Decolonising the Internet

It’s Wednesday, I think, so I’m in Dublin for the first face-to-face AoIR conference since AoIR 2019 in Brisbane. It’s genuinely delightful to be amongst this wonderful community again at last. As usual, the conference starts with the conference keynote by Nanjala Nyabola, addressing the conference theme of Decolonising the Internet. She begins by noting that the vast majority of people experience the Internet in a foreign tongue; and it is appropriate to address this issue in Ireland, which has had its own history of having its national identity and language suppressed for so long.

Nanjala’s keynote is based on research which worked to translate keywords from Internet research into Kisuaheli, and the assumption from others has always been that this was an AI and natural language processing project; but it was not, and the real question is what it means to be human in the digital age. This was also her first academic paper in Kisuaheli; it was a project in decolonisation. How does this even happen? Ultimately, as so often, the story begins with the arrival of the British: the colonisers. And too many people in the world still don’t know what it means to be colonised – the damaging, scarring disruption of history and culture; one of the darkest and bloodiest chapters in human history that reorganised societies for the economic benefits of imperial power, and a form of bureaucratised murder, systematised rape, and legitimised robbery.

In Kenya this lasted less than a century, but in the last decade of colonisation alone tens of thousands were killed, often simply for speaking out against oppression. In addition, lives were disrupted by introduced pests and diseases. This form of structural violence was documented in the files of the occupiers, but the larger loss of culture is less measurable, and the patterns of colonial administration often still continue. And the intention of the violence was to reorganise society to make money; to create ideal labourers – including by rooting out local languages by force in schools. That legacy still endures, and the trophies of this violence still remain in British museums, while culture is still being reclaimed and relearned.

Decolonisation is about freedom, and this means asking: freedom from what? The question of language is especially important here. When the Kenyan government required the people to sign up to a new digital identity system that was built on the old colonial identity system, this continued the colonial system; but the activism against this new system had to continue to use English terms because equivalent terms in Kisuaheli or any of the other 44 Kenyan languages did not exist. This continues the legacy of colonial oppression – and the instrumentalisation of language is a key colonial project. This means that decolonising the mind by developing equivalent terms in people’s own languages is a critical project: language is the most important vehicle through which colonial power held the soul prisoner.

To decolonise Africans intellectual thought, therefore, it is necessary to use African languages. Just looking at language can give us a sense of the size and scope of the task of decolonisation that still lies ahead. Moving away from English requires concerted political action – decolonisation, a move away from colonisation that recognises the scope of the damage that colonisation caused, and cannot repeat it. But English is the leading language of the Internet, and of the platforms that are now experienced as ‘the Internet’ by most people in the world, and this has been widely perceived as normal by Internet users themselves; there isn’t even much effort to support an Internet in diverse other languages.

The first biometric ID system was built in apartheid-era South Africa; digital identification systems are the legacy of this work, and there have been many cases now that demonstrate the colonialising violence that such systems exert on the populations that are governed by them – in Afghanistan, Assam, and elsewhere. Some of this violence is structural, and some of it is physical; the size of the misinformation struggle is exceptional, and it is global. The choice to underinvest in content moderation in the global south was a business decision, and has had direct impacts on the Internet users there, who already experience the Internet in translation. Failing to keep people there safe is a choice, to view human existence through a specific lens, a colonial lens that focusses own the financial bottom line rather than on the people affected. ‘Move fast and break things’ also means breaking people.

This is structural violence that serves to reorient the labour of the people who have the least to the benefit of the people who have the most; it enables the massive data harvesting that makes these systems more efficient, but puts these data out of reach of the populations concerned. What does this mean in contexts where commercial power is completely abstracted from the people from whom it is collected; how can we overturn this situation to create a more free, open, and egalitarian Internet? There are many examples of such systems, and their effects amplify as they are trained on each other’s data. Data extraction without concomitant investment in community creates the opportunities for data colonialism.

The question in decolonising language is not whether people can write in English, but whether they ought to. Most Kenyan government services are only available in English, for instance, even through the vast majority of the population only speak English as a distant third language; so who do we assume here is the end user of these technologies? Kenya is not even the worst case in this context, but the burden is shifted to the speakers of these languages without, and exploited by the platforms. Again, the project here is not an AI project. Decolonisation is work: not just creating words and a lexicon, but also the mechanisms and structures to disseminating these terms and enabling speakers to adopt them; this means talking to and building with people, it is centred on rebuilding the threads of community that the violence of colonisation has destroyed.

And Kisuaheli isn’t necessarily a language of choice for Kenyans either, so there needs to be an open acknowledgement of the structures of power that these languages represent. Decolonisation needs to be about restoration and restitution: it needs to yearn towards balance with nature, and build an Internet based on and enriching human experience; it needs to make the materials as widely and freely available as possible, and generate value well beyond the idea of technical efficiency. The machine exists to serve the people, not the other way around. This may also be good for the natural environment.

And there is no win-win in decolonisation: it is disruptive and uncomfortable work, but this discomfort is necessary. This also means avoiding the monetisation of the new structures being created; power does not concede easily and needs to be fought, and this will inevitably reduce the privileges of some. Things will have to break in order to make room for the new: this will create a feeling of being adrift and confused for many of the people who will have to concede their privilege in order for all of us to be better off in the end.