Beyond matters of fact
«We cannot characterize political ecology by way of a crisis of nature, but by way of a crisis of objectivity» (Latour,
2004 p. 19). The joint crisis of modernity, science and nature leads Latour to assert that the first real step towards a real political ecology is to change the way we conceptualise objects in order to then be able to act on them in terms of design. Latour defines matters of fact as the modern way of conceiving of objects, and it is interesting to note how the characteristics of this perspective illuminate many of the problems we are facing in AI urbanism. First of all, matters of facts are objects defined with clear boundaries and which are considered for their internal efficiency (Latour,
2004). It is easy to recall how efficiency is at the heart of the technosolutionistic rhetoric on the urban application of AI and how this object is evaluated in itself as a bearer of order regardless of context. Traffic (Caprotti & Liu,
2022), policing (Tulumello & Iapaolo,
2021), urban infrastructure (Cugurullo et al.,
2024), housing (Rosen et al.,
2021), urban management and design (Son et al.,
2023): all these domains are supposed to be made more efficient through AI because this is an object bearer of order and rationalisation. However, such conception problematically understands AI as a mere instrumental technology to be applied to various fields. Not as a technology capable of completely redefining the fields into which it is grafted, by changing their logic (Coeckelbergh,
2022). Moreover, through this feature, AI is conceptualised as a technology separate from society. Especially AI companies and a relevant portion of the research in computer science and engineering promotes a vision in line with the idea of science that we criticised in the previous section.
The second characteristic concerns the concealment of the technical and material production mechanisms of objects (Latour,
2004). In the AI context, this characteristic develops in two senses. Firstly, the great advances in AI in recent years are based on the large amount of data that our society makes available on which machine learning systems can be trained (Lee,
2018; Floridi,
2022). However, the quality and quantity of data that literally informs the ability of these systems to act can contain biases, errors, inaccuracies that affect the very functioning of these tools (Floridi,
2022). Moreover, as already mentioned, all steps in the production of AI have considerable environmental costs. Like many digital technologies, the material trail connected to these tools is extremely articulated and starts from the extraction of rare materials to then arrive at their application through the assembly of components, the structuring of subsystems and the programming of platforms (Sætra,
2023b). Moreover, AI specifically adds material costs in terms of emissions and water consumption, which are also linked to the training phase and management of data centres (Van Wynsberghe,
2021; Coeckelbergh,
2021; Brevini,
2021;
2023). In the urban context this hiding of the logical and material production processes of urban AI is certainly an obstacle to a just and sustainable use of these tools.
The third characteristic concerns the fact that the implications of these objects are always seen as external to the objects themselves, as if belonging to different universes. In the case of urban AI, this approach is particularly evident when considering the social, ethical, and environmental implications that accompany the use of these tools. These implications are very often addressed in diverse fields such as sociology, ethics, urban planning, environmental sciences, not affecting the conceptual and technical composition of these systems. Finally, according to Latour (
2004), the consideration of objects as matters of fact never includes the long-term risks associated with them. These risks are interpreted as historical contingencies that may occur but are not included in the initial definition of the object. Two examples are extremely interesting in the field of urban AI. Firstly, AI is becoming decisive as an infrastructure in our societies both on its own and as a mechanism for enhancing and supporting existing infrastructures (mobility, waste and electricity management systems, for example). The current rush to artificially upgrade our infrastructures conceals the medium- to long-term risk of getting locked into the compulsory use of AIs which, as we have seen, come with considerable environmental costs (Robbins & Van Wynsberghe,
2022). Making AI fundamental for infrastructure, without taking its costs into account, risks leading us into an unsustainable future (ivi). Finally, the increasing autonomy of urban AI is leading and may lead in the long run to questioning the very concept of the city as something created by humans for humans (Cugurullo,
2021). Indeed, while AI is profoundly changing the ways in which cities are planned and managed (Batty,
2023; Son et al.,
2023), even the very composition of the citizenry is affected by these logics: in experimental settings such as the Neom project in Saudi Arabia a robot may even become a full citizen (NEOM: Made to Change, n.d.). These poorly considered risks could lead to significant changes in what we mean by
city, thereby challenging its very definition.
Towards a non-modern theory of design
As we have seen by analysing the literature on smart and eco-cities, the simple application of technology to increase the efficiency of our cities does not seem to be enough in the Anthropocene (Cugurullo,
2021). Some systemic changes are needed in the way the habitability of our cities is thought about (Elmqvist et al.,
2019); a new design theory seems necessary to try to assemble technology and urban space together with the goal of sustainability (Yigitcanlar,
2018). Through Latour’s thinking, we have realised that we need to change the way we look at AI by abandoning the moderns’ concepts of science and nature and going beyond the perspective of matters of fact. According to the French philosopher, the concept of design has five specific principles concerning a new theory of action that can be useful, in the context of this paper, to conceptualise the relationship between urban AI and sustainability.
