The environmental impacts of Big Data initiatives draw attention to their material substrate and infrastructural dimensions. Such materiality is not value-free, instead it has political and social implications, as has been highlighted by scholars in the field of internet studies (Sandvig 2013), media and cultural studies (Coté 2014) and anthropology (Pink et al. 2016), who draw on social constructivist and post-human theories (like Actor-Network Theory) and human–computer interaction approaches (Leigh Star 1999). Going beyond the study of the symbolic and language-related aspects of the virtual world, these scholars turned towards an examination of the material character of the internet: they describe the material constitutions of networks (Fuller 2003; Manovich 2001), wireless connections (Mackenzie 2010) and internet infrastructures and critically explore their connections with epistemic practices or power structures (Sandvig 2013). This “material turn” in internet studies and Science and Technology Studies considers material and physical infrastructure as socially constructed, and societal relations as constructed by such infrastructures that benefit some and marginalise others. In this sense, technological choices are considered as cultural, social, political and ethical. Pioneer scholars in this field, like Susan Leigh Star (1999), have described and made apparent physical parts of the internet that would otherwise be invisible (wireless signals, buried wires, fibre optic lines)Footnote 10 in order to uncover the tacit, intangible labour and the “politics” that are involved in practices and routines around these systems. Such practices are not only related to the design of digital services and systems, but also to their use and governance (DeNardis and Musiani 2016). Internet and media studies have therefore moved from an understanding of virtual communities as disembodied identities to a focus on the continuous relationship between online and offline activity thus challenging the virtual versus real distinction (Pink et al. 2016). In many cases, challenging the relationship between the digital and the material and acknowledging the material character of digital infrastructures, content and context is a way to question the online/offline dichotomy.
More recently, a growing body of research in cultural and media studies has also provided insight into the situatedness and material substance of data centres (Burrington 2014; Hogan and Vonderau 2019; Holt and Vonderau 2015; Hu 2015; Taylor 2017a; Vonderau 2017) Besides highlighting the environmental implications of the materially heavy infrastructure that enables the data economy, these studies discuss its political and cultural dimensions. Through discourse analyses and ethnographic studies, this research has shown how the cloud is localised within specific geographical and historical contexts with some important geopolitical consequences. For example, in her anthropological inquiry into the relocation of Facebook servers in peripheral areas of Sweden, anthropologist Asta Vonderau (2018) shows how this move has contributed to redefining the local identity of the region, creating new geographical balances where the old national peripheries become new centres of the global cloud, which has triggered conflicts between communities. Media theorist Mél Hogan has described how Big Tech promotes a narrative of environmental concern proposing themselves as the most suitable industry to manage natural resources (Hogan 2018). At the same time, however, they also generate data and fast connection demand, thus encouraging consumption and sustaining their own economy: their sustainable discourses coexist with their neoliberal practices.
The environmental implications and political and cultural dimensions of the data revolution have strong normative implications that have so far been ignored by ethicists. This is surprising because not only philosophers and ethicists of technology, together with STS scholars, have reflected on the moral implications of technical design choices of digital systems in shaping human experiences, actions and identities (see for example Achterhuis 2001; Oosterlaken and van den Hoven 2012; Verbeek 2005; Winner 1999), also, the ethical reflection on implications of Big Data has boomed in the last 10 years (Mittelstadt and Floridi 2015). Despite this relevant scholarly work, the topic of the environmental implications of Big Data and their material infrastructure has gone largely unremarked.
The paper now discusses three ways in which the materiality of data has ethically relevant implications and would therefore benefit from research in this field. Firstly, it explains the implicit normativity in the vocabulary that is currently used when discussing the governance of data initiatives, and how this normatively laden vocabulary influences understandings of responsible behaviour. Secondly, it explores the internal tensions between current data initiatives and environmental policies and points out the need for a thorough assessment of benefits and risks. Thirdly, it analyses issues of fair distribution in the context of decision-making on data storage practices. For each of these aspects it will be pointed out why these issues should matter to ethicists, how they fall within their interests and how ethicists could contribute to the debate. The last section of the paper further reflects on the role of the ethicist and some of the policy implications for these aspects.
