Mikkel Stein Knudsen & Jari Kaivo-oja:
Recently, researchers have started to investigate the potential positive/detrimental role Artificial Intelligence (AI) might play in achieving the global Sustainable Development Goals (SDGs). This is a natural squaring of two prominent twin transition megatrends likely to shape global developments over the coming decade. We provide a snapshot of the recent literature.
The advances of AI have been trumped as vast and considerate in recent years, and it is now hard to find images of the future, which does not position significant roles for AI. SITRA’s analysis of megatrends highlights that AI applications permeate society. AI technologies feature prominently in the Finnish 100 Opportunities for Finland and the World and the European 100 Radical Innovation Breakthroughs for the future. The new, first annual Strategic Foresight Report of the EU (European Commission, September 2020) underlines that “Digital technologies and related business models, including Artificial Intelligence (AI) and the platform economy, will impact the job market”.
Several studies and surveys suggest that artificial intelligence platforms and apps help humanity towards sustainable development goals (SDGs). However, this optimistic assumption about this future trend is by no means to say that people should not use their brains and intelligence. It is important to identify the issues relevant to decision-making in which AI can support more sustainable development processes and choices with high-level impacts.
AI needs to be complemented by other instruments
Let´s take an example of human intelligence and AI. The Tinbergen Rule, the rule that was named after one of the first two Nobel laureates in economics in 1969, is a basic principle of effective policy. Distinguishing between policy targets, on the one hand, and policy instruments, on the other hand, Jan Tinbergen (in 1952) argued that to successfully achieve 𝑛𝑛 independent policy targets, at least the same number of independent policy instruments are required. The Tinbergen Rule distinguishes three types of variables: (1) Data, (2) Target Variables, and (3) Instruments. This rule has become known as the Tinbergen Rule. The Tinbergen Rule is very relevant in planning and building transition scenarios to reach SDGs. This means that tailoring policy instruments in the backcasting scenarios of SDGs should consider the Tinbergen Rule. All potential AI applications, which do not take the Tinbergen Rule seriously, are probably creating more policy failures than policy successes.
We can conclude that human intelligence and AI are complementary system planning issues – just mentioning the Tinbergen Rule as an example.
AI and SDGs
The Sustainable Development Goals (SDGs) consist of 17 globally agreed goals with 169 explicit targets guiding the national and global efforts of sustainable development. It is important to understand the extent to which the different goals and targets are linked with each other (Mainali et al., 2018), and to keep in mind that there might be both synergies and trade-offs when combining the individual SDGs. An interdisciplinary study from Nature Sustainability (Nerini et al., 2019) e.g. shows that efforts to combat climate change can reinforce all 17 SDGs, but also undermine efforts to achieve 12.
Since the links between SDGs themselves are complex and involves synergies, it is no surprise that the role of Artificial Intelligence is also complex. The role of artificial intelligence in achieving the Sustainable Development Goals by Vinuesa et al. (2020), provides the hitherto strongest overview of the links between AI and SDGs. With the use of a consensus-based expert elicitation process, the researchers conclude that AI may enable the accomplishment of 134 targets, but also undermine the accomplishment of 59 targets. Figure 1 illustrates this across the 17 different SDGs.
Fig 1. Source: Vinuesa et al., 2020
The analysis also shows that more academic publications are demonstrating the positive enabling potential of Artificial Intelligence (in SoMe-terms, #aiforgood) compared to those demonstrating the risk of inhibiting the SDGs. This might be explained by reporting bias (researchers are more likely to research and publish about potential new opportunities than about potential pitfalls) or by the fact that the discovery of detrimental effects might require a long-term study, wherefore we may only get to read academic research on this year from now.
While the article in Nature Communications provides numerous examples of how AI might be detrimental, another new article, just published online September 3rd, explicitly sets out to expose the political economy of environmental costs of Artificial Intelligence. International Relations Professor Peter Dauvergne takes on Artificial Intelligence as much more critical terms, as can be seen in Figure 2 below which cites some of the main conclusions of the paper. AI reinforces the status quo, enriches corporate billionaires and transnational companies, accelerates the extraction of minerals and fossil fuels, turbocharges consumerism, and results in an ecological displacement.
