Decolonising Artificial Intelligence 3

· Read in 6mins · 1297 words ·

The Artificial Intelligence we believe to be global, is far from it. It is localised. Contained. Within restricted geographies and people. As the pressure of containment builds and the kettle begins to whistle, the call sounds.  How do we make global AI truly global? 

This call will be made in different ways. Inevitably, a call will be made to decolonise artificial intelligence. We understand why. Because, as much as we aspire towards an ideal scientific-self—the scientist that is forward-looking, sceptical and inclusive—we often fall far from this ideal. At times, we replay a colonial-looking world view. We rely on inherited thinking and sets of unquestioned values; we reinforce selective histories; we fail to consider our technology's impacts and the possibility of alternative paths; we consider our work to be universally beneficial, needed and welcomed.

And so the call sounds. #Decolonise. Decolonise our Minds. Decolonise Science. Decolonise All the Things. This is the call that resonates today, loudly; louder than it ever has.

The call for decolonisation in artificial intelligence1 2 is yet to reach its full volume. It is a topic that I suspect is on the horizon, and one we will need to engage with. I am conflicted by a view that considers decolonisation as irrelevant to the challenges of developing human-centred artificial intelligence (AI), versus decolonisation as the critical thinking with which we shape a self-reflective and responsible community of AI designers. This essay is a reflection on my conflict, perhaps yours.

Defining Decolonisation

Decolonisation emphasises self-ownership. For land and peoples and nations, it emphasises the restoration of 'land and life' to its original inheritors. Self-ownership, as Fanon says3, is a 'precondition for the creation of new forms of life'. To create new ways of living and being, we need self-ownership.

For knowledge, decolonisation emphasises the restoration of self-confidence. We need self-confidence to create new thought and ways of thinking.

There are three views of decolonisation that give me a more concrete understanding4.

As a de-centring of European knowledge. One can remove the colonial power, back to England or to France, but their legacies of thinking, psychology, economics, and culture remain. Decolonisation asks us remove Europe—or the 'west' more generally—as a point of reference. We should reject an imitation of the West in our scientific work. We should assert our own identities. We should re-centre our knowledge on approaches that restores our global histories and problems and solutions. For Ngũgĩ wa Thiong'o, this meant rejecting the English language as the unassailable medium of teaching and discourse. For we technologists, this could be re-centring our motivations, not always on the problems of silicon valley, but on our own localities and questions and backgrounds.

Towards additive and inclusive knowledge. To remain globally competitive, and to address our key intellectual and social challenges, this form of decolonisation asks us to work together to use existing knowledge, but to explicitly recognise the value of new and alternative approaches. We should include other histories, like the ongoing correction of history that includes the Asian and African contributions to science5 6. And we should support environments in which new ways of creating knowledge can flourish. What we must not do is include, and then relegate new voices to the periphery. Of relevance here is the open-source, open-access and open data movements in which we are already active participants.

Engaging settled knowledge. We are asked to empower each other to take a critical view, always. Where does our knowledge come from? What does it include and leave out? Whose interests does it support? What are the assumptions by which our thinking is governed? What and who is silenced? This is the view that seems most relevant to our practice. Here, we are already engaged in areas of fairness, accountability, transparency, safety, robustness, verification and governance.

AI will result in objects of culture, and its use will have impacts on the way our societies work (it already has). And this is why we should consider decolonisation seriously. But we are already engaged in these issues, under different headings and, importantly, in relation to their underlying technical questions. This brings the utility of decolonisation as a tool into question.

Not a Metaphor

The famous paper by Tuck and Yang entitled 'decolonisation is not a metaphor'7 enters:

'Decolonisation brings about repatriation of indigenous land and life; it is not a metaphor for other things we want to do to improve our societies and schools

E. Tuck and K. Wang, Decolonization is not a Metaphor, 2012.

This is an important clarification. The key message they leave us with, is that it is problematic to use decolonisation as a placeholder for all the ways we wish to engage with social justice. It leads to a loss of meaning.

