How Societies Select Their Social Sciences

In 2026, China renewed a decade-long effort to construct what it now calls an autonomous knowledge system in philosophy and the social sciences. The language and institutional setting are specific to China, but the underlying question is broader: when a society concludes that inherited systems of knowledge no longer adequately explain its circumstances, should the organization of knowledge itself change?^[1]

China provides a catalyst rather than an answer. Its contemporary project makes explicit something that normally occurs through dispersed institutional change. Social-science disciplines are historical institutions. They developed under particular social conditions, were supported by governments, universities, foundations, professions, journals, and intellectual communities, and eventually acquired the methods, credentials, career structures, and standards through which they reproduce themselves.

If society changes, the knowledge it values may change with it. But intellectual supply does not respond mechanically to social demand, and institutions do not select from a menu generated independently of them. The two co-evolve. Existing selection systems shape what scholars have incentives and resources to invent; new intellectual possibilities can subsequently alter what institutions recognize as valuable.

Artificial intelligence now provides a contemporary test of that process.

Disciplines as historical settlements

Modern social science emerged during a transformation of social life. Industrialization, capitalism, urbanization, bureaucratic government, democratization, migration, and new forms of class organization created questions that older combinations of political philosophy, history, political economy, jurisprudence, and statistics could no longer comfortably contain.

The thinkers later identified as founders of separate disciplines frequently crossed the boundaries their successors inherited. Marx moved among political economy, philosophy, history, and social theory. Weber studied markets, religion, authority, bureaucracy, law, and historical development. Durkheim worked to establish sociology as a separate science while addressing problems touching morality, religion, psychology, education, and political organization.

The later division of this terrain into specialized disciplines produced enormous gains. Specialization allows intellectual capital to accumulate. Economists can develop sophisticated techniques without reconstructing psychology or political philosophy for every problem. Psychology can build experimental traditions suited to cognition and behavior. Sociology can accumulate distinctive theories and methods for social structure, organizations, institutions, networks, and status.

Disciplines became technologies for organizing intellectual labor.

Their borders, however, are historical institutions rather than natural features of society. Ross (2003) describes the recognized twentieth-century social sciences as emerging through the separation and negotiation of previously overlapping bodies of knowledge. Over time, a successful institutional solution to the organization of inquiry came to resemble the organization of reality itself.

That distinction matters whenever reality begins changing faster than the intellectual institutions built to study it.

Knowledge regimes co-evolve with intellectual supply

It is tempting to explain the development of the social sciences by saying that societies encounter new problems and create the knowledge required to solve them. That is too deterministic.

Intellectual innovation has dynamics of its own. Scholars encounter contradictions in existing theories. New mathematical techniques appear. Better datasets make previously inaccessible questions tractable. Researchers import ideas from neighboring fields. Methodological dissatisfaction can reorganize a discipline even when no government, market, or social movement has requested the innovation.

The credibility revolution in empirical economics is an important example. Better research design, quasi-experimental methods, improved data, and changing professional standards altered what applied economists regarded as convincing evidence. The transformation arose substantially from dissatisfaction within economics with existing empirical practice and subsequently changed which questions economists could credibly answer (Angrist and Pischke 2010). External society did not first request a credibility revolution and then receive one.

Yet it would be equally misleading to treat intellectual supply as independent of the institutional environment selecting among ideas.

A discipline that rewards a particular form of evidence will attract students trained to produce it. Journals that privilege certain methods make investment in those methods more valuable. Funding systems generate research infrastructure around some problems while leaving others thinly supplied. Successful programs create graduate courses, professional networks, data resources, and conceptual vocabularies from which the next generation of innovation emerges.

The relationship is therefore one of co-evolution under institutional selection.

Governments, universities, foundations, firms, journals, professional associations, political movements, and funding institutions select among intellectual possibilities. Those selections also reshape the environment in which future possibilities are generated.

A stylized sequence is recursive: changing social conditions and existing intellectual possibilities affect institutional selection; selection expands particular methods, questions, and research styles; those choices alter incentives and intellectual capabilities; new possibilities emerge; institutions select again.

