‘Trustworthiness of AI is mainly a socio-technical concept’

About Emma Ruttkamp-Bloem

• Emma Ruttkamp-Bloem is a Professor of Philosophy at the University of Pretoria and heads the Department of Philosophy at this university.

• Her fields of research expertise include ethics of technology and the philosophy of science. She focuses on the impact of social robotics on the human condition, the ethics of autonomous weapon systems, AI and sustainability in Africa, and fair, accountable and transparent machine learning. In addition, she focuses on problems relating to scientific realism and the structure of scientific theories.

• She currently participates in research projects on ethics of AI, scientific realism and the structure of theories and epistemic justice.

• Ruttkamp-Bloem was elected chair of the Bureau to oversee the final drafting of the UNESCO Recommendation on the Ethics of AI.

What is the importance of UNESCO’s global Recommendation for the Ethics of AI for the African region in particular?
‘In general, I think it is especially important for regions such as Africa, Latin America and the small island states to have a seat at the international table. Yet, let me speak from an African perspective. When it comes to AI development and research, a lot has been done the last years. At the same time, the needs of Africa have not really been recognised internationally. We are always seen as some kind of dumping ground of the Global North. We help to improve AI models to make better predictions and we provide the labour. Despite that, we do not have any access to decisions that really pave the path for standards for responsible use and development of AI. Plus, in most cases we do not own the data we generate for transnational companies. This is a serious concern.’

Can you give an example of the uneven relationship you have described?
‘There is one despicable example of the self-driving car companies. In order to label the data of city traffic situations – data that can be used for the models to learn - these companies use people who live in terrible conditions in Western Africa. They work under dire conditions, which you can compare with sweatshops. Thus, they are exploited and truly live in terrible circumstances. Let this sink in for a moment: these people are providing information to develop a technology that they will never have access to. Moreover, they are paid almost nothing. There is nothing fair and responsible about that.’ 

It worries you.
‘Sure and it is not just this example that illustrates the vulnerable position of the Global South. Another example is that we in Africa do not own any of the data that is being collected by big Western companies such as Amazon, Google, Facebook and Uber. There is a serious issue with data sharing in Africa. That is the case in situations where apps just get dumped on us.

For example, a digital doctor app, that was first introduced and used in Tanzania, turned out to be worthless. Because, all the data in the app came from the Global North. It took almost two years to be useful in the African country. It got framed and sold as if it would solve all kinds of medical problems. While in fact, it actually created a myriad of more problems.’

What do you consider is the most important issue to bring into the discussion for the UNESCO recommendation from an African perspective?
‘The most important part, which is something that I find missing in the European perspective, is that the value of trustworthiness should not only be seen as a technical issue.’

What do you mean by that?
‘I think it is much needed to more clearly define what we are talking about when we refer to trustworthiness. In Europe there is a lot of focus on trustworthiness in the sense of robustness, safety, accuracy of the AI systems itself. It even means accountability, explainability.
While I strongly feel that trustworthiness should actually be viewed as a socio-technical concept. Which means that the flip side of trustworthiness is: respect for the users. By looking at it in this way, it is clear there is a strong link between trustworthiness and principles like the human dimension, human oversight and determination, diversity and inclusiveness and privacy. Principles that relate to the impact of AI systems on humanity and how humans interact with those systems.

It is important that not just the systems are trustworthy. Actually, in my vison trustworthiness means first and foremost that humans have to be able to trust AI technologies. But tech companies must realise that this trust has to be earned, in every context, continuously, every step of the way.’

Earning trust is a delicate matter – what do you suggest?
‘Let us start by being real and realise that a good recommendation it not only about creating a document for which governments have just ticked all the technical boxes for fairness, accountability  and so on. Because that does not necessarily mean that people feel that they can trust AI systems. I would love to see a concerted effort to really demonstrate that in all aspects where AI technology has an impact on humanity, this impact is regulated. This will lead to trust. However, this is not an easy job as many countries have different perspectives on how to ensure that no one is left behind.’  