Too often is decolonisation used for political symbolism. Too often are claims for decolonisation used to raise opposition, without genuine concerns. Too often is decolonisation used to signal an enemy. Our science will not be advanced through a world-view based on empty symbolism and opposition: together we can move beyond the too-easy narrative of them-vs-us, coloniser-vs-colonised, metropole-vs-south, west-vs-rest.

More worryingly, decolonisation increasingly seems to be used as a replacement for Transformation. The price of transformation cannot be paid by allowing ourselves to be distracted by the language of decolonisation and delaying the work of deep social, institutional and personal responsibility for change.

We can learn from the questions that decolonisation raises and the strategies it suggests. But we can also maintain our focus on the important questions of social justice. I will continue to demand for a simpler language. Let us say what we mean and want, to aim to be understood, to be more precise. And we can do this by keeping the challenges we identify and their scientific and technical basis in close proximity.

Lessons for Artificial Intelligence

There is a reason for concern! What else can we see when read in the New York Times8 about one future path for our countries:

"Unless they wish to plunge their people into poverty, they will be forced to negotiate with whichever country supplies most of their A.I. software — China or the United States — to essentially become that country’s economic dependent."

Kai Fu Lee, The Real Threat of Artificial Intelligence,  June 2017

We immediately recognise the colonial nature of this possible future. When Ian Hogarth writes of AI nationalism9,  we recall the hubris and implications of the nations that sought Empire. We are not oblivious to the fact that a form of imperialism based on data10
and its ownership is possible (if not underway). We do recognise a forming cyber-colonialism11 that expands as censorship increases and online freedoms are curtailed.

This is where decolonisation plays its role. The solutions that have been tried in the restoration of land and life in the post-colonial age, can be ours to learn from and reuse. The basis of this solution is in self-ownership and its protection.

Fortunately, as a field we have the basis of such protections already. We can continue to strengthen open-source software, open-data, and open-access science— publishing more, not less; we can further support accessible machine learning frameworks, and accessible scientific communication; and we can continue to find solutions to the challenges of fairness, privacy, safety, verification, and governance. And we can go further, by always challenging our settled assumptions and world-views as we expand the frontiers of our knowledge.

The only AI that empowers and works for the benefit of humanity is a truly global AI. Making global AI truly global will not be easy. We have heard the call.  

This essay was written while listening to My Queen Is Ada Eastman from the Sons of Khemet. You can also read a related essay on The Price of Transformation, or one of my earliest writings on Marr's Levels of Analysis.

--Shakir, decolonising the mind.Sunset alongthe Regent's canal in October.


  1. Decolonising AI at the Leverhulme Centre for future of intelligence.
  2. Video: Decolonising Artificial Intelligence? by Professor Genevieve Bell
  3. Frantz Fanon, The wretched of the earth, 1963
  4. J. Jansen, As by Fire: The end of the South African University, 2017
  5. Clapperton Mavhunga, What Do Science, Technology, and Innovation Mean from Africa? June 2017
  6. Caroline Tagny, Feminism online in west and central Africa: identities and digital colonisation, May 2016
  7. Tuck E, Yang KW. Decolonization is not a metaphor. Decolonization: Indigeneity, education & society. 2012 Sep 8;1(1).
  8. Kai Fu Lee, The Real Threat of Artificial Intelligence, New York Times June 2017
  9. Ian Hogarth, AI Nationalism, June 2018
  10. Data Colonialism: Critiquing Consent and Control in “Tech for Social Change”, Model View Culture, June 2016
  11. The Dawn of cyber-colonialism, June 2014

3 thoughts on “Decolonising Artificial Intelligence

  1. Reply Eren Sezener Oct 12,2018 1:42 pm

    cyber-colonialism link is broken

  2. Pingback: Artificial Intelligence/Machine Learning Roundup #106 | Daily Artificial Intelligence & Machine Learning Curated News

Leave a Reply