This process can fail at several points. A society can urgently need knowledge that its intellectual institutions never generate. A well-funded research agenda can prove sterile. A methodological innovation can transform a field before outsiders understand its importance. An intellectual tradition can remain institutionally dominant after the social conditions that once favored it have changed.

Knowledge regimes are therefore neither passive reflections of society nor autonomous intellectual machines. They are evolving systems in which social conditions, institutional incentives, and intellectual innovation continually reshape one another.

Different environments sustain different mixtures

Continental Europe and the United States share much of the same intellectual ancestry. Neither can sensibly be reduced to theory on one side and evidence on the other. Their institutional histories nevertheless sustained different mixtures of social inquiry.

European traditions preserved substantial space for scholarship crossing sociology, philosophy, political thought, history, and cultural criticism. Questions about capitalism, rationalization, modernity, alienation, social order, and historical transformation remained legitimate theoretical objects even when they did not immediately resolve into bounded empirical relationships.

Hartmut Rosa’s work on social acceleration illustrates the value of that space. Its first move is conceptual. It asks whether technological acceleration, accelerating social change, and the acceleration of everyday life constitute a structural feature of modernity. Once such an object has been identified, it can be measured, criticized, refined, or rejected. Concept formation and empirical adjudication perform different intellectual functions.

American social science developed in a more strongly professionalized disciplinary environment. Universities, foundations, reform movements, public agencies, policy institutions, and increasingly specialized journals rewarded forms of social knowledge capable of being decomposed, measured, compared, and subjected to formal evidentiary standards. Ross (1991) documents the attraction of natural-science models in the development of American social science, while Fourcade (2009) shows that even economics developed different professional identities and institutional relationships in the United States, Britain, and France. Wagner et al. (1991) similarly document the interaction between social-scientific knowledge and modern policy institutions.

These are differences of institutional mixture, not national intellectual essence. Knowledge moves internationally, but institutions filter, reward, and reproduce it locally.

The resulting equilibria also shape future intellectual supply. Once an empirical research infrastructure becomes large, innovations that improve empirical research become more valuable. Where broad theoretical inquiry retains institutional legitimacy, scholars can continue generating concepts that would be harder to sustain inside systems rewarding narrower contributions.

The knowledge regime and its intellectual products gradually reproduce one another.

East Asia and the localization problem

East Asian experience provides a compressed version of this process. Modern social-science systems were frequently imported, reconstructed, and localized while the societies using them underwent rapid transformation.

There is no single East Asian model, and the trajectories of Japan, South Korea, Taiwan, and mainland China should not be collapsed into one story. The recurring problem is narrower: theories and categories developed from European and American historical experience were repeatedly applied to societies following different sequences of industrialization, state-building, family transformation, urbanization, political development, and economic growth.

That generated debates about localization. Korean sociology, for example, explicitly wrestled with the relationship between Western theory and Korean reality (Park and Chang 1999). Comparable debates appeared elsewhere in the region.

The contemporary Chinese case is analytically distinct because the attempt to modify the knowledge regime has become an explicit institutional project. Sociology itself had been removed from the Chinese university system during the 1952 reorganization and began to be restored after the late 1970s. In 1983, Lucie Cheng and Alvin So described that reconstruction as moving “toward the Sinification of Marxian sociology,” foreshadowing a problem that would reappear decades later in broader form.

China’s current autonomous-knowledge-system initiative does not matter here because it provides a template for other countries. It matters because it makes deliberate selection of the knowledge architecture itself visible. A process that normally emerges through dispersed universities, professions, funders, and journals has become an explicit national intellectual objective.^[1]

That visibility brings another property of knowledge regimes into focus. Selecting what knowledge should flourish is one problem. Determining how selected knowledge can be shown to be wrong is another.

The correction channel

Every knowledge regime contains institutional pressures. Governments direct resources, foundations choose priorities, journals determine what counts as publishable evidence, universities decide whom to hire, professional associations define prestige, and markets reward some forms of knowledge more generously than others.

