Networks,narratives and neocoloniality of AI for Climate Action

Abstract

This article results from a cross-disciplinary study which unfolds the neocolonial nature of the application of AI technologies for climate action by examining the actors’ network and discourses around this global practice. The study demonstrates that tech businesses from the Global North are currently leading the AI for Climate Action project, with certain non-profits and civil society organisations helping these actors to advance their core profit-driven ambitions. They facilitate discourses, such as ‘leveraging AI’, ‘AI innovation’, and ‘responsible AI’ to build techno-solutionist narratives and concurrently legitimise their networked relationships and actions. Drawing from our exploratory findings, we bring the question of geopolitical power imbalance into the forefront of this AI for Climate Action discussion. We highlight the deeply problematic realities around this practice that includes the minimal participation of the Global South actors in accessing the critical digital infrastructures to build AI technologies and their lack of control over the strategic decisions around the applications of such technologies in their own climate contexts. We contend that AI for Climate Action, due to the Global North-centric corporate interests, further exacerbates climate injustice by marginalising the populations who are already disproportionately impacted by climate change.

This article is a part of special theme on Datafied Development. To see a full list of all articles in this special theme, please click here: https://journals.sagepub.com/page/bds/datafied_development?pbEditor=true

Introduction

AI technology developments heavily depend on data centres (van Es et al., 2023), requiring a staggering amount of energy to compute, and contribute to a huge carbon footprint (Crawford, 2021) by draining natural resources and exacerbating pollution. However, AI can also help in efficient climate prediction and mitigation; some researchers accuse the industry actors of avoiding environmental costs consideration while advertising AI as a ‘quick fix’ of climate problems (Brevini, 2020), but others go beyond such binary by suggesting that it is possible to realise the potential of AI to address climate problems by developing these technologies responsibly (Cowls et al., 2021). Additionally, AI Ethics researchers bring more complex questions of justice (Coeckelbergh, 2021) and sustainability (Van Wynsberghe, 2021), but we find these discussions rarely focus on Global South actors’ ownership in AI and implications of these technologies when applied in their climate contexts.

In this article, we contribute to AI and climate justice related conversations by illustrating the lack of participation of Global South actors in AI for Climate Action. We argue that AI for Climate Action is a global initiative dominated by diverse corporate and non-profits and show the concentration and interconnectedness of Global North actors in this enterprise. We highlight how the corporate-capitalist discourses that these actors collaboratively produce and propagate can harm the Southern stakeholders and unfold the neocolonial dynamics underpinning the network and narratives of the global AI for Climate Action practice. In this effort, we briefly communicate the results of our empirical study involving cross-disciplinary methods such as ‘issue mapping’ (Rogers, 2019) and critical discourse analysis (CDA) (Fairclough, 1995) and understand these findings as an instantiation of neocolonialism by drawing from the colonial implications of both AI (Couldry and Mejias, 2020; Mohamed et al., 2020) and climate problems (Sultana, 2022).

AI for Climate Action as a global north-centric network of practice

AI technologies used for climate action and the actors employing them are interdependent, forming a heterogeneous network through which they are dynamically assembled (see Latour, 1996). On this premise, we use the method of ‘issue mapping’ to analyse and visually represent the connections between the actors, organisations, or issues of contemporary global AI for Climate Action practice (method details are in Appendix). This method assists digital researchers to identify relationships between actors, and how these actors can align or diverge on particular debates or controversies (Marres, 2015). Using this method, we find a hyper-linked URL network across the web through automated ‘dynamic link sampling’ which represent the actors involved in AI for Climate Action who share common web-pages, research papers, documents, events, databases, and audio-visuals that eventually bind this network.

The Issue Crawler application in our study discovers an array of websites which reveal three categories of key actors in the network of AI for Climate Action practice (Figure 1). These include (a) corporate organisations (such as PricewaterhouseCoopers), (b) non-profits (such as Global Partnership on AI) and (c) global bodies (such as World Economic Forum). The mapping (Figure 2) highlights incoming and outgoing links among the actors within the network, illustrating how they interact and connect with each other despite being different types of organisations.

Figure 1.

Three key categories of organizations doing AI for Climate Action from the list of weblinks generated through Issue Crawler, May 2022).

Figure 2.

Corporate companies (Boston Consulting Group and Tomorrow AI) doing AI for Climate Action are shown connecting with World Economic Forum (Generated through Issue Crawler, May 2022).

