Illustration by Undraw

UNDP, the SDGs and the Role of Innovation Involving Machine Learning

The United Nations Development Programme, Sustainable Development Goals and applications within the field of Artificial Intelligence

Alex Moltzau
15 min readJul 11, 2020


I have long wanted to consider United Nations Development Programme and how machine learning can be used within the organisation. This brought me to writing an article about artificial intelligence (AI) and UNDP.

To me the United Nations Development Programme is one of the most inspiring networks out there.

It advocates for change and connects countries to knowledge, experience and resources to help people build a better life for themselves.

As an organisation it operates in 177 countries and is funded by voluntary contributions from UN member states.

The UNDP works internationally to help achieve the Sustainable Development Goals (SDGs).

Before talking of innovation or the field of artificial intelligence it might be wise with (1) a discussion of the evolution of the concept of sustainable development, (2) why they are important and (3) greenwashing before proceeding to (4) practical applications in the field of artificial intelligence.

1. The Evolution of the Concept of Sustainable Development and a Discussion of the 2030 Agenda

The sustainable development goals has a long and varied history that stretches back far beyond the conference in Paris in 2015 where they were agreed by most of the countries in the world. One could say perhaps that they began with the wish for a more peaceful coexistence after the second world war with the creation of what later came to be the United Nations in 1945, since collaborating to prevent future atrocities and nations helping each other in a more coordinated manner has been an increasing focus since then. The ideas of multilateralism and cooperation between countries is central to the creation of the sustainable development goals, and these agreements that nations make for a better world (at least the attempt) at conferences or gatherings between nation leaders.

Thereafter ideas began taking shape on the topic of sustainability in the 1970s. A notable idea was that of the idea regarding sustainability, proposed in the book Limits to Growth (1972). The novel idea was using computer simulations to model the world and our resources. One could argue this idea was Malthusian in its fright of the future projection (not taking into consideration innovation), however what it did was say that there are limited resources too (not only people). This book had a variety of authors one whom was Norwegian. Norway’s first female prime minister Gro Harlem Brundtland was appointed as the Chairwoman of the commission (often known as the Brundtland commission) that was to produce the famous report “Our Common Future” released in 1983. In it she placed a strong emphasis on the concept of sustainability.

It had a generational perspective in its thoughts of sustaining the planet, in a sense handing over the future to our children, that resonated well with many leaders. It was not then that the sustainable development goals came into being, but it did contribute to varied discussions at places such as the conference in Rio that later may have led to the development of the process leading up to the goals. Before the Sustainable Development Goals (SDGs) were to occur there was an initiative for Millennium Development Goals (MDGs). These were not made in a wide participatory manner, rather it was made by a smaller gathering of people, however with its fewer goals and focus on poverty reduction as well as food/starvation it can be argued to have been more specific and narrow (although many say it did not place enough focus on the environment). A famous success from this process was China with its massive reduction in the number of poor living in their country. Within this framework and building on the success the beginning of the sustainable development goals were taking shape.

The 2030 Agenda with the Sustainable Development goals and their development was very different from the MDGs in the manner first and foremost that they themselves were developed. As mentioned previously the MDGs was made in a far less inclusive and participatory manner while the SDGs were made together with more countries, businesses and public engagement to some degree. People, Planet, Prosperity, Peace and Partnerships (5Ps) being central to the engagement was enacted far more in practice than had previously known in history been done on such a wide scale. Getting so many countries to agree is an incredible feat of multilateral diplomacy and cooperation between nations. Due to the fact that so many had to agree it can be argued that the goals themselves became muddled or harder to understand. On the flip side there were not only 17 SDGs there were also made 169 different indicators that were placed within the different goals. It was novel too that there were a focus on consumption and production to a greater degree. These goals were not only made to ‘develop’ poorer countries, yet could be equally applicable in most cases to other countries that considered themselves to be more developed.