There is little analytical value in contrasting politically influenced knowledge with an imaginary science untouched by institutions. The more useful question is where the correction channel lies.

Suppose an institution, discipline, or intellectual community strongly favors a particular explanation, method, or research direction. What must exist for a competing account to correct it?

Four properties matter. A challenger first needs standing: legitimate access to the arenas in which the claim is evaluated. Criticism that cannot enter relevant journals, professional debates, funding competitions, or institutional decisions has little corrective force. The challenger also needs some degree of independence from the institution whose premise is being challenged. Alternative sources of intellectual authority, employment, funding, publication, or reputation make costly disagreement possible.

Correction further requires capacity. Critics need access to data, methods, personnel, time, and the empirical object itself. Formal permission to disagree has limited value when no one can assemble the evidence necessary to make disagreement persuasive. Finally, correction requires consequence. A successful challenge must be capable of changing something: publication standards, curricula, funding, professional reputation, institutional practice, or prevailing belief.

These properties lie on continua.

Decentralized systems can fail badly. Elite departments, top journals, concentrated funding, citation cascades, or shared ideological assumptions can narrow effective correction without formal central control. A centralized knowledge regime can contain competing universities, intellectual factions, empirical pressures, and demands for international credibility that create corrective capacity within the system.

The relevant comparative question is therefore not whether a system is centralized or decentralized in the abstract. It is: who has standing to challenge an important premise, how independent and capable is that challenger, and what happens if the challenge succeeds?

There is also a temporal dimension. A correction mechanism that eventually identifies error may still perform poorly when the object being studied changes faster than the correction cycle. That possibility becomes increasingly relevant under rapid technological change, though it requires separate empirical treatment.

The correction channel helps explain why established disciplines possess an advantage when new problems arrive.

Why disciplines are difficult to displace

Specialization does more than lower the cost of producing knowledge. It lowers the cost of recognizing familiar forms of error.

Economics has developed conventions for distinguishing persuasive from weak identification. Experimental psychology has standards governing design, measurement, and replication. Historical research has techniques for evaluating sources and competing accounts. Sociology has accumulated methods for connecting individual observations to social structures and institutions.

None of these correction systems is complete. Their value lies in repeated use. Researchers learn what a weak claim in their field looks like because generations of earlier researchers have already made related mistakes.

This gives established disciplines an epistemic advantage as well as an institutional one.

A newly assembled research structure combining economics, psychology, sociology, and technology studies may see more of a connected phenomenon. It does not automatically possess a common procedure for deciding when the combined explanation is wrong. Errors can occur inside each component, but they can also arise at the interfaces: an economic response can be incorrectly connected to a psychological mechanism, or an organizational finding can be transferred into an individual-level claim without sufficient justification.

Recombination therefore must construct correction mechanisms while integrating knowledge.

Disciplines are not only cheaper places to specialize. They are cheaper places to be wrong in ways that trained peers know how to detect.

This principle applies to deliberate knowledge-system construction as well. China’s autonomous-knowledge-system project, to the extent that it seeks intellectual forms departing from inherited arrangements, must develop correction mechanisms appropriate to those forms. New coordination does not itself create new error-correction capacity.

Institutional persistence therefore reflects accumulated epistemic infrastructure as well as organizational inertia.

AI enters the existing system: absorption as the historical baseline

Artificial intelligence operates simultaneously as a research technology and as a source of social transformation. Its first institutional effect is already visible: existing disciplines are absorbing it.

This should be the historical baseline.

Major research technologies have repeatedly altered disciplines without eliminating their boundaries. Computing transformed empirical work. New econometric methods changed economics. Imaging changed psychology and neuroscience. Large administrative datasets altered research across multiple fields while leaving professional jurisdictions recognizable.

The early AI pattern resembles this history.