Fig. 2. Based on and featuring quotes from Dauvergne, 2020.
While Dauvergne is perhaps the most focused on highlighting the dark sides of the AI-evolution, numerous other recent papers also present the duality of AI as an enabler/inhibitor of sustainable progress. Noteworthy examples include Di Vaio et al., 2020, Truby, 2020, Goralski & Tan, 2020, Sharma et al., 2020, and Mohamed et al, 2020.
The duality of AI-progress
The dual imaginary of progress and challenge is reminiscent of the EU data privacy regimes analysed by our colleague at the Finland Futures Research Centre, Matti Minkkinen, in his brilliant doctoral dissertation published earlier this year. Here he shows three key imaginaries in the European data privacy debate: (i.) ‘Continuous growth’, (ii.) ‘tragic loss’, and (iii.) ‘Europe as a hero’. This Janus-faced duality of positives and negatives coupled with a discourse of a potentially unique European third way is also easily detectable in the European AI-discourse. The opening paragraphs of the EU White Paper on Artificial intelligence follow this exact trilogy of imaginaries, cf. figure 3 below.
Fig. 3. White paper on Artificial Intelligence (EU, 2020), annotated with data imaginaries (Minkkinen, 2020)
EU also positions AI and digitalization as fundamental pillars of the so-called twin transitions, namely simultaneous the green and digital transformations of society and businesses. AI and other enabling technologies are essential for building a New Green Deal, e.g. the bold and ambitious idea of creating a new Digital Twin of the Earth, Destination Earth.
However, the EU-approach reflects the frame that while AI can deliver great potentials, left unchecked it is more likely to harm. We need new initiatives and new regulations to steer the development and the deployment of technology in the right direction. This same frame exists in the majority of literature on AI and SDGs.
Truby (2020) provides a prime example:
- “Promising and feasible possibilities of AI-driven developmental progress are being overshadowed by the current unfettered experimentation with untested AI technologies in markets and societies” (…)
- “The case will be made that the design of AI software would benefit from pre-emptive regulation based on international principles, and secondly that such principles include a sustainable development purpose.”
We try to illustrate a way forward in Figure 4 below. At the same time, we must make use of the new technological opportunities helping the world on the path to sustainable development, and make sure that sustainable principles also permeate the development and deployment of Artificial Intelligence.
Fig. 4.: Adapted from Knudsen & Kaivo-oja, 2018.
A key element in this is to limit the important mismatch highlighted across numerous articles, not least in the contribution of Vinueasa et al. (2020): “research suggests that AI applications are currently biased towards SDG issues that are mainly relevant to those nations where most AI researchers live and work”. If AI technologies are designed and developed for richer and more technologically advanced nations, they “have the potential to exacerbate problems in less wealthy nations” (Ibid.).
In short, AI for the global Sustainable Development Goals requires establishing a global AI ecosystem.
How do we progress from here?
Based on the snapshot of recent literature, we list a few key priorities for the future development of sustainable AI.
- Transformation to an environment-friendly ICT-sector (reduced energy use, use of renewable energy, sustainable mining of raw materials, reduced e-waste)
- AI for the global, greater good; inclusive approaches beyond AI as solutions for the chosen few.
- Tackling algorithmic bias and algorithmic coloniality (decentralized AI, algorithmic transparency, codified ethics, certifications (?), and regulatory oversight).
What we need is an approach, which stands on the shoulders of the remarkable work in recent years on ethical and Principled Artificial Intelligence (Fjeld et al., 2020; LaGrandeur, 2020), and combines this with a stronger planetary focus. This will be a key challenge during the years to come.
Mikkel Stein Knudsen
Project Researcher (M.Sc., Pol. Science), Finland Futures Research Centre, Turku School of Economics, University of Turku
Research Director, Finland Futures Research Centre, Turku School of Economics, University of Turku.