An inter-agency initiative to measure the goals was initiated too. Different indicators received a tier depending on a few different factors. Tier 1; had a clear measurement and steady streams of data from the different countries; tier 2 had clear measurement yet no steady stream of data; and tier 3 indicators within the goals had neither. Most of the indicators were relegated to the first tier (about 100) while fewer were said to be of the second tier (about 50+) while far less were said to be of the third tier (10+). Thus the governments or nations involved seemed relatively certain it would be possible to move forward with measurement. On the other hand measuring is difficult, because not all nations have had steady statistics agencies, and not all statistics agencies are equally as trustworthy in each separate state. It becomes difficult, and some may argue problematic, that there is so much to measure. Thus giving time and resources to measure instead of using these resources to address the issues themselves is a troubling aspect of the SDGs. Authors such as Easterly has argued that they are too vague, while other again have compared them with shorter lists of goals (one article even compared them to the ten commandments of Moses).

In addition it is still usual to use GDP (Gross Domestic Product), a form of measurement that can be more harmful to the environment as it places impetus on what is produced as a goal in itself rather than being conserving or moderate for the environment. If a land can produce to export then it is harder to understand how these goals are linked together, if it helps ‘SDG1 no poverty’, yet it harms the goal of sustainable consumption is it then still a justified action towards the goal? This is less clear and the focus on inclusive growth may forget the important talk of the trade-offs that are apparent or the sacrifices that has to be made for certain countries to even attempt reaching the goals. Norway without oil, is that a sacrifice or necessity? What sacrifices does a country have to do to enable their move towards a more prosperous and equitable future, despite for example the prospect of Decent Work, that is another SDG that is wished to be reached in the short term. What SDG is a catalyst or related to each other that is hard to tell immediately with clarity beyond doubt.

Thus SDGs operate in contested space with a virtuous label, so it may be hard to argue for one SDG as opposed to another, because they are all connected it is hard to know what goal to approach first or whether all should be approached in tandem. This may be the case for businesses as well that can be cherrypicking what part of the 2030 agenda they would like to achieve depending on their business targets. It is possible to do green-washing by claiming, aggrandizing and sharing a goal such as ‘3: good health and wellbeing’ simply because you are already selling pharmaceutical products. Studies have shown that most businesses do not relate their goals, at least in written strategies, to the different indicators within the goals. Despite this one could of course argue that it is better now that people within businesses have a clearer mandate to pursue these improvements due to a stated goal by the government.

2. Are the SDGs an Important Tool for Global Governance of Sustainable Development?

Yes, the SDGs are an important tool for global governance for sustainable development. However simply stating this is of course not enough. When challenged it is hard as well to say how these goals actually matter at all, and a classic argument is that business as usual is still occurring. If the Sustainable Development Goals (SDGs) really matter then how can it be said to do so, if nations are doing what they would do regardless: then are the SDGs not worse than useless? At worst they can shroud the harsh reality with their rainbow colors, although it is easy to love rainbows, lofty goals remain aloft of the tragedies on the planet. If you look around at your friends, at your family, are they changing because of the SDGs? It is easy to argue on behalf of abstract ambitions with concrete points and claim a measured improvement, that is ideal for policy-makers. Yet one could think that these goals do offer a great deal of grief too because of their public nature, although even for the educated public to arrest or hold to account those who enact or implement policy in their lack of focus is not at all simple.

Firstly the SDGs are not meant to be an absolute requirement for a state, and that is perhaps contentious, because who would police the emissions if they were to be policed? How would we know if a country polluted into an ocean through their rivers and to what extent? Much can be measured, but does not necessitate improvement, indeed we have trade wars and countries enact consequences such as punishing a country and their economy. However who is punished and how for their lack of urgency on the sustainable development goals. We could take a realist perspective in this way for the sovereign states have no clear responsibility except policing/protecting their own nations. However if we consider Robert Putnam’s Two-Level games perhaps if the sustainable development goals are spread enough the local public can place more pressure on their local leaders while the international leaders can create the multilateral pressure to enforce these norms. Still, countries have no set requirement beyond the initial commitment.

Secondly, financing the development goals is not an easy matter either. There is a commitment to a development budget beyond the SDGs, and Norway contributes a percentage of their GDP to this purpose. However it is clearly outlined that if we take the example of the climate crisis other countries around the equator and elsewhere will be hit much harder by the changing climate. Some islands will sink into the ocean, other countries will have little or no fresh water left (no water no humans), and others will experience extreme weather (floods, hurricanes, etc.). What happens if we do not achieve the sustainable development goals? Especially in regards to the climate crisis the world will be poorer and less habitable for humans for a long time.

