Extending the standards: AI,global sociology,and North–South collaboration

Aksoy et al. (2026) offer a welcome and timely statement. We write not to dispute its three principles but to widen their frame. The standards debate, as the statement presents it, is read largely through the lens of European and North American sociology, and we argue that two developments, the changing geography of sociological knowledge and the disciplinary reach of artificial intelligence (AI), together make a wider frame necessary, and that the two are best addressed at once.

The first development concerns where sociological knowledge is produced. Production remains heavily concentrated in the Global North, and theories and datasets generated there tend to travel as if they were universal, while large parts of the Global South remain comparatively under-studied and under-resourced (Connell, 2007; Mosbah-Natanson and Gingras, 2014). This is a deficit in the discipline’s coverage of the social world. It is also, we suggest, an opportunity. A statement on standards written at a European jubilee could read the geography of the field as a settled background. We think it is better read as an open question about whose puzzles the discipline chooses to treat as central.

The second development is the arrival of AI in social research, and here our difference with the statement is one of emphasis. Aksoy et al. frame rapid changes in information technology, massive digital data, and large language models as challenges to European sociology and related disciplines (Lazer et al., 2021). Read in the context of the European Academy of Sociology (EAS) jubilee, a European framing is natural. We would argue, however, that the substantive challenge is general. Large-scale computation and generative models reshape what counts as feasible research everywhere at once, in Oslo and Oxford no less than in Nairobi or São Paulo, and a vision of standards that locates the challenge in one region risks understating how general it is.

What, then, can AI do, and what can it not? It is already capable of performing parts of the research pipeline, large-scale data processing, the classification of text and images, and the detection of patterns across corpora that no single team could read. It does not yet do the explanatory theory building that the statement rightly prizes, the work of linking macro-conditions to outcomes through specified mechanisms. Nor does it reliably do the contextual interpretation that good qualitative research requires, or the prior judgement of which empirical puzzles matter. The risk, as Aksoy et al. note in their discussion of prediction and explanation, is that classification crowds out understanding. The opportunity is that once routine analytic tasks become cheaper, researchers are freed to spend their effort where machines cannot substitute for them.

This brings us to the unevenness of coverage. Many societal issues remain under-studied, not because they are unimportant but because research capacity is limited and data infrastructure is thin across much of the world (Hilbert, 2016; Schroeder, 2018). The result is a large and uneven frontier for both quantitative and qualitative sociology, where basic description is often still missing and where theory built elsewhere has rarely been tested. Cheaper computation lowers one barrier to working on this frontier. It does not remove the others, since the data, the local knowledge, and the methodological capacity to use AI tools well remain unequally distributed.

We therefore propose a reorientation that the statement’s logic already invites, namely deeper and more genuinely reciprocal collaboration with researchers in the Global South, oriented toward questions that matter there, including those motivated by intellectual curiosity rather than only by policy salience. Such a reorientation does three things at once. First, it mitigates the AI shock by opening research terrain that predictive tools cannot exhaust, since explanation, measurement, and theory in under-studied settings are precisely the tasks machines cannot yet perform. Second, it begins to correct the empirical and theoretical deficit, extending and stress-testing sociological theory against a far wider range of cases. Third, it enhances knowledge diffusion and helps close the methodological gap in the use of AI for social research (Mökander and Schroeder, 2022), so that new methods spread across the field rather than widening the distance between core and periphery. They are one move with three payoffs.

It might be objected that this dilutes the standards the statement defends, trading rigour for breadth. The opposite holds. On the analytical view that Aksoy et al. articulate, where explanation runs through mechanisms linking macro-conditions to outcomes, the generalizability of any proposed mechanism is itself an empirical question, and extending tests across a wider range of social settings is therefore not a concession to inclusivity but a precondition for the kind of theory the statement defends; a mechanism that holds only where it was first observed is a local regularity, not yet a mechanism. The same demanding requirements should govern work on Global South questions as much as on Northern ones. Open science is the natural vehicle here, since shared data, code, and materials are what allow distributed and unequally resourced teams to collaborate and replicate one another’s work.

In sum, the geography of the discipline and the arrival of AI are usually discussed apart. Taken together, they point to a single response. A sociology that extends its standards to a wider world, and that does so through reciprocal collaboration rather than the export of Northern templates, is better placed to absorb the AI shock, to fill its own empirical deficit, and to remain credible when it speaks on matters of public importance. EAS, with its founding aim of promoting integration and excellence, is well positioned to lead such a turn. The standards are the right ones. The question is how far they are allowed to travel.