Abstract
In this commentary on Azadeh Akbari's article, “Uneven Datafication,” I build on her work by offering further ways that the concept of uneven datafication can be used, developed, and elaborated. I briefly outline two different theoretical expansions built on Akbari's analysis of uneven datafication. First is the question of uneven legibility and the power of inscrutability—or how we account for negative cases when people and places are not subjected to datafication. Second is how uneven datafication fits into the longer history of development in various productive forces for the purpose of turning subjects into objects of value extraction. By fleshing out this extended universe of uneven datafication, I hope to add further depth and utility to Akbari's concept.
If datafication is a core feature of the socio-technical systems that dominate our lives and shape the world, and if unevenness is a core feature of the political economic regimes that also dominate our lives and shape the world, then it makes sense that any critical analysis of—and challenge to—technological capitalism must account for the dynamics and impacts of uneven datafication. For many scholars engaged in critical debates about the political economy of digital technology and the social justice issues raised by data extraction, the concept of uneven datafication will feel familiar. Yet, in her article, Azadeh Akbari (2026) gives the practice a name and substance that goes beyond any intuitive grasp which might come from seeing two ideas joined together in conceptual unity.
Akbari (2026: X) offers an analysis of uneven datafication as a way “to describe the persistent creation of unevenness in data production, extraction, and control operating on a multi-scalar level.” In doing so, Akbari crystallizes many core concerns in political economy and pushes past the tendency to use terms like “colonialism” as evocative metaphors in fields like Critical Data Studies for universalist practices of extraction. Instead, Akbari's paper helpfully draws deeper connections across related bodies of work in Geography, postcolonialism, uneven development, and dependency theory. She provides a contemporary analysis of information technology that is directly situated in foundational work on colonialism as an ongoing process rooted in historical material relations, which now manifests through global digital infrastructures.
The value of a concept, for me at least, depends on what I can do with it. What new forms of understanding does it reveal, what new ideas for research does it generate, what new approaches for analysis does it unlock? The value of uneven datafication became apparent as I engaged with how Akbari develops the concept through three “interrelated notions”. First, the production of de/re/territorialised spaces of digital colonial capitalism. Second, the annihilation by dispossession of humanity-as-data. Third, the asymmetrical circuits of data-driven value creation and capture. The article stands on its own, but while reading, I kept thinking of many more ways that this concept could be used, developed, and elaborated. So I want to use my commentary to briefly outline two different theoretical expansions built on Akbari's analysis of uneven datafication. By fleshing out the extended universe of uneven datafication, I hope to add further value and utility to this concept.
Uneven legibility
Uneven datafication directs our attention to questions about the production of power/knowledge—what data is created, in what ways, by and about whom, for whose benefit and harm? These are the concerns that largely motivate critical work documenting the processes and effects of datafication. The general assumption in so much of that work is that the total transformation of everything, everybody, everywhere into data by the forces of capitalism is already well underway (just see my own past work in this vein: Sadowski, 2020). To be sure, those processes are very real and hyperactive, but they are not totalizing or universal. Not all people and places are touched by datafication in the same ways, and importantly, many of them are ignored by datafication for various reasons.
To know the world is to exercise power over it, but power also comes with the ability to decide what forms of knowledge are created and what data is neglected or obscured. Uneven datafication highlights that the absence of information is just as important in our analyses of power/knowledge. We should also be accounting for the negative cases: when is data not created, who is not subjected to datafication, how are some things made legible while others are made inscrutable? 1
Akbari (2026) does not directly explore these points about presence vs absence, legibility vs inscrutability in her analysis. But as a concept, uneven datafication does more than call forth concerns about the differentiated ways that people are enrolled into practices of subsumption and circuits of extraction by data-driven colonial capitalism. As Akbari observes “not all human life is treated identically in datafication processes” (Akbari, 2026). We should also expand its theoretical utility to raise issues about how people are excluded, removed, or hidden from systems of datafication. The explicit connections Akbari draws between uneven datafication and uneven development point us to the ways that the growth machine of (colonial) capital also relies on structural forms of stratification and dispossession, obfuscation, and abandonment.
In addition to the stark contrasts between cycles of investment and exploitation in different places (Rodney, 1972), the same place can also be subjected to simultaneous dynamics of over-investment (say, into real estate or police forces) and under-investment (say, into community centers or social services). Expanding these dynamics through uneven datafication, we see how certain populations are also both over-datafied (say, by a bureaucracy used for immigration enforcement) and under-datafied (say, by a census used to allocate community resources).
The complex realities of uneven datafication mean that in some cases being legible as machine-readable can be incredibly harmful to the datafied subject. Indeed, this is the main focus for critical scholarship on data, surveillance, and algorithmic governance. While in other cases, being legible can be beneficial—perhaps as a password that grants access to helpful systems or services—or even necessary for survival: such as a passport that proves citizenship and the rights it confers.
On the flip side, being inscrutable because you are unrepresented or missing from datasets, or because your data traces are so thin they are meaningless, also carry benefits or harms depending on the specific context and your social position. Inscrutability may be a boon if you are a powerful figure trying to hide from prying eyes or a protester trying to evade arrest by automated facial recognition. Or it may be a curse in the case of machine learning systems used for modeling exposure to climate risk or allocating welfare to vulnerable groups.
