of how gender relates to the development of software systems suggests that some 92% of all coding is the prerogative of males (Statista, 2024). There are few explanations for these differences which clearly relate to the way our minds work as well as power structures in society but for which we do not have easy explanations. My own view is that this will never really be understood and the last thing we might expect is that some sort of artificial intelligence will provide the basis for such an explanation.

The history of the computer in terms of hardware, software and probably dataware is not only male dominated but when it comes to those who rise to the top and are associated with the most dramatic advances, the list of female contributions is sparse. Lovelace (1837, 2024), Lord Byron's daughter, is credited with inventing ‘programming in the early nineteenth after century, Hopper (1947, 2024) who worked in the Harvard computer lab in the late 1940s but was really seconded from the Navy, was instrumental in the development of early operating systems, and much more recently Fei-Fei (2023) invented ImageNet which is a key development for machine learning in generating images, and for which I am guessing and hoping she might ultimately get a Nobel Prize. But these contributions simply serve to show that CDTs and related technologies largely depend on a masculinity that dominates the field. Although it is hard to figure out what all this means for what is constructed and how it is used, gender must be somehow factored into the production and consumption of such computational procedures and devices.

The focus on games has a less skewed distribution of contributions by gender but a lot of the content with respect to war games, adventure games, and even puzzles like AI games such as Alpha-Go continue to betray a male bias which is undoubtedly reflected in their construction. I must admit to not knowing what that bias is in terms of usage of any routine software where I assume the balance of gender is more even. But as our analysis rarely focuses on designers and users but more on the games or CDTs themselves, many would argue that although a balance should reflect the contributions of those who build CDTs and those who use them, even this assumption is contestable. In Gillian's essay, there are several other themes, but the importance of her contributions is to turn the evaluation of CDTs as an example of contemporary software systems on their head, arguing that those who build and use such systems must reflect on the role of gender in the process as well as race and ethnicity on which there has been little work despite an undercurrent of speculation for many years. In a sense, her essay is in the vanguard of how we should begin to evaluate many different technocratic constructions and applications with a view to extending their use to reflect the relevance of race and gender in computation.