Cities, Climate and Inequalities

Digital technologies to combat climate change: Between problem and paradox

December 2022

Marie-Luc Arpin, professor at the School of Management (Université de Sherbrooke); Sébastien Gambs, director of the Canada Research Chair in Privacy and Ethical Analysis of Massive Data and professor at the Department of Computer Science (UQAM)

Introduction

Over the past decade, the meaning of the word “urbanization” has undergone a surprising and structuring change. Today, against the backdrop of crises (environmental and health, social and humanitarian) which urbanization has become associated with (UN, 1976, 2016), the word evokes above all a set of problematic phenomena, if not a crisis in itself. Concurrently, however, urbanization is becoming a new vector of hope: “a transformative force” from which “sustainable and equitable urban futures” can emerge (UN-Habitat, 2016). In the fight against climate change, in particular, cities are now assuming a decisive role in planning and operationalizing the transition of behaviours toward “low-carbon” lifestyles (IPCC, 2022; Revi et al., 2014).

Still, confidence in this transition is not a given. It brings blind spots and comes with implicit promises and conditionalities regarding digital technologies which, rightly and wrongly, are becoming more prevalent as the climate crisis deepens (IPCC, 2022; Palm and Nikoleris, 2021). Massive flows of data, particularly from mobile devices for individual use, are significantly (re)shaping urban dynamics, so much so that any urban leadership in climate action will have to integrate it into its vision. While megadata are giving citizens and communities new capacities for adaptation, they are also making them vulnerable in new ways. In this sense, the opportunities brought by contemporary urbanization are a double-edged sword, and urban climate action can no longer be conceived in isolation from the (paradoxical) social dynamics of digital technology.

1) Literature review: Digital technology as a vector for urban climate action

In this article, we summarize the findings of the research report Synthèse des états de l’art sur les enjeux éthiques de la ville intelligente (Synthesis of the state of the art of the ethical challenges facing the smart city), authored by us and published in 2022. Our focus lies on megadata (big data) and the artificial intelligence (AI) systems that enable their analysis. We also discuss the “smart city,” which refers to the vast and largely imperceptible influence of AI in the urban context, particularly when deployed by platform companies (big tech). Apart from this dominant trend, other types of use of digital technology are possible. This focus is warranted because the double paradox which characterizes AI illustrates clearly the social dynamics that, if more subtly, accompany other technologies (5G, the Internet of Things, blockchain, etc.) and their integration in the urban context.

As posited in the research report, which draws on a corpus of literature with titles or abstracts that included the key words “city,” “digital,” “ethics” or their synonyms, the social science literature often fails to recognize this double paradox insofar as it frames and highlights the challenges of digital technology in urban environments as a problem, in other words, it subjects digital issues in urban environments to the soluble-problem model^[1] (Vérin, 1984). That said, the social science literature does effectively document the challenges of digital technology in cities. In this sense, it can serve as an empirical basis for theorizing the relationship between urban climate action and digital technologies. The literature demonstrates that in the second half of the 2000s the previously pluralistic discourse of the “smart city” became the preserve of large specialist companies^[2] (Albino et al., 2015; Halpern et al., 2013, 2017; Hollands, 2008). Through these companies, data collection, processing and management practices in urban environments were developed with disregard for (or exploitation of) the imbalance of forces and knowledge. These relationships have been poorly framed to date and are subject to various forms of abuse that need to be denounced (CEST, 2017, 2020; Hatuka et al., 2020).

From the outset, these companies based their rhetoric on environmental urgency and the supposed need to generate new knowledge for combatting it (Halpern et al. 2017). This rhetoric was essentially retained as the discourse transitioned to the so-called science of the city (Caprotti et al., 2017; UN-Habitat, 2016), which, fueled by the megadata that are constantly circulating within it, seeks to optimize mobility and energy efficiency, agriculture, crisis management and so on. However, this highly mediatized discourse fails to sum up the relationship between digital technology and urban climate action. On the contrary, since “everyday life is invented with a thousand ways of poaching” (de Certeau, 1990, p. xxxvii; own translation), other trajectories exist, as the literature reports. The AI solutions promoted under the dominant rhetoric so far tend to exacerbate the historical issues of urbanization—for example, in terms of socioeconomic disparities, exclusion of marginalized populations or environmental injustice (Bender et al., 2021; Green, 2019; Hatuka et al., 2020)—but urban climate action nonetheless remains driven by spontaneous adaptation strategies, new collaborative modalities or unusual alliances made possible by participation in megadata flows and citizen appropriation of AI (Guldi, 2016, 2018; Picon, 2015, 2016). The uses of digital technology go far beyond the dominant rhetoric and even spill over into projects that are critical of the growth imperative and of platform-companies’ abuses of power and tendency to hide or dissimulate their carbon footprint (Caccamo, 2019; Durand-Folco, 2017).

However, the critical uses of digital technologies are not immune to the paradoxical dynamics of contemporary urbanization, whereby collective action tends to turn into maladaptation^[3] despite any other intentions (Schipper, 2020). As the climate crisis pushes problem-solving to its limits, the paradoxes of digital technology are essentially no more than an amplified reflection of a much deeper phenomenon.

Photo: Madhis Mousavi (Unsplash)

3.1 The climate emergency and the paradox of the rebound effect

Recently, the rebound effect has been the subject of growing interdisciplinary attention, and has even appeared in the popular media (see e.g., Bédard, Les Affaires, March 15, 2018; Harvey, Le Devoir, November 7, 2015). In the smart city, it typically corresponds to the paradoxical phenomenon whereby a technology, public policy or circular economy strategy designed to reduce the consumption of an energy resource turns against itself, provoking behaviours or systemic dynamics that cancel out some or all of the expected environmental gain (Font Vivanco et al., 2018). Take, for example, the rising use of machine learning in a manufacturing context, which enables gains in energy efficiency and thus reductions in greenhouse gas (GHG) emissions. If these efficiency gains were to translate into increased production, as might be expected,  then the gains would be partially or totally offset. This means that GHG emissions could then be increased compared to the initial situation without AI. Without even taking account of the intrinsic carbon of AIs, we would then be faced with a direct rebound effect on the scale of a manufacturing company or an entire industrial sector.