The Ethics in the Metaphor: From Language to Actions
Media and lay discourses often suggest that the digital is somewhat immaterial. As pointed out by several commentators, using the term “cloud” to refer to computing and internet networks is a misleading metaphor (Holt and Vonderau 2015; Hu 2015; Taylor 2017a). It suggests something impalpable, fluffy, untouchable, light, and transparent. This language strategically obscures the materiality of the infrastructure as well as its geographical presence and environmental impact. Cloud computing is in fact a highly tangible and touchable assemblage of material and heavy stuff. The material substrate of the cloud is made of cables, wires, servers, and shelves in buildings in every corner of the globe. Similarly, the language used to refer to data as an “unlimited and superabundant resource” implicitly suggests that data are virtual goods, always present, an ever increasing and never-ending resource, in contrast to other resources (such as oil, water, or land) that are limited in quantity and need careful management. However, how data relies on limited resources in order to be stored and processed suggests that it is not superabundant.
The use of this type of language in relation to the storage of data induces false assumptions as it tends to hide what the “material turn” in internet and media studies has highlighted: that there is a continuity of offline and online life. The digital has material implications for the physical “real” world, but the language used obscures this. Places, infrastructures, and buildings all play a role in constituting the online world, and vice versa: online behaviour also has material implications in the physical “real” world. False assumptions, created by misleading language, influence people’s understanding of the data universe and economy, and shapes their attitudes towards it. For example, it suggests that the virtual environment is drastically different from the material, physical environment in which we live our offline lives, where we eat, sleep, watch movies, have sex, and throw away our garbage. However, digital behaviour interferes with the physical environment and not always in a positive way. Buying or selling bitcoins, storing large amounts of photos on cloud services, binge watching TV are all online activities, but their implications are not relegated solely to the digital environment as they also require material and limited resources (space, water, electricity, and fuel) and therefore have serious implications for the natural environment.
Metaphors and language suggesting otherwise are culpable of obscuring the physical character of data and its implications for the offline world. Data consumption is no less environmentally problematic than material goods consumption; and the paperless, computer and data intensive office is not an ultimate solution to environmental issues, but instead creates new ones [see also Tenner (1996)]. Highlighting the material consequences of online behaviours and their impacts on the environment is crucial in order to draw attention to individual and institutional responsibilities in terms of the common good. This is even more relevant in societies where we acknowledge the importance of protecting the environment by incentivising “green” activities aimed at reducing individual and collective carbon footprints. Production and storage of data can be environmentally problematic and, as such, the sustainability of digitally-rich activities and behaviours should be part of the conversation on Big Data.
Besides hiding responsibilities, the language referring to data as an “unlimited and superabundant resource”, which only needs to be collected and managed in order to be used in a meaningful and socially beneficial way, implicitly calls for incentivising policies and activities that promote data initiatives. The risk is that the logic of such language lends governors and policy makers to think that by promoting numerous data initiatives they are always promoting the public good. Again, however, the environmental costs of these implications should be considered when assessing how, or whether, these initiatives foster the public good.
According to a pragmatist view on ethics (Keulartz et al. 2002; Lucivero 2016), ensuring that the normative assumptions in industry and policy discourses questioned are disclosed, falls into the remits of ethics. Vocabularies have a moral character as they justify some actions and forbid others as well as contributing to knowledge and distributing responsibilities. Reconsidering the validity of such metaphors and using a vocabulary that highlights the material and physical character of data storage could help in enhancing awareness of the consequences of some actions and allow space for discussing individual and institutional responsibilities towards data handling, storing and production. Ethicists can have a role in facilitating such discussion as a first step in questioning the presumed benefits of data initiatives.
The Ethics in the Visions: Evaluating Data Initiatives
The second morally relevant consideration behind the material nature and environmental impact of Big Data is that such impacts should be considered when assessing the overall value of these initiatives; this is important in relation to the societal value of sustainability that, according to UN and industry visions, Big Data can promote. With increasing understandings of the relationship between human intervention and the natural environment, public health authorities and governments have placed much emphasis on the need to reduce anthropogenic changes to the environment that can have negative impacts on human health. Take for example the UN 2030 Agenda for Sustainable Development that was adopted in 2015 by 193 countries (United Nations 2015a).Footnote 11 Nine of the 17 goals explicitly refer to the environment and sustainability as a value: not only as goals clearly referring to policies actively promoting environmental goodsFootnote 12 but also as goals linking innovation and economy with the overarching aim of sustainability. These include, for example, ensuring access to affordable, reliable, sustainable and modern energy (goal 7) the development of clean industry and environmentally sound technologies and industrial processes (goal 9) or the sustainable management and use of material resources (12.2); the reduction of waste generation through prevention, reduction, recycling and reuse (12.5) and the encouragement of companies to adopt sustainable practices and integrate sustainability information into their reporting cycles (12/6); the importance of raising awareness (12.8), and the importance of taxing inefficient and unsustainable practices (12.c). Sustainable development is strictly linked to the improvement (or non-deterioration) of the environmental conditions in which people live and to the protection of the ecosystem.
