15 reasons to govern AI with 17 SDGs
Updated on 02 November 2023
As AI continues to shape our reality, there are many calls to establish guardrails to guide future AI developments. Hundreds of AI ethical codes are outlined. New AI laws are proposed worldwide. Amidst all these initiatives, we are missing potential AI guardrails that are in plain sight: sustainable development goals (SDGs). The SDGs are the most comprehensive coverage of economic, human, and environmental development codified in 17 SDGs elaborated via 169 targets. They cover – among others – climate, health, gender, and inequality, from general principles to specific monitoring indicators.
So far, AI is mainly considered a technical way to realise the SDGs through the use of new applications and tools. Here’s why we should use SDGs as guardrails for AI governance.
Why are SDGs appropriate AI guardrails?
1. The SDGs are ready to be used. In AI governance, timing is crucial, as it was also stated clearly in the call for a 6-month pause in significant AI development signed by thousands of scientists and tech leaders. Other various calls for immediate action that describe AI as posing an ‘existential threat’ to humanity, highlighted this urgency for action. Any new development of AI rules and guardrails will involve lengthy and complex negotiation processes. Yet, SDGs can be deployed immediately as AI guardrails.
2. The SDGs have global legitimacy. At a time when it is challenging to have almost any global agreement, SDGs could be the only solution for AI governance with global buy-in. They have already been agreed upon by states in a legitimate setting. Through this legitimacy, SDGs can help build trust, which will be critical for future AI developments.
3. The SDGs are comprehensive in their policy coverage. They can address AI’s wide range of impacts on our core values, from respect for human lives and human dignity to the realisation of human potential via education and work. The SDGs are codifications of basic ethics and can replace hundreds of AI ethical codes that will take valuable time to create and agree upon.
The interactive illustration below shows how values at the heart of the SDG framework, such as affordability, inclusiveness, human-centeredness, and accountability (a total of 11 values), interconnect with AI.
4. The SDGs are operational and specific. SDGs can be operationalised in guiding AI developments through the matrix of goals, targets, and indicators. SDGs are an effective bridge between general principles and operational practices.
5. The SDGs are interdisciplinary. The main challenge of AI governance is connecting policy silos. The SDGs connect most policy fields, such as human rights, economic development, cultural values, and more. Metaphorically speaking, SDGs can provide guardrails not only for ‘one road’ – be it ethical or technological – but for the road network (see the interactive illustration below)
6. The SDGs are pro-development. There are major concerns that AI will trigger new and deeper digital divides for countries that do not have the means to participate in the AI race, which means most countries except the USA, China, EU, and a few other players. By applying the SDGs, we can address the risk of AI-driven divides within societies and globally.
7. The SDGs are inclusive. Most SDG targets for achieving inclusion could be applied to AI developments. Since inclusion is at the core of the SDGs in the societal development of youth, women, and other marginalised groups.
8. The SDGs are diverse. Currently, AI is shaped by available data that comes mainly from developed countries and the Global North. In contrast, the SDGs were developed considering various cultural, religious, and local practices. The SDGs can guide AI development where it lacks the most – in diversity.
9. The SDGs are local. AI tends to have a very global and generic approach. Take ChatGPT that provides answers without taking into consideration local cultural contexts. Making AI more local is a major challenge, given the lack of local data and AI capacities in countries and communities worldwide. SDGs can help overcome this challenge like no other as they are anchored in the problems and concerns of citizens and local communities.
10. The SDGs are not centralised and imposing. Currently, a few companies are monopolising main AI developments and the AI market. The SDGs provide policy space for avoiding the centralisation and monopolisation of AI developments.
How can we deploy SGDs as AI guardrails?
11. AI itself can be used to develop AI guardrails. A rule-based AI model can be developed, starting with SDG targets and indicators. It can be enriched by text-based machine learning built around considerable texts and data gathered via SDGs reporting over the last eight years. Such an AI system could be developed transparently as an open-source project, enabling countries, citizens, and companies to monitor its development and deployment. Once in place, the AI model/system would be used as an ‘evaluation’ tool to determine whether specific AI solutions contribute to or go against particular SDGs.
12. The SDGs should be complementary to other AI laws, standards, and declarations, including the OECD and UNESCO principles, the EU AI Act, Chinese AI regulations, and the Council of Europe’s planned convention on AI and human rights.
13. The environment, sustainability, and governance (ESG) principles are one channel to link the SDGs to the tech industry. Given the complementarity of SDGs and ESGs, the ESGs can be used to ‘translate’ SDGs to professional languages and cognitive framing used by businesses worldwide. By developing AI platforms, companies would also monitor their adherence to ESDGs.
14. The SDGs should be part of the evaluation of investments in AI, as this technology attracts a lot of capital. Investors should evaluate AI companies and projects based on their alignment with the SDGs.
15. The SDGs as AI guardrails should be communicated effectively. The prevalent view is that SDGs are some ‘do gooder’ approach with limited impact on ‘real life’. However, they are very realistic and practical governance tools. Thus, The UN, governments worldwide, and tech communities must put more effort into communicating the narrative of the SDGs as AI guardrails.