The University of Chicago’s AI in Social Science Conference provides one limited but useful observation because it is intentionally cross-disciplinary. Its 2024 program included The Market Effects of Algorithms, British Industrialization and Cultural Change: Evidence from the Use of Proverbs, and Journalist Ideology and the Production of News: Evidence from Movers. These are recognizable social-science problems in which algorithms, language models, or computational analysis expand measurement or research capability. The conference brought together participants from economics, sociology, law, behavioral science, and AI.^[2]

The September 25–26, 2025 meeting retained the same broad conception of how AI might change social science and how social science might shape AI while presenting work ranging from AI-enabled hypothesis generation and causal inference on text to established questions in poverty, policy, education, and political economy. The October 8–9, 2026 edition has been announced using closely similar language about methods, datasets, technologies, and cross-field exchange; because that meeting has not yet occurred, only its announced institutional framing is observable at the time of writing.^[2]

One conference series cannot establish a general pattern. It does illustrate what disciplinary absorption looks like.

Under that pathway, economics becomes better at AI-assisted economics, sociology at AI-assisted sociology, psychology at AI-assisted psychology, and political science at AI-assisted political science. New intellectual supply generated by AI evolves inside existing professional selection systems, while those systems gradually adjust to reward new forms of work.

Given the history of specialization and the correction advantage disciplines already possess, absorption is the base-rate expectation.

When should recombination have an advantage?

The alternative is not interdisciplinarity for its own sake.

Many broad subjects can attract scholars from several disciplines without generating much intellectual integration. Researchers may attend the same conference while studying separable components of a topic. The umbrella becomes interdisciplinary while the explanatory work remains multidisciplinary.

Genuine recombination should be more valuable under a stronger condition: interface dependence.

Interface dependence exists when explaining a phenomenon requires resolving causal connections that lie between established disciplines. One component cannot be treated as a self-contained black box without materially weakening the explanation.

Consider a professional worker adapting to AI. A technological capability changes the productivity and market value of a skill. Firms reorganize tasks. Occupational definitions and expectations of competence change. Educational institutions adjust. Social comparison shifts as adoption diffuses unevenly. Individuals interpret those changes using expectations about careers, status, and achievement formed under earlier technological conditions. Psychological responses affect later choices about work, training, consumption, family, and risk.

Economics, organizational research, sociology, psychology, life-course research, and technology studies each contain part of this chain.

If the research question concerns the wage effect of AI, disciplinary absorption may be fully adequate. If the question concerns how changing technological capability travels through markets and institutions into altered expectations and human behavior, explanatory errors may occur precisely where one discipline hands the mechanism to another.

That is where problem-centered recombination could generate a return large enough to offset its additional coordination and correction costs.

The relevant comparison is therefore not disciplinary research versus interdisciplinary research. It is: how much explanatory value is lost when the interfaces are left unresolved?

The greater that loss, the stronger the potential return to recombination.

Observing interface dependence before institutions change

Interface dependence cannot serve as a predictor if it is inferred only after successful integration occurs. It must leave an observable trace beforehand.

The most useful trace is handoff failure.

A disciplinary model may perform well within the domain for which it was constructed and then lose explanatory or predictive power when the outcome turns on a variable that the model treats as fixed but a neighboring field treats as endogenous.

The worker-adjustment example illustrates the point. A labor-economic model may generate a coherent prediction from changes in wages, productivity, skill prices, and retraining returns. Yet observed adaptation may depart systematically from that prediction when expectations about competence, career timing, identity, or attainable progress themselves change with exposure to AI. The economic model need not be defective within its domain. The failure occurs at the handoff from economic incentives to expectation formation and behavioral response.

Such interface dependence can leave several traces in an existing literature before any interdisciplinary institution forms.

One is systematic handoff residuals: a model predicts well until outcomes become sensitive to a neighboring mechanism, at which point errors acquire structure rather than remaining noise.

Another is competing closure. Two disciplines can model adjacent parts of the same process while closing the causal chain differently. A variable treated as an exogenous input in one field is treated as an endogenous outcome in another. The two approaches may consequently generate incompatible predictions even though each is internally coherent.

A third trace is repeated patching. A literature may repeatedly add proxies, reduced-form controls, behavioral terms, fixed effects, or residual categories to capture a mechanism that another discipline studies directly. Such patches can be appropriate for a bounded research question. Their repeated importance is nevertheless evidence that the disciplinary boundary itself may be carrying explanatory cost.