Notably, none of the actors discovered from the issue mapping exercise, except Africa AI4D, is geographically located in the Global South. This absence is striking since climate vulnerability is bound up with the realities of many countries of the Global South. These countries currently have far less capability to adapt to climate change than those with high incomes (Levy and Patz, 2018). Our findings imply that Global South actors have weaker access to infrastructure and resources to apply AI technologies in their own climate contexts. As a result, they have less control over the trajectory of this practice and policy or regulatory decision making.

Our study also reveals evidence on the growing Global North-centric corporate interests in AI for Climate Action. With non-participation from the Global South stakeholders, this creates opportunities for these corporations to deploy AI applications by targeting Global South countries as climate prone areas with little market barriers. The Western AI industry evidently prioritises profit-making over people's stakes (Zuboff, 2019), Nost and Colven (2022) show that some of these actors are already facilitating surveillance, greenwashing, and commodification in disguise of providing climate solutions. We argue such corporate capture of AI for Climate Action is deeply problematic for the already vulnerable Global South populations, since deployment of AI is further exploitative and can reproduce colonial harms and inequities for them (see Birhane, 2020).

AI's political ecology researchers increasingly focus on restoring control of land, resources, energy, and technology for those who reside on the periphery of such extractivist Global AI practices (Taffel et al., 2019). Building on, we draw parallel between AI and global climate politics. Global South countries produce the least greenhouse gases yet are more adversely affected by climate change than high-income countries with industries and big corporations (Levy and Patz, 2018). The control of AI-based climate solutions by Global North corporations will exacerbate such climate injustice, further marginalising the populations already disproportionately impacted by climate change as naturalised ‘consumers’ of AI technologies.

We have discovered that some non-profits and global bodies (for example, World Economic Forum) play facilitating roles in the network of AI for Climate Action by supporting corporate actors to advance their core profit-driven ambitions. By closely studying the web pages and links, we find that these actors legitimise the AI applications in climate context by producing research, slogans or documents, which corporations use and share to manifest consent regarding their actions. Our mapping represents web-based interactions between these actors and corporations (Figure 2), which instrumentalise transnational and interregional interactive communications at the core of this network of practice. These interactions buttress the worldwide flow of narratives to ‘globalise the globalised’ (Fairclough, 1995). Many of these enabler organisations are not directly involved in AI or commercial activities, but their financing comes from the donations of corporations and Western governments. They keep the capitalist values and interest of AI for Climate Action alive through discursively shaped narratives, which we highlight in the next section.

The capitalist logics of AI for Climate Action narratives

Through issue mapping exercise, we discover that two key documents were recurrently shared within the network, which implies that corporations and their enablers are purposefully promoting these resources: (i) ‘Climate AI: How artificial intelligence can power your climate action strategy (2020)’, developed Capgemini Research Institute — a corporate research organisation; and (ii) ‘CLIMATE CHANGE AND AI: Recommendations for Government Action (2021)’, developed by the Global Partnership on AI (GPAI) in collaboration with other agencies (all located in the Global North). We take these two most frequently shared documents within the network of AI for Climate Action as critical objects in our study, since analysing these reveals the underlying interests of the actors that are driving this global practice.

A close reading draws attention to the high frequency keywords used in these documents, which include the terms ‘leverage’, ‘innovation’ and ‘responsible’. Our analysis (See Figure 3) shows that both documents, although one endorsed by a corporate and another by the enabler organisations, recurrently use the word ‘innovation’ while referring to their actions. Capgemini's document uses ‘leverage’ and GPAI's document habitually uses the word ‘responsible’. Furthermore, we utilise the CDA approach to examine the hidden power relations and capitalist ideologies implicit in these discourses (see Fairclough, 1995) that shape the overall narratives of this practice. Three layers of analysis — the actual text, the discursive practices, and the larger context that affects both the text and the discursive practices — are all combined into one and their relationships are determined by CDA (Fairclough, 1995). We understand how these discourses reinforce structures of power in this global practice (cf. Foucault, 1969) and the capitalist interests of the actors involved. Our analysis helps us find three dominant narratives in AI for Climate Action: ‘leveraging AI’, ‘AI innovation’, and ‘responsible AI’.

Figure 3.

Brief insights from critical discourse analysis part of the study.

Capgemini's report states that high-performing organisations can guide us on aligning and leveraging AI to achieve climate goals. The term ‘leverage’ in the context of Climate AI warrants scrutiny. It comes from the word ‘lever’, a physics term for a mechanism that trades force for movement, and in finance, it refers to borrowing money to make purchases with the expectation that future earnings will exceed the borrowing costs. In financial terms, ‘resource leverage’ is ‘getting the most from the least’ and the desired results are frequently much bigger than the resources used to achieve them (Hamel and Prahalad, 2013). Tech companies seek to ‘stretch’ and maximise their goals and innovate AI when used in the context of climate change, enabling them to produce profits with minimal resources (for political economy of AI, see: Luitse and Denkena, 2021). ‘Leveraging AI’ as a form of ‘smart power’— an interplay of both ‘soft’ and ‘hard’ power. The soft power is the narrative that this leverage is advantageous for supposedly all countries, including those in the Global South without any concrete evidence. The hard power is the potential application of AI technologies by the Global North companies in the Global South without resistance or contextualisation.