When a policy is put in place to combat climate change, the potential gains are greater, but so is the scale of the rebound effect (Dauvergne, 2022). In terms of mobility, for example, the installation of a fleet of shared electric vehicles has the potential to improve energy efficiency and the carbon footprint. That said, the opposite could also hold true, for, if a critical mass of people were to substitute not their car but public transport with electric vehicles, GHG emissions would increase (Herrmann et al., 2014).

In addition, the modeling of rebound effects for the purposes of decision support or climate policy development poses major challenges. Typically, economic models are combined with environmental impact models such as life cycle assessment (Font Vivanco and van der Voet, 2014). With the advancement of complementary analytical methods (e.g., deep learning), hybridization continues, giving rise to models that are increasingly sophisticated (OECD, 2019) but also increasingly energy-intensive (Almeida et al., 2018; Wu et al., 2017). As a result, the causal chains to be understood are so extensive, uncertain and entangled that even the best models are unable to establish the necessary and sufficient conditions for avoiding a rebound (Zink and Geyer, 2017). It is as if the rise of digital environmental modelling has contributed to constantly broadening the scope of the phenomenon being modelled, thereby making it ever more elusive.

Whatever its scale, the rebound effect presents itself as a paradoxical dynamic that should not automatically be assimilated to the soluble-problem model. When it comes to urban climate action, the paradoxical perspective at least gives us the ability to recognize that the digital remedy is also the poison.

3.2 Personal data security and the paradox of predatory emancipation

It is against the backdrop of the climate emergency, alongside other issues pertaining to good urban governance (such as security or social equity), that sensors are multiplying in common-use urban spaces, both public and private. But while the smart-card type of usage associated with smartphones remains under the partial control of its users, the networks of sensors (cameras, pollution or noise sensors, etc.) weaving their way invisibly through the urban fabric mark the advent of more intrusive approaches. In the near future, for example, we can expect the Internet of Things to increase the number of sensors in urban areas as well as the number of interfaces that can exchange data more or less freely.

When it comes to the AI systems that enable the analysis of the data continuously emitted and captured by these networks of connected objects, the issues of social inequity and exclusion are numerous and abundantly documented. They are typically associated with the biases inscribed in the training data of learning algorithms, which are then introduced to AI-assisted decision-making processes. The phenomenon of predictive policing has drawn attention in recent years, with algorithms designed “to predict where and when crimes are likely to take place, but also and above all to guide, supervise and regulate police work” (Benbouzid, 2018, p. 223; own translation). However, so far they have tended to exacerbate inequities and violence, particularly toward racialized populations (Asaro, 2019; Niculescu-Dincă, 2018)—an issue that is not specific to predictive policing. All the already existing social biases are likely to be reinforced whenever an AI system assists in making a decision with a high social impact (Bender et al., 2021), for example, in the supposedly more objective allocation of healthcare services (Obermeyer et al., 2019). Moreover, in the name of good governance principles, such as safety, equity and inclusion, the multiplication of AIs could make it virtually impossible for individuals to travel without a digital trail, compromising not only privacy but also the very values of equity and inclusion this technology was intended to defend in the first place (Castets-Renard et al., 2020; CNIL, 2017).

It is in this sense that the urban uses of AI give rise to a paradox of predatory emancipation analogous to the paradox of the rebound effect. Whatever the intended purpose, the data that feed it always carry both an emancipatory potential and a risk of predatory practices. Here again, the paradoxical perspective does, if minimally, enable us to recognize that the digital remedy is also the poison.

Photo: Yeshi Kangrang (Unsplash)

In practice, it is not all that simple, as the raison d’être of the emerging 5G technology is precisely to improve the energy efficiency of flows, and therefore drastically increase the data rate in circulation for the same amount of energy (Wu et al., 2017). It is worth noting, however, that the environmental benefits associated with digital sobriety translate perfectly to the social level via the principle of data minimization (EU, 2018). In other words, by collecting only the personal data that is necessary for the purpose for which it is being processed, and according to the “informational norms” specific to a particular social context (Nissenbaum, 2018), we give ourselves the means to generate a virtuous circle both environmentally and socially.

^[1] Problem-solving is the “capacity [of an individual or collective] to consider the probable regarding the certain, the desperate regarding assurance, chance regarding conduct, and the harmful regarding perfection” (Vérin, 1984, p. 262; own translation).

^[2]  The speech given in November 2008 by Samuel J. Palmisano, then CEO of IBM Corporation, fully embodied the trend of the moment, while heralding the discourse that has now become dominant.

^[3] E. Lisa F. Schipper (2020), author and member of Working Group II (WG2) of the 6^th IPCC Assessment Report, defines maladaptation as follows: “Maladaptation is not just a waste of time and money; it is when an action results in conditions that are worse than those which the original strategies were trying to address. Maladaptation is therefore not being poorly adapted to climate change; it is actually a process whereby people become even more likely to be negatively affected by climate change.” (p. 409)

To cite this article

Arpin, M-L., Gambs, S. (2022). Digital technologies to combat climate change: Between problem and paradox. In Cities, Climate and Inequalities Collection. VRM – Villes Régions Monde. https://www.vrm.ca/digital-technologies-to-combat-climate-change-between-problem-and-paradox