These traces allow interface dependence to be investigated before observing whether institutions later recombine. Researchers can search existing literatures for recurring prediction failures at handoffs, incompatible assumptions about which variables are endogenous, and persistent reduced-form treatment of mechanisms modeled explicitly elsewhere.

The logic of the test therefore runs forward rather than backward:

observable handoff failure → potential return to integration → institutional response → intellectual integration, absorption, shell formation, or re-specialization

Recombination no longer supplies the evidence that interface dependence existed. Interface dependence can be diagnosed before the institutional outcome is known.

Cognitive science and the danger of confusing institutions with integration

Cognitive science provides a useful caution because it demonstrates how easily organizational evidence can be mistaken for intellectual integration.

The field brought together psychology, linguistics, computer science, neuroscience, philosophy, anthropology, and related traditions around the study of cognition. It acquired journals, degree programs, faculty positions, professional societies, and academic departments.

By conventional institutional indicators, this looks like successful recombination.

Núñez et al. (2019) challenged that interpretation. Examining two bibliometric and two socio-institutional indicators, they concluded that the original multidisciplinary program had not developed into the coherent interdisciplinary field envisioned by its founders. They found cognitive science largely subsumed bibliometrically by cognitive psychology and reported substantial lack of curricular consensus among degree-granting institutions.

That interpretation was contested. Topics in Cognitive Science published a cluster of responses, including McShane et al. (2019), who argued that Núñez and colleagues’ evaluation criteria mischaracterized the health of the field. The dispute need not be resolved here.

Its methodological lesson is enough: institutionalization and intellectual integration are separate variables.

A field can acquire departments, faculty lines, degrees, journals, and careers without producing sustained integration across its parent traditions. Institutional structures can form a shell around intellectual activity that remains fragmented or becomes dominated by one parent field.

The case also suggests a possible explanation. A broad object such as “cognition” can sustain many adjacent research programs without forcing them to resolve one another’s mechanisms. Where interface dependence is weak, organizational proximity may not create strong intellectual integration.

That possibility is directly relevant to AI. “AI and society” is an extremely broad label. It can support dozens of valuable research programs while creating little need for a common explanatory architecture. Narrower phenomena whose causal chains repeatedly fail at disciplinary handoffs provide a more demanding test.

Four observable states

The development of new knowledge structures is better understood dynamically than as a set of mutually exclusive endpoints.

Disciplinary absorption occurs when existing fields incorporate new methods or substantive problems while retaining their intellectual centers, professional standards, and correction mechanisms.

Institutional shell formation occurs when centers, departments, programs, conferences, or journals acquire interdisciplinary labels without corresponding integration of theories, evidence, or intellectual traffic.

Problem-centered integration occurs when a phenomenon generates sustained explanatory exchange among multiple fields because important mechanisms cannot be adequately decomposed.

Re-specialization occurs when successful integration develops a stable conceptual core, standardized training, professional identity, its own correction machinery, and increasingly recognizable boundaries. Recombination has generated a new specialization.

These states can form paths rather than terminal categories. A center may begin as a shell and gradually integrate. Problem-centered integration may remain porous for decades. Successful recombination may eventually specialize and become another discipline. A nominally interdisciplinary structure may instead be absorbed intellectually by one parent field.

The historical process can therefore reproduce disciplinary boundaries even when it begins by trying to cross them.

How would we know?

Institutional indicators remain important because they reveal whether a new intellectual structure has acquired the capacity to reproduce itself.

Funding patterns show where resources are being directed. Hiring reveals whether organizations possess independent authority over careers or borrow faculty from established departments. Publication structures indicate where professional reputation is generated. Degree programs and curricula reveal what knowledge institutions believe must be transmitted. Promotion and tenure show which intellectual standards ultimately determine professional survival.

The effective boundary of a discipline is not where scholars can attend a conference. It is where they can be hired, published, credentialed, and promoted.

Cognitive science demonstrates that these indicators are not sufficient. They must be paired with evidence about what occurs intellectually inside the institutions.