Is the African perspective sufficiently included in this UNESCO recommendation on the Ethics of AI?
‘What I find striking and very positive is that there were four experts from each region in Africa part of the expert group that wrote the first draft for the recommendation. That is not unique in the context of UNESCO. What I do find unique, is that it concerns compiling an instrument on the ethics of AI. Until now, these kinds of ethical guidelines for AI all came from the Global North. For these guidelines, there may have been at most one representative from Africa. Usually African representation has been in the minority and there always tends to be some imbalance. Or worse, the African perspective is brought in as some kind of token to presume inclusiveness. But that is really not the case here and I think that is very refreshing.’

Another thing that is refreshing is the fact that in the UNESCO recommendation values are the foundation – not legal norms.
‘That is so important to me. Values are not legal norms, but they are supposed to guide the shaping of legal norms. In addition, ethics is much more dynamic than international law. All legislation should of course be in compliance with human rights, principles and standards and with international law in general. There is a very urgent role for ethics, especially with this kind of fast-moving technology. By putting values central, we were able to bring in more refined perspectives from all over the world, to make it more inclusive.

Let me give the example of a new value we added, that was first called ‘living in harmony’. We brought in specific ideas/concepts from Africa, Latin-America and Asia and combined them into a value which involves and recognises interconnectedness. This is a deep value that comes from Africa. It is called Ubuntu, which means: ‘I am because we are’. So, if I deny another human being his or her dignity, I deny my own dignity. Another element we added in this new value is the harmonious co-existence between humans and the natural environment. That comes from Latin America and Asia. It is closely related to their way of looking at life as a whole.’

That is quite distinctive.
‘It is something that,  I am sure,  is missing in the dominant Northern or European vision. I think that these things are taken for granted in Europe, so Europeans do not have to point this out. Where for us in Africa, this is a rule, telling us who we are and everything that we do. It is important to frame ethical regulations in terms that all cultures can relate to. This is what makes this recommendation unique and comprehensive. Moreover, we ensure inclusiveness in this way.’

Next to the values, there are also policy actions. What do you consider the most important one,  especially for Africa?
‘For sure, that is capacity building for AI ethics. Not in the empty sense of the word, because you cannot talk about the ethics of AI, if you do not know what AI is. Therefore, what we are trying to do now in South Africa is to introduce a public Digital Ambassador Program. This is very concrete and has been done very successfully in Rwanda as well.

Thus, awareness needs to be created. In order to really achieve something, you need to focus on individuals. We need all African people to understand why it is important to talk about the ethics of AI and make them understand what the impact of AI is on their lives. Or what it potentially can be. So, we need them to understand the possible disruptive nature of AI technologies, as well as the possible benefits.’

That is a big challenge all over the world.
‘Awareness of AI fundamentals is just the first building block to involve civil society. However, you have to realise that in Africa there are a lot of people who do not have access to electricity. In order to reach them, you literally have to send people,  ‘ambassadors’, to their communities.

What I like to emphasise is that in the recommendation we do say that it is important that we have Public Awareness Programs. Yet, this comes with the realisation that Public Awareness Programs in a continent like Africa only reach people in cities. Where a huge part of countries consist of rural populations. Even in cities, the poorest of the poor will not have access to these Public Awareness Programs. So, for that reason, you need individuals to interact with local communities. They need to be talking with them about this, and explain the positive and negative effects and how it can impact their lives.

It is challenging when you have to talk about such a complex issue to people that have never even seen a cell phone and do not have electricity. Basically, they often do not care about this, because they have different needs. However, in the end, their lives will be impacted too. Unfortunately, that is just the way it is. In their case, it can mean that they will be pushed away further and further. We cannot have that. That is exactly the reason why it is important to work with dedication on AI literacy.’

Even in the Netherlands where 98 percent of the households have access to internet and use cell phones, capacity building is very challenging. Where to start in the rural areas of Africa?
‘First of all, it is important to make it less abstract. Of course, in an international recommendation we can only speak of it in a more abstract way to achieve global approval. But to involve ordinary people, we need to speak about concrete examples. Johannesburg, for instance, is now full of surveillance cameras. What we need, and ideally this recommendation contributes to this, is more and more people saying: ‘I want to know what happens with that information. This is my basic right: privacy and protection of my information.’ If people start asking these questions and insist on getting the answers, and receive guidance to know what rights they have in terms of responses, we will  build a truly ethical and global AI community.’