– – –
The project ‘Manufacturing 4.0’ has received funding from the Finnish Strategic Research Council [grant number 313395].
– – –
Dauvergne, Peter (2020). Is artificial intelligence greening global supply chains? Exposing the political economy of environmental costs. Review of International Political Economy. Ahead-of-print. DOI.
Di Vaio, Assunta, Palladino, Rosa, Hassan, Rohail & Escobar, Octavio (2020). Artificial Intelligence and business models in the sustainable development goals perspective: A systematic literature review. Journal of Business Research. 121: 283-314. DOI.
European Commission (2020). White Paper: On Artificial Intelligence – A European approach to excellence and trust. Brussels, 19.2.2020. COM(2020) 65 final. Web: https://ec.europa.eu/info/sites/info/files/commission-white-paper-artificial-intelligence-feb2020_en.pdf
Fjeld, Jessica, Achten, Nele, Hilligoss, Hannah, Nagy, Adam & Srikumar, Madhulika (2020). Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-based Approaches to Principles for AI. Berkman Klein Center for Internet & Society: https://dash.harvard.edu/bitstream/handle/1/42160420/HLS%20White%20Paper%20Final_v3.pdf?sequence=1&isAllowed=y
Goralski, Margaret A. & Tan, Tay Keong (2020). Artificial intelligence and sustainable development. The International Journal of Management Education. 18: 100330. DOI.
Knudsen, Mikkel & Kaivo-oja, Jari (2018). Bridging Industry 4.0 and Circular Economy: A New Research Agenda for Finland? Tulevaisuuden tutkimuskeskuksen blogi. Web: https://ffrc.wordpress.com/2018/09/12/bridging-industry-4-0-and-circular-economy/
LaGrandeur, Kevin (2020). How safe is our reliance on AI, and should we regulate it? AI and Ethics. DOI.
Mainali, Brijesh, Luukkanen, Jyrki, Silveira, Semida & Kaivo-oja, Jari (2018). Evaluating Synergies and Trade-Offs among Sustainable Development Goals (SDGs): Explorative Analyses of Development Paths in South Asia and Sub-Saharan Africa. Sustainability. 10(3): 815. DOI.
Minkkinen, Matti (2020). A Breathless Race for Breathing Space: Critical-analytical futures studies and the contested co-evolution of privacy imaginaries and institutions. Turun Yliopiston Julkaisuja – Annales Universitas Turkuensis. Web: https://www.utupub.fi/bitstream/handle/10024/149425/AnnalesE55Minkkinen.pdf?sequence=1
Mohamed, Shakir, Png, Marie-Therese & Isaac, William (2020). Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence. Philosophy & Technology. DOI.
Nerini, Francesco Fuso, Sovacool, Benjamin, Hughes, Nick, Cozzi, Laura, Cosgrave, Ellie, Howells, Mark, Tavoni, Massimo, Tomei, Julia, Zerriffi, Hisham & Milligan, Ben (2019). Connecting climate action with other Sustainable Development Goals. Nature Sustainability. 2: 674-680. DOI.
Sharma, Gagan Deep, Yadav, Anshita & Chopra, Ritika (2020). Artificial Intelligence and effective governance: A review, critique, and research agenda. Sustainable Futures. 2: 100004. DOI.
Tinbergen, Jan (1956). Economic Policy: Principles and Design. North-Holland. Retrieved from http://hdl.handle.net/1765/16740.
Truby, Jon (2020). Governing Artificial Intelligence to benefit the UN Sustainable Development Goals. Sustainable Development. 28(4): 946-959. DOI.
Vinuesa, Ricardo, Azizpour, Hossein, Leite, Iolanda, Balaam, Madeline, Dignum, Virginia, Domisch, Sami, Felländer, Anna, Langhans, Simone Daniela, Tegmark, Max & Nerini, Francesco Fuso (2020). The role of artificial intelligence in achieving the Sustainable Development Goals. Nature Communications. 11: 233. DOI.
– – –
Article picture: pixabay.com