Goals and values, such as sustainability, therefore, clearly permeate the visions of policy makers and broader society and are used to justify policy decisions. They are guiding normative visions for policy makers and broader communities as they propose more desirable futures (Grin and Grunwald 2000; Jasanoff and Kim 2013). Policy visions of Big Data for sustainable development (or health) can be considered as sociotechnical imaginaries with a performative role. In the world of the EU, data are “an essential resource, for economic growth, competitiveness, innovation, job creation and societal progress in general”. Data initiatives are promoted within the context of modern economies and are expected to have great benefits for populations and individuals. The normative dimension of such visions emerges in the fact that they depict images of futures that are presented as desirable for societies and individuals (for example, healthier and more sustainable) and suggest ways to realise them. This normative dimension suggests that data are good, not only for individuals and private actors, but also for societies at large. However, data initiatives can also have negative environmental impacts. Therefore, on closer examination, the guiding values of data economy policies seem to be at odds with the guiding values of sustainable development predicated in the same policies.
Visions of Big Data for sustainable development (fighting, for example climate change) collide with the reality of Big Data, which can work against sustainable living. The two visions are not necessarily incompatible and can be considered as two sides of the data revolution coin: as we have seen in the previous section, ICT and sustainability have an ambiguous relationship and yet there are ways for ICT to contribute to sustainability. To further complicate the matter, sustainability itself is a “contested value” as it is conceptualised in different and sometimes conflicting ways to normative underpinnings (Van de Poel 2017). For these reasons, one cannot expect complete consistency in policy visions of Big Data for sustainable development. At the same time, it is important to make policy makers, funders and institutions aware of the ambiguity of the relationship between sustainability and Big Data in order to nuance expectations and initiate a dialogue about underlying values, priorities and trade-offs. Ethicists can contribute to more responsible public policy by raising awareness of these ambiguities, fostering a weighing of benefits and drawbacks of data initiatives, uncovering hidden value conflicts and making governing bodies and public institutions more critical of the rhetoric of data optimism strategically advanced by Big Tech. For example, Big Data projects could be made accountable for assessing their environmental impacts, in the same way that they are required to account for the way they tackle issues of privacy and security. Documents such as the United Nations report on Big Data for Sustainable Development discussed above, do not list environmental issues amongst the challenges of Big Data initiatives as they focus primarily on issues of privacy, security and fair representation and access (United Nations (IEAG) Independent Expert Advisory Group on a Data Revolution for Sustainable Development 2014).
As explained in “The Weight of the Cloud and the Environment” section, environmental harm is still uncertain and difficult to measure. This uncertainty suggests that current measures of environmental impact assessment may fall short when asked to provide a definitive answer or a clear-cut weighing of benefits and drawbacks. This complexity demonstrates the need for an ethical and philosophical reflection on impacts that takes account, not only quantitative aspects, but also of qualitative, behavioural and moral dimensions.
Understanding how data initiatives mobilise different values and discussing how to balance them is therefore crucial to a sound ethical discussion. Ethicists can also help in the assessment of the broader benefits and drawbacks of data initiatives for health. Such assessments are not a one-off, preliminary, “measuring” exercise (wherein impacts are quantified and weighted before specific data initiatives are given a green light) but require a continuous reflection that works in parallel with data initiatives and which also focuses on qualitative impacts.
The Ethics in Decision-Making: Allocating Resources and Changing Landscapes
A third ethically problematic aspect related to the material nature of data relates to the criteria that drive decision-making processes around allocation of data-relevant resources. Data storage and use depend on the availability of land and resources such as water and fuel. Since these resources are scarce, we cannot exploit them limitlessly. Technical and logistical solutions to use less of these resources are a way of dealing with this problem: these solutions include, for example, the establishment of data centres in Nordic countries where less energy is needed to cool down servers.Footnote 13 However, these solutions may create new problems: for example, the geographical location of data centres raises questions of legitimacy and autonomy concerning decisions over the landscape and use of the natural environment. Data infrastructures not only alter the physical space, but also the socio-economic organisation within that space and beyond. Data centres, for example, are currently modifying natural and urban landscapes with their distinctive architectures (Burrington 2014; Taylor 2018). Vonderau shows that in placing data centres in the peripheric North, aspirations of local change face the reality of limited jobs offered to local communities in a highly automated industry (Vonderau 2018). These geopolitical revolutions in specific areas raise questions of how these investments benefit local residents, how the new arrangements change social and economic configurations both at the local and the national level and who gains and who loses in these changing interactions. For example, in the case of digital device disposal discussed above, the location of disposal facilities and informal recycling centres in low- and middle-income countries raises questions concerning the fair distribution of benefits and damages of the digital revolution, where more vulnerable populations suffer the environmental consequences of rich countries’ increasing digital demand. Furthermore, it is legitimate to question who should be involved in such decisions regarding location and management of such infrastructures.