Thirdly, we are learning more about global governance of sustainability as we go. Global governance of sustainable development particularly in regards to the environment is not an entirely new practice and it is my strong opinion that we will grow to learn more as time passes. Global governance of this scale and ambition has never been seen before in the history of mankind. Simply as an innate understanding that we are connected to the same issues it could be argued that the SDGs in this way are an invaluable tool. This clear view of issues that have been agreed upon creates a shared ownership of the issues. It must be said that many do not agree on the specifics, but as a framework for shared values through the colors and names alone it is something that can be shared and discussed by policy-makers, businesses and the public regardless of where you are from. The SDG4 Quality Education as an example does have some common indicators, yet in a more vague sense means different things in China than in the US currently.

As discussed in the introduction this can be seen as a crux or a fatal flaw in the tool for global governance, however if we look at it from a more human perspective rather than the maths or by the numbers we can see agreement. Policy-makers who use the SDGs in a discussion will find a platform that can be shared with other countries in enacting change in their local communities. This can be described either in this positive light or on the other end more as a global policy ‘beauty pageant’ where those who enact policy parade their success to the applause of other countries. It could be too that the negative perception as a consequence from state leaders enact a pressure on country representatives or citizens to shift their focus. There are many other pressures on state representatives that require attention so it is hard to think they would focus solely on the sustainability of a given venture. There is a mention previously of the trade-off, sometimes possibly for some it may seem a question of peace or prosperity rather than peace and prosperity.

The 2030 Agenda does not handle in an explicit way how they goals are going to join into each other. Operating in a war zone with the SDG16 Peace, Justice and Strong institutions is particularly problematic. It should preferably be an operative goal that can be used to help solve conflicts as much as prevent, but the local-global approach is not necessarily conducive.

An example from experience is the implementation of the Sustainable Development Goal 11: Sustainable Cities and Communities. With this goal the problematic aspect we found in the discussion with a central member of the UN Habitat was how much it was influenced by one particular director in the ambition for infrastructural development. When these type of measurements are agreed upon they may not of course necessarily be what constitutes sustainable development in all possible cases, rather it will be to some extent a generalization from a member of the UN system that then proliferates across the world. When asked by a government representative to pursue this goal without the critical engagement from the given representative there is a great problem that occurs on the local level.

The local level is after all where the notion of implementation lies, so it is vital that the policy-makers who are set to implement these goals to a great extent understands the pragmatic approach that is required to make this happen. When a policy-maker moves from color and number to indicators and implementation there is a wide variety of issues that arise in each separate part. First if you impose a goal onto a civil group and set out to measure based on this there may be a variety of contributing factors to the issue that are ignored in the pursuit of indicators that make little sense on-the-ground or in the pragmatic lived reality of individuals of a city.

If that is the case is it then a good tool? If the policy-maker somewhat has an understanding that this goal is interconnected with other goals that must be harmonized to make it possible to a larger degree then we do perhaps have a good tool or framework. If a policy-maker lacks flexibility and is obsessed with reporting their measurements to win the international ‘beauty pageant’ to receive the praise of an international community then it may not be the case at all. From experience it does not seem either as if there is an awareness of the trade-offs or the sacrifice in this process, rather it is more of an project to be fulfilled and then congratulated. Saying it is an important tool does not indicate it is either efficient or effective, however it needs at least to be effective in the implementation (even if you cannot expect result straight away). Economists will quickly ask if it is efficient (not all of course), and then measure the effect. Still as aforementioned paragraphs indicate it does to a large extent depends on the outlined measurement.

