AI Data Extraction

Our resident Humanities and Anthropology Lead, Dylan Kwande, is doing research into the field of deep data and the impact of data extraction on marginalised communities. He purports that data should not be extracted simply for optimisation and economic reasons but must be owned and managed from an ethical and politically responsible foundation

What’s the value of data?

Data encodes power and can have profound effects:
Lives are rendered legible
Knowledge counts
Realities are erased

Data’s real value lies in its capability to shape the future:
Norms
Markets
Governance

From a decolonial perspective, data represents sedimented social labour and is not a neutral resource. Its value comes from lived experience, cultural knowledge, and historical context – often extracted from marginalised communities without consent or compensation.

Treating data purely as an economic asset, reproduces colonial logics of extraction.

Why should firms be concerned about where their AI data comes from?

Purpose:

• Where data comes from shapes who benefits and who bears the harms of the output from AI systems.
• Training AI with data extracted through coercive, opaque, or unequal conditions can reproduce epistemic injustice – systematically disadvantaging already marginalised groups.

Risk:

• • Data provenance is a moral and political risk, not just a compliance issue.
  • Firms risk legitimising exploitative knowledge systems, undermining trust, and facing societal backlash.

Regulation – governance regimes are shifting toward data rights, consent, and reparative justice.

From a decolonial lens, “bad data” is not just inaccurate; it reflects histories of domination and exclusion.

Recycling AI proceeds back to the source of the data

Fairness – derisk innovation, strengthen social licence and model an economy of AI grounded in fairness rather than dispossession.
Reciprocity – reframes AI from extraction to delivering benefits to the source of the data. This recognises communities not just as raw inputs, but as co-producers of value.
Social impact – this can take the form of revenue sharing, infrastructure investment, or participatory governance. Strategically, it builds legitimacy, trust, and long-term data sustainability. Normatively, it aligns AI with reparative justice, countering colonial patterns where wealth flows outward while harms remain local.