Proximity to data centres and geographical distribution of data centres raise questions of fairness in the context of a growing data economy. For example, it has been highlighted that the proximity or distance from relevant infrastructures determines conditions of access to services. In 2013, architecture reporter Andrew Blum, published the book Tubes: Behind the Scenes at the Internet where he describes 2 years of research in which he attempted to better understand the materiality of the internet by visiting the real places on the map where the internet is built, provided, connected and distributed (Blum 2013). In describing the physical details, buildings, people and infrastructures that make the network a real and tangible thing, capable of modifying geographies and the landscape, Blum challenges the metaphors of the digital as a space beyond place. Not only does he show that the internet is a place but also that data is located in some places more than others. The proximity to data servers allows information to travel at a faster speed, which is quite relevant for certain businesses. The fact that the old telecom company Verizon has recently sold a building in central Manhattan to a real estate agent for data companies, highlights the increasing need for data companies to locate in urban areas, such as central Manhattan, in order to allow stock market companies to work at the speeds necessary for their business (Carp 2013). As the physical location of infrastructures determines speed of access to services, decisions concerning the location of centres cannot ever be neutral, a matter which has societal implications. So, if proximity to data infrastructures has competitive implications for the fast accessing or processing of data, who will decide the activities that require more (and more proximity to) data storage and processing facilities? Furthermore, what criteria would drive decisions on whether priority should be given to healthcare databases, stock market programmes, or social media and consumer storing servers? The location of data centres will also have implications for who has faster access. This is increasingly important in the context of a data economy where access to data means access to wealth and economic development, as this implies that the choice of place will also have an impact on local economies. Given that data centres are increasingly being moved to cold countries, as warmer climates are not suitable for their placement, how can the growth of economic gaps between warm countries and cold countries be avoided? Leaving these decisions in the hands of the commercial sector could be problematic as issues concerning the common good and justice may be overlooked in favour of the economic gain of a few private actors.
Another set of questions related to fairness in deliberations in contexts of scarcity concerns the criteria to decide what can be stored and what needs to be deleted. In a context where digital material occupies physical spaceFootnote 14 and cannot be endlessly kept, decisions need to be made with respect to what information or knowledge should be kept or deleted. Interestingly, Jonathan Mirand and colleagues have explored these issues in the project “The Archive Documentary”Footnote 15 where they highlight how, as digital space needs to be decluttered, some digital production will inevitably need to be erased to make space for more. Their project aims at creating an archive that saves artistic and cultural productions as often these are considered less appealing for industry and, as such, they become lost. More generally, this artistic project reminds us that when decisions concerning the placement of data centres or the storing of data are involved, criteria that justify the legitimacy of these decisions need to be explored. Who decides what is important and what the criteria are for determining the value of data?
These sets of questions show that data initiatives do not build on superabundant resources and although data can ideally be produced in unlimited ways, the resources that are used for their storage and processing (e.g. water, physical space, electricity) are finite. In this context of scarcity, decisions of resource allocation need to be made. What is up for discussion is: how the data revolution distributes environmental benefits and losses among groups and communities, what type of inequalities it creates and who is responsible for them, whom should be involved in the decisions about distribution of scarce resources and environmental outcomes, what criteria will need to be considered (e.g. not only economic and efficiency-based criteria, but also those based on fairness) and what values should guide these choices (e.g. economic utility, cultural value). These are questions of social justice and as such would be enriched by an ethical analysis. Political theorists and environmental activists have been raising similar questions under the banner of “environmental justice” (Bullard 2005; Schlosberg 2009, 2013). In these discussions, the distributive issues around environmental benefits and disbenefits have been explored, but also the structural reasons for such imbalances as well as issues of participatory parity. As the digital revolution creates new inequalities that have motivated the need for addressing issues of “data justice” (Dencik et al. 2019; Taylor 2017b), so new issues of environmental justice arise, as not everyone “enjoys the same degree of protection from the environmental and health hazards” (US EPA, n.d.). Moreover, not everyone has equal access to the decision-making process involving data infrastructure. It is at this intersection between data justice and environmental justice—where the populations that currently suffer more from the environmental implications of ICT are also the ones that are at the present time less likely to benefit from the digital revolution—that more work needs to be done.