One dimension is intellectual traffic. Do contributing disciplines cite, challenge, and build on one another, or does most research remain within parent-field networks? Bibliometric data can provide an imperfect measure.

A second is conceptual and curricular convergence. Programs need not teach identical material, but durable integration should eventually generate some shared understanding of the problems, mechanisms, methods, and evidentiary standards that participants are expected to know.

A third is boundary permeability. When the underlying phenomenon changes, can the structure incorporate a previously peripheral discipline or new method because the explanation requires it? Or does the institutionalized mixture begin defending itself as the definition of the field?

These indicators should be interpreted alongside the ex ante evidence of interface dependence. Handoff failures tell us where integration might have intellectual value. Institutional and bibliometric evidence then tells us how the knowledge system responds.

No numerical threshold separating integration from coexistence is imposed here. Determining such thresholds is an empirical task rather than something that should be stipulated conceptually.

A longitudinal research program could track defined AI-related problem areas using citation flows, faculty appointment networks, funding categories, publication venues, graduate curricula, career outcomes, and documented handoff failures in the underlying literatures. The objective would not be to count interdisciplinary labels. It would be to observe whether the production and correction of knowledge itself begins crossing established boundaries.

What the present evidence permits us to say

Current evidence is preliminary.

The most defensible observation is that disciplinary absorption dominates the visible core of social science, while interdisciplinary activity is expanding around institutes, conferences, collaborative projects, and other structures at the institutional periphery.

The Chicago conference series is consistent with that pattern, not proof of it. It combines participation across fields with research that remains largely interpretable through established substantive and methodological traditions. The announced 2026 event also shows continuity in institutional framing, but its October program remains in the future at the time of writing.^[2]

There are good reasons for absorption to remain dominant. Disciplines possess accumulated knowledge, training pipelines, publication systems, professional identities, and correction channels. AI makes those disciplines more capable while creating new substantive questions for them to study.

Whether peripheral recombination becomes more than organizational proximity depends in part on interface dependence.

If AI-generated social phenomena can be decomposed into components that existing disciplines handle adequately, absorption may remain efficient. Cross-disciplinary conferences and centers can coordinate communication without altering the deeper architecture of knowledge.

If important phenomena repeatedly generate prediction failures at disciplinary handoffs because mechanisms treated as fixed in one field are endogenous in another, the return to genuine integration rises.

And if such integration succeeds, it confronts a further historical choice: remain organized around a changing phenomenon, or institutionalize stable boundaries and become another specialization.

The present essay does not determine which path will dominate. It specifies what would distinguish them.

The AI test

The history of social science gives no reason to regard the current disciplinary architecture as either permanent or obsolete.

Disciplines developed because specialization solved real intellectual problems. They accumulated methods, concepts, standards, and communities capable of recognizing familiar errors. Their persistence reflects productive and corrective capital as well as organizational inertia.

Knowledge regimes nevertheless co-evolve with the societies and intellectual technologies around them. Institutional selection shapes which forms of intellectual innovation become attractive, while innovations can alter the institutions doing the selecting.

China’s contemporary autonomous-knowledge-system project makes one form of deliberate realignment visible. Its significance here is not that it provides a model for other societies, but that it demonstrates how explicitly a knowledge architecture can become an object of institutional policy. The same case also reminds us that reconstructing a knowledge system requires attention to the correction mechanisms through which the new structure discovers its own errors.

AI poses a different experiment. It has already entered established disciplines as a research technology, where absorption is proceeding rapidly. At the same time, it is changing social reality across markets, organizations, institutions, information systems, professional identities, and individual experience.

The hinge is interface dependence, and interface dependence can be observed before the institutional response. Recurrent prediction failures at disciplinary handoffs, conflicting assumptions about which variables are endogenous, and repeated reduced-form patching of mechanisms developed elsewhere provide evidence that disciplinary decomposition is becoming costly.

The subsequent institutional outcomes can then be distinguished.

If AI-related inquiry continues to reproduce itself primarily through established departments, journals, curricula, professional standards, and disciplinary intellectual networks, disciplinary absorption remains dominant.