3. ‘Greenwashing’ and how it impacts the achievement of the SDGs

Greenwashing is the practice of using a false premise related to an argument of environmental and societal benefits of a product or a service. It could also be outright lying about your service or product through this process rather than giving a false premise. Another way for greenwashing as a practice to operate is simply attempting to make your service or product be perceived as ‘green’ or more beneficial in this same sense. This is not a set definition, and there are likely more ways to define greenwashing, however this will be an attempt to exemplify greenwashing giving one example for each variation:

  1. The false premise: this could be the partly state-owned Equinor cutting all of their emissions in the production of oil. Although this sounds helpful only 5% of emissions come from production of oil while 95% comes from the combustion of oil. Therefore stopping the production of oil and the transition towards other energy sources would likely be better.
  2. Outright lying: the OK Tedi Mining facility in Papua New Guinea is known to be one of the greatest environmental disaster in the mining industry. They claimed they would build a tailings dam, and they ended up never doing this, as such an outright lie. This lie can often take the shape or form of a promise, and it could of course be argued that this is a broken promise rather than an outright lie, however the difference in this regard is not great. After being discovered and sued for a large sum it continued with its practices while it created a nonprofit based in Singapore to own the mining facility. They promised again to build a tailings disposal, but they did not do so.
  3. Perceived as beneficial: Coca Cola announced a bottle they called Coca Cola Life. This bottle had a green wrapping and claimed to be more sustainable or green due to its use of a different type of sugar, as well therefore being more healthy for you because it did not use unnatural/artificial sweetener.

This practice of greenwashing impacts the sustainable development goals because it makes people lose trust in the concepts and words, thereby it can reflect into distrusting the strategies. As a consequence of this there can be a great amount of distrust towards these concepts: green, sustainability and climate crisis.

As mentioned previously there are research to show that many corporations know of the goals and mention these in the strategies, yet do not specifically say how they are going to address these or what indicators they will use. From experience it is often that a project ends up at the marketing or communication department instead of the product department, and becomes a way for a company to communicate that they are green or sustainable rather than actually being responsible. This can be the case on a national level as well, so that these goals are not anchored or synergies are not created between departments in the government. Instead it becomes a practice of communicating progress instead of initiating projects.

4. From theory to practice

I wrote a blog post specifically focused on the intention of wanting to move towards a more practical project:

This is after all day 403 of #500daysofAI, a personal project where I have written one new article every day about artificial intelligence — with the aim of doing so for 500 days.

So what am I going to do? I gave a hint with my article yesterday about natural language processing.

I am not any form of advanced practitioner within the field of artificial intelligence, although I know of many applications and much theory particularly within social science discussions of the topic.

What I thought to do specifically was analyse documents within the UNDP, specifically the Country Programme Documents.

“Country programmes are the operational backbone of the organization. The Independent Evaluation Office (IEO) conducts Independent Country Programme Evaluations (ICPEs) to support the formulation of UNDP’s strategy at the country level. IEO aims to strengthen UNDP’s accountability to the Executive Board and national stakeholders and to promote corporate learning through the ICPEs.”

This can be done in the form of language, however I have not seen a library either that tried to build a dataset from these text to make it easier for others to analyse or understand the overall effort of the UNDP.

As you might understand from this text I believe sustainable development to be important, although it is not an unproblematic concept or practice. I would argue that these tools are important in governance. Yet, strategy and implementation may vary greatly from country to country.

In that sense an important part of using machine learning in analysis is the data. The UNDP has gathered data on each country in a variety of ways, and they have encouraged country programmes written through country programme documents (CPDs). These CPDs can be made ready for analysis so that not only I may benefit from understanding the workings of the organisation, others may gain from being able to directly access these documents.

On the topic of greenwashing it is important to mention that the very practice of undertaking digital activities can contribute to a great deal of emissions. I do not propose in any way that these considerations should not be made. I may only use a modest amount of processing power for this project, additionally due to a lack of any further resources, however it may be well worth to calculate should computing resources increase. It must be said that building a dataset is not a neutral activity, and the dataset could be used for other purposes than intended.

If you are reading this and you have any thoughts or advice I would be happy to receive them. If you think that document analysis and research on evaluation seems like a weak proposal after the text you read I understand. I am very open for other ideas.

This is #500daysofAI and you are reading article 403. I am writing one new article about or related to artificial intelligence every day for 500 days.



Alex Moltzau

AI Policy, Governance, Ethics and International Partnerships at All views are my own.