The point is not to say that datafication is a land of contrasts; who can say whether legibility or inscrutability is good or bad? But rather that his first theoretical extension of uneven datafication demands that we also pay close attention to this dialectical relationship between who is made present and who is left as absent in datasets. That we recognize how processes of over- and under-datafication are happening at the same time, to the same people, in various ways, through different systems, all with highly uneven distributions of benefits and harms that fall on predictable patterns of power and privilege.
Objects of value extraction
According to the technical language of computer scientists and software engineers, any “entity” that is targeted by a machine learning system is labeled as “objects” (Kalluri et al., 2025). This includes the “object” most often targeted by these technologies: humans. A world made into data is a world without subjects, only objects that can be consumed, targeted, and valorized by machine learning systems. While subjects have social rights and moral value, objects are here defined as only having technical features and financial value that can be exploited without worrying about other concerns. An object has no capacity for (self-) ownership and thus cannot be dispossessed since it cannot possess anything in the first place. As Akbari notes, “loss of self is not unique to data colonialism; it is a continuation of centuries of colonial violence” (Akbari, 2026).
Akbari further explores this objectification by datafication when discussing how people can become so embedded and constituted by digital infrastructures that they become a “code/body”—a new ontological relation formed by interdependency between the digital system and physical body. “The code/body functions as an uneven spatial trap, physically or virtually, by exposing some individuals to continuous surveillance, control, and profiling, ultimately treating them as a moving part of the machinery of uneven datafication” (Akbari, 2026). This calls to mind Marx's (1990: 548) discussion of “dead labor”, or the labor embedded in tools, machinery, and other materials used in production, and how capital is always innovating ways to treat workers as merely a “living source of value,” who are no better than any other component and are regularly subordinated to the dead labor of technological systems that facilitate value extraction. There is a long history of social techniques used to annihilate the subjectivity of human groups by rendering them as objects that can be captured and controlled by violent systems of value extraction.
Akbari's theorization of uneven datafication is useful for the ways it historicizes contemporary relations, situating them in invasive practices of territorialization and dispossession, and showing how they have evolved over time. Thinking with Marx, this concept invites us to consider how uneven datafication and digital colonial capitalism fit into broader historical developments in the modes of production. That is, how specific arrangements of productive forces (e.g., labor, tools, machinery, knowledge, resources) and social relations (e.g., power, classes, trading, legal codes, organizational structures). When Akbari (2026) notes that data colonialism “is a continuation of centuries of colonial violence,” we can go even further in tracing this long economy of violence and the colonial capitalist development of extractive modes powered by different productive forces, structured by different social relations, and based on different forms of capital (slaves, machines, data).
In my recent work, I have argued that centuries of capitalist investment and development of technology have been driven by the ambition to build what I call the “perpetual value machine. In short, the machine would be a way to create and capture an infinite amount of surplus value without needing any human labor to produce that value” (Sadowski, 2025: 122). Through this machine, capital can finally eliminate the power of labor as both a mortal threat to the supremacy of capital and a vital component to the production of value. I discuss this idea in the context of artificial intelligence as the latest—perhaps even greatest—attempt at creating a technological system that can at least appear to operate like a perpetual value machine.
However, Akbari's analysis helps us see that we do not have to be so literal when discussing machinery and its performance of inhuman value production. The “centuries of colonial violence” is also a long history of turning people into machines; the productive forces of treating humans like chattel and capital through evolving systems of colonial enclosure, imperialist expansion, and inhuman extraction—inhuman meant here as cruel, as cold, as non-human. We can expand Akbari's own historical purview—and her analysis of datafication, objectification, territorialization, and dispossession—by tracing the direct lineages of perpetual value capture further back.
We could start with the plantation as a form of colonial automation in which slaves are scientifically managed as mechanic capital (Browne, 2015). Indeed, manuals for operating sugar plantations described “abstracted visions of the plantation as a mechanical device under the control of its operator” (Whittaker, 2023: 16). Historical work by Meredith Whittaker (2023) details how British industrialists like Charles Babbage, inventor of the first mechanical computer, were directly influenced by the operation of plantations in their designs of automated machinery, management techniques, and surveillance systems that were intended to discipline labor and accelerate production in industrial factories.
We can trace these logics and influences of colonial capitalist extraction onward through subsequent changes to the social relations and productive forces that underpin different modes of value extraction. Leading, of course, to our contemporary moment dominated by the imperial dreams and endless frontiers that motivate the AI industry. A more robust historical materialist analysis offers a way to expand Akabari's conceptual work and strengthen her argument that digital colonialism is not just a metaphor of the past overlaid on the present, but is instead the latest development in capitalist innovations for perpetual value capture.
Conclusion
Akbari (2026) makes a welcome intervention into bodies of work on “capitalist transformations” that are “rich and extensive”—but also divergent and disconnected. Uneven datafication strikes me as one of those concepts that, upon hearing it sounds so obvious and intuitive that it's amazing nobody coined the term previously. Fortunately, Akbari does more than just mint a concept; she also goes a long way to fleshing out its many features and connections across a large literature on—and long history of—colonialism and its many styles of violent extraction. It's a tragic fact that the practices of colonial capitalism are constantly renewing themselves, expanding into new domains, and shape shifting in response to—and as drivers of—developments in the social relations and productive forces that structure our world. The concept of uneven datafication offers solid ground for advancing the kind of critical work needed to understand these developments, and perhaps even hinder their progress.
Footnotes
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