The first two principles visibly contrast with contemporary trends in AI urbanism. The first relates to humility, that is the willingness not to be always foundational and revolutionary (Latour,
2008). The design of urban AI does not have to follow the path of radical disruption, which is evident for instance in AI urbanism projects such as the linear city The Line in the Neom project in Saudi Arabia, which aims to be «a revolution in urban living». The very slogan of the Neom project is «made to change» (NEOM: Made to Change, n.d.;
THE LINE: a revolution in urban living n.d.) to signify a radical break with the past all within that concept of modernity that Latour intends to overcome and that in contemporary urbanism we can define as
transurbanism (Cugurullo,
2021), that is, the idea that technology will produce an autonomous solution to urban problems. The great power of AI and the great fragility of Gaia would suggest, instead, an experimentalism of a different kind (Bulkeley,
2021). A diffuse and small-scale experimentalism that could test the actual functioning of urban AIs in the ecosystems in which it fits. This approach would avoid the already highlighted risk of being trapped in environmentally expensive and socially damaging infrastructures (Robbins & Van Wynsberghe,
2022). In this light, the attention to limits highlighted both in the history of urban ecology (Endlicher,
2007) and in Latour’s reflection (Latour & Schultz,
2023) is extremely relevant. Being revolutionary and foundational might have been a reasonable approach in the face of Galileo’s inert nature capable of absorbing every shock, but it appears to be a suicidal strategy in the magmatic context of Gaia.
The second characteristic is closely linked to the first one and involves attention to detail. The idea is to have an almost craftsman-like attention to detail in design and not to proceed without considering the implications of actions or thinking that these will only be addressed at a later stage (Latour,
2008). In the design of sustainable urban AIs, therefore, it is key to first understand and publicise the impact of these technologies on the various collectives that populate the city, namely how the placement of robots, AVs or software agents can have costs that are distributed differently in the urban space (between centres and suburbs, between different ethnic communities, between different social classes, between people of different ethical orientations, between different ecosystems) (Bulkeley,
2021). Furthermore, a sustainable and detail-oriented urban AI cannot fail to incorporate within itself an indication of the environmental cost (materials, energy, emissions) associated with its assembly, training and deployment to enable citizens and policy makers to make an informed assessment of its use (Van Wynsberghe,
2021).
The third characteristic concerns the focus on purpose (Latour,
2008). A sustainable urban AI must move away from the logic of efficiency as a fundamental purpose (Hodson & Marvin,
2010; Ahvenniemi et al.,
2017; Yigitcanlar,
2018; Mora et al.,
2021). In fact, efficiency is what we might contradictorily define as an instrumental purpose: something is efficient with respect to a purpose and not in itself, and we must always ask ourselves whose benefit a technology is efficient for. A sustainable urban AI must emerge from the goals discussed by the collectives that populate the city and arise from that dialogue between science and community that we examined in the previous section (Latour,
2017; Turnhout et al.,
2020). Scientific reflection and technological solutions must arise from the concrete interests and needs of communities, overcoming a modern conception of science. The fourth characteristic concerns the fact that design is always re-design (Latour,
2008). That is, the design of urban AI is always situated in a context of humans and non-humans with specific needs. The critical literature on AI urbanism has already highlighted how the application of urban AI takes little account of contexts (Bratton,
2021) and how it is often believed that AIs should be the main intelligences in the cities of the future (Lynch & Del Casino,
2020), that is, the ones capable of coordinating the others. It has also been pointed out that this mindset contradicts any claim to sustainability (Palmini & Cugurullo,
2023). Sustainable urban AIs must first and foremost be put at the service of preserving the habitability conditions of the various collectives by dialoguing with the natural ecosystems that are the foundation of urban life. They must also fit into the relationships between intelligences (human and non-human) in order to foster cooperative cohabitation while respecting specific differences. Situating urban AI in Gaia means putting it at the service of habitability and differences and not of efficiency and homologation. The last characteristic concerns the importance of politics, the idea that one should always discuss whether design is right or wrong, suitable or unsuitable for a certain context. Following Latour «no designer will be allowed to hide behind the old protection of matters of fact. No designer will be able to claim: “I am just stating what exists”, or “I am simply drawing the consequences of the laws of nature”, or “I am simply reading the bottom line”»(Latour,2008). Designing a sustainable urban AI therefore means abandoning the problematic concept of progress that animated moderns in thinking both AI (Coeckelbergh,
2022) and urban AI (Cugurullo,
2021). This design culture must not only ask how AI can change the city but also how the urban ontology - made of differences, clashes, cooperations and balances - can change the very logic of AI.
The suggestion that can be made to design sustainable urban AI through Latour is to start seeing it as a matter of concern. With this term, the French philosopher designates the new type of objects that we have to deal with in this time marked by a global climate crisis. These new objects have no clear boundaries but must be defined in relation to the environment in which they are located; their material, scientific and technological production has to be made visible; they must be not independent of their implications and impacts but determined by them and must by definition acknowledge the possible long-term risks associated with their use (Latour,
2004). According to Latour, as we have seen, in order to establish a true political ecology, it is necessary to question the concepts of modernity, science and nature that have accompanied us so far and to change our way of understanding objects - from matters of fact to matters of concern - and to change them through a new design theory. We believe that sustainable urban AI design needs this reflection to break out of the impasse created by the juxtaposition of smart and eco urbanism. In an urban reality in crisis that needs to be redesigned, an urban planning that merely repeats the paradigms of the past seems to be ineffective and, worst, harmful. The era of AI and the Anthropocene demands a radical reflection on the way we inhabit the world, and Bruno Latour’s philosophy certainly points a way forward in this regard.