Sankofa

What does the past mean to a future-facing technology? Some might argue that the vast repositories of data have come to bear because of a string of “pasts” that have been mined. Some might suggest that the past is data inherently. This immediately begs asking what past is remembered by whom, and in what ways. What might it mean to go beyond a flattened understanding of the past? Can it mean more than a string of patterns that have emerged over time? What will we gain if we reflect on the past beyond just a limited understanding of it? What happens if the past is not “documented” in the format of data that we know today? Does it risk being lost forever?

Sankofa, a concept from the Akan people of Ghana, invites us to reflect on these questions deeply, and cultivate a reflective approach to engage with the past.  

Understanding Sankofa

Before proceeding, we note that Sankofa is a concept from the Akan people of Ghana and is part of a living visual and philosophical tradition. We approach it with care, with the understanding that concepts from African traditions have historically been appropriated and stripped of context when they travel into other contexts. This note in our archive is an attempt to honour the depth of this concept while being transparent about the limits of our interpretation.

In the Akan culture, the idea of Sankofa is represented by a bird that flies forward with its head turned backward. The Twi expression associated with Sankofa is “Se wo were fi na wosankofa a yenkyi,” which means “It is not wrong to go back for what you forgot.” Sankofa is a lot more than bringing up memories, focusing on nostalgia, and thinking of the past. It is a claim about the relationship between the past and future and reminds us that moving forward well requires us to actively retrieve all that has been left behind or taken away from us.

In Ghanian thought, Sankofa has been the underlying wisdom for actions to address the ruptures of colonialism and the transatlantic slave trade, both of which involved the deliberate destruction, erasure, and appropriation of cultural memory, knowledge systems, and social structures. Sankofa refuses erasure by asking us to go back, retrieve, and carry forward all that was destroyed and erased.  

What Sankofa unsettles in AI governance

As an “emerging” technology, AI has been inherently tied to the future and associated with all things “futuristic.” Rhetoric pushing for AI literacy has also consistently deployed the “follow along” or “be left behind,” suggesting that the adoption of AI is not negotiable. Within this space, the past appears in two ways: One, as data in the form of training datasets, historical patterns, legacy systems, and gaps and inadequacies in extant systems; and two, as the outdated, “left behind” world that did not go along with the flow of AI uptake. Sankofa, however, asks us to reflect on this reality, and see the past as something to which we owe an active debt through our governance.

The AI industry has been built on two specific kinds of forgetting, where one emerges from the omission of whole communities and the other comes from the flattening of complex human lives into bite-sized data units that are appropriate to train a system. By not accounting for communities whose data were extracted without consent, workers whose efforts went into labelling training data under exploitative conditions, vast language and knowledge systems that were absorbed into large models without attribution, and the historical patterns of harm, discrimination and bias that are being encoded and amplified by automated systems, the AI industry has centred some to the exclusion of several others. This structure has meant that the benefits of the tool come to a few, while several others bear the cost of those benefits.

Similarly, in gathering data, the AI industry sees a whole human being, a complex whole with a range of nuances and social relationships, dependencies, needs, wants, and behavioural patterns, as a mere datapoint. In flattening a whole complex web of interconnections, emotions, feelings, thoughts, beliefs, and ideas and much more, this structure has meant that humans have been homogenised into countable units that can be fit into a machine’s formula.

Sankofa invites governance to respect the past. It invites us to ask what AI has already forgotten, and what it could mean to go back for this knowledge. It asks us to pay attention to which knowledge systems were excluded and which knowledge systems were appropriated with consent excluded. It tells us that governance cannot begin with asking “What harms will this cause in the future,” but must rather start with “What harms have already been caused by the emergence of this technology, and what do we owe whom by way of repair for these harms?”

Sankofa also reminds us of the many historical biases and discriminations that are embedded in our worldviews and thus find themselves a place within the AI systems we build. It asks us to recognize these past harms and commit ourselves to repair. It is a call for us to practice epistemic repair by restoring what was taken, attributing what was erased, and naming everything that was deliberately written out, omitted, or forgotten.  

References

Stanley, J. T., & Chukwuorji, J. C. (2024). Sankofa: Learning From the Past to Build the Future—Introduction to the Special Issue on Aging in Sub-Saharan Africa. Innovation in aging, 8(4), igae031.

Opare, G. (2025). Sankofa as Praxis: Ontological Healing, Memory, and Decolonial Return. Cultural Studies↔ Critical Methodologies, 15327086251412926.

Wiredu, K. (1996). Cultural Universals and Particulars: An African Perspective. Indiana University Press.

Gyekye, K. (1997). Tradition and modernity: Philosophical reflections on the African experience. Oxford University Press.

Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. In Algorithms of oppression. New York university press.