Capitalist organisations use innovation to pursue profit, with investment in innovation crucial for business cycles (Courvisanos, 2012). Schumpeter (1939) argued that capitalism reduces innovation to routine, handled by specialists to ensure predictable outcomes. Innovation shifts cost curves, dominates capitalist life, and drives competition. Capitalist lobbyists often promote deregulation to foster capital expansion, frequently under the guise of innovation. We note that GPAI report (2021) suggests to, ‘deploy AI-for-climate innovation support in a manner that aligns the incentives of innovators and market incumbents’ (p. 10). This raises questions about who these ‘innovators’ are, and whether these innovations will be locally led and owned or merely serving techno-solutionist practices (Morozov, 2013). We cast doubt on ‘AI innovation’ drawing from how the European Union has come under fire for over-emphasising ‘innovation’ as the answer to Europe's economic and social problems (de Saille, 2015). Fougère and Harding (2012) argue ‘the understanding of innovation articulated in the Oslo Manuals has become so dominant that it is close to such a hegemonic status’ (p. 17). We maintain that current uncritical and dubious pro-innovation discourse in AI for Climate Action needs to be scrutinised, particularly when this innovation is taking place in Global South contexts.

Our findings show that GPAI's report mentions AI should be ‘responsible’ on every other page, whereas Capgemini's document lacks this emphasis. Responsible AI is not a core value or activity of most big tech. At best, it operates as a practice of ‘ethics-washing’ (Bietti, 2020). The ‘responsible AI’ discourse also obscures how power operates through AI systems and that AI fairness cannot be achieved in societies that are fundamentally unjust (Weinberg, 2022). For the notion of ‘responsible AI’ to have a meaningful critical purchase, it needs to be specified in terms of what it means in the context of the Global South and climate action, because it is now largely used in Western contexts to justify the usage of AI by tech companies. For the Global South actors, normalising these capitalist and ethics-washing narratives is akin to participation in systems of domination and inequality that will further strengthen asymmetries of power.

AI for Climate Action as a neocolonial practice

Our findings unfold the neocolonial dynamics of AI for climate action showing the dependence and control over Global South through economic and political means by the powerful actors in the Global North that limit local development and autonomy (Nkrumah, 1965). First, Global North-centric corporate interests obstruct Southern actors’ ability to develop and participate in local AI innovation for climate action, and second, they are not well-positioned to effectively resist the techno-solutionist and hegemonic narratives due to their lack of wealth, resource, and institutional capacity. It is likely that they will replicate the Western idea of ‘AI innovations’ by being ‘imitators’ or ‘adopters’ (Rogers et al., 2014). Such imitations are sometimes identified as ‘technology transfer’, but this transfer can be precarious and exploitative (Bozeman, 2000). We contend that neocolonial traits of Global North tech companies, in the (dis)guise of climate action, will result in repercussions of colonialism in the present day — what scholars refer to as the continuing force of coloniality through Western domination (Quintero, 2012). AI is deeply implicated in heteropatriarchy, racial capitalism, white supremacy, and colonialism and these mechanisms help the AI industry to sustain its power through a networked and dispersed global order (Tacheva and Ramasubramanian, 2023). Such neocolonial practices have further unintended consequences when applied in climate contexts. For example, AI Now researchers, drawing from previous examples of mass surveillance, warn that AI could significantly violate human rights if applied on climate refugees (Dobbe and Whittaker, 2019).

The neocolonial implications AI for Climate Action need to be understood differently from the coloniality of AI. Scholars increasingly argue that the climate problem is a colonial problem (Sultana, 2022), and stress that how today's environmental disasters affect people in the Global South should be understood in relation to British and European colonisation (Barber, 2021). By attempting to tackle climate change, which is largely their creation, and additionally entrenching AI's own coloniality, the Global North players enable double-layered harms in the Global South. In this article, we reveal the domination of Global North actors in AI for Climate Action and the necessity of resisting their exploitative narratives. We emphasise on avoiding such situations where Global South stakeholders are merely recipients and casualties of AI impacts but misidentified as ‘beneficiaries’ of Western-designed climate solutions.