Towards Indigenous Understandings of Artificial Intelligence

Well, we’re finally here: AoIR 2018 in Montréal has begun. We start with the keynote by Jason Lewis, who addresses the continuing rise of white supremacy in recent years. He begins by referencing the novel Riding the Trail of Tears, which discusses a retracing of the removal of the Cherokee from their traditional lands through virtual technology, and the possibility of Indigeneity in a digital earth.

But such a perspective clashes with white supremacy, which is well established in societal power structures even without further action to entrench it more deeply. Jason compares this with the multi-layer hardware and software stack that digital interfaces operate on; we are subject to the regimes that the stack places upon us and have no meaningful way to escape them. In much the same way, white biases are a feature, not a bug of contemporary society at every level; in software, biases beget biases because new data and new systems are built on old data and old systems, and perpetuate their built-in assumptions, and the same is true in societal protocols. This is a millennia-long process or epistemological inertia.

The growing role of algorithmic systems further entrenches this inertia: machine learning builds on existing data and perpetuates its patterns; it does so on the basis of proprietary and often undisclosed logics. Contrary to the industry hype, machine learning is not free of bias; it merely represents the encoding of bias, as Kate Crawford has argued. Machine learning is where the philosophical shit hits the real-world fan.

So, again, bias is not a bug of the system; white supremacy is built deeply into the fabric of contemporary society. If it is a feature of the system, how do we respond? In the novel Jason referenced at the start, the participants in the system evolve it beyond its original design parameters: this requires imagination, and the rise in white supremacy at the present moment therefore represents a failure of our collective information to shape our reality towards different ends. What if we took a fundamentally different approach?

If humans are not the height of creation, then we might be able to make kin with machines, animals, objects and rethink the aim of artificial intelligence as going beyond serving humans. Can artificial intelligence escape the anthropocentric box? Could this engender a greater depth of mutual respect between human and non-human kin, rather than perpetuate a treatment as tools or even slaves.

But this approach cannot be described simply as an Indigenous epistemology, because there is no single Indigenous perspective or voice. Particular world-views arise from particular territories; knowledge is relational, and this is true in the physical as well as the digital world, especially as these are increasingly interwoven. How can Indigenous peoples understand this new territory, and claim it as their own? How would they engage with AI at various levels of self-awareness?

Different Indigenous cosmologies will produce very different perspectives on these questions – and they should not simply be appropriated by outsiders, but should be used to inspire new ideas. And again, there is no single ‘Indigenous’ perspective even within the same culture.

Further, how would such understandings be operationalised in actual software design; what new and different biases might this introduce into the layers of the hardware and software stack that in turn build on it? The question of the provenance of these ideas remains, even if the provenance itself is now different, and needs to be made more transparent.

Jason’s Initiative for Indigenous Futures is pursuing these concepts further. It is translating languages such as C++ and C# into Hawai’ian, for instance; it develops fictional explorations of how Indigenous cultures may engage with AI; and it investigates how Indigenous categorisations of animate and inanimate might translate to intelligent systems. What, in short, does it mean to be Indigenous in the digital realm; how might we get to a point where the digital is a good place for Indigenous peoples?

One thing is clear, though, Jason says: we must try, because we flourish only when all of our kin flourish.