If interdisciplinary organizations proliferate while theories, citations, curricula, and evidentiary practices remain fragmented or dominated by parent fields, institutional shell formation has occurred.

If research organized around defined phenomena generates sustained intellectual exchange across fields, resolves mechanisms at their interfaces, develops credible cross-field correction procedures, and remains open to changing disciplinary inputs, problem-centered integration has emerged.

If that integrated structure subsequently develops standardized training, distinctive professional standards, stable conceptual boundaries, and its own self-reproducing correction machinery, a new specialization has formed.

These are not predictions about what should happen. They are observable alternatives.

The historical record supplies the mechanism. Handoff failures identify where integration may have value. Institutions and intellectual traffic reveal the response.

That is the AI test.

---

Notes

1. China as catalyst. On May 17, 2016, Xi Jinping called for accelerating development of philosophy and social sciences with Chinese characteristics. During an April 25, 2022 visit to Renmin University, he stated that accelerating that project ultimately meant constructing China’s own “autonomous knowledge system.” In May 2026, around the tenth anniversary of the 2016 “5·17” speech, official and institutional commentary again centered the autonomous-knowledge-system formulation. On May 17, 2026, Xinhua reported a new instruction calling for faster construction of the autonomous philosophy and social-science knowledge system and for better answers to the questions of China, the world, the people, and the times. The Fifteenth Five-Year Plan outline, published in March 2026, also calls for accelerating construction of that system. These sources document institutionalization of a knowledge-system objective; they do not establish the scientific validity of its output. See Xi’s May 17, 2016 speech; People’s Daily discussion of the April 25, 2022 Renmin formulation; Xinhua, May 17, 2026; and the Fifteenth Five-Year Plan outline. Back 1 Back 2
2. Chicago AI in Social Science Conference. The 2024 conference was held September 19–20, 2024, and the 2025 conference September 25–26, 2025. Both framed the event around AI’s effects on social science, social science’s effects on AI use, and cross-disciplinary work involving new methods, data, and technologies. The 2026 edition is scheduled for October 8–9, 2026. Because this essay is dated August 2026, the 2026 event is used only as evidence of continuity in announced institutional framing; its program and outcomes are not treated as observed evidence. See the 2024 conference, 2025 conference, and 2026 announcement. Back 1 Back 2 Back 3

References

Angrist, Joshua D., and Jörn-Steffen Pischke. 2010. “The Credibility Revolution in Empirical Economics: How Better Research Design Is Taking the Con out of Econometrics.” Journal of Economic Perspectives 24(2): 3–30.

Cheng, Lucie, and Alvin Y. So. 1983. “The Reestablishment of Sociology in the PRC: Toward the Sinification of Marxian Sociology.” Annual Review of Sociology 9: 471–498.

Fourcade, Marion. 2009. Economists and Societies: Discipline and Profession in the United States, Britain, and France, 1890s to 1990s. Princeton: Princeton University Press.

McShane, Marjorie, Selmer Bringsjord, James Hendler, Sergei Nirenburg, and Ron Sun. 2019. “A Response to Núñez et al.’s (2019) ‘What Happened to Cognitive Science?’” Topics in Cognitive Science 11(4): 914–917.

Núñez, Rafael, Michael Allen, Richard Gao, Carson Miller Rigoli, Josephine Relaford-Doyle, and Arturs Semenuks. 2019. “What Happened to Cognitive Science?” Nature Human Behaviour 3(8): 782–791.

Park, Myoung-Kyu, and Kyung-Sup Chang. 1999. “Sociology Between Western Theory and Korean Reality: Accommodation, Tension and a Search for Alternatives.” International Sociology 14(2): 139–156.

Rosa, Hartmut. 2013. Social Acceleration: A New Theory of Modernity. New York: Columbia University Press.

Ross, Dorothy. 1991. The Origins of American Social Science. Cambridge: Cambridge University Press.

Ross, Dorothy. 2003. “Changing Contours of the Social Science Disciplines.” In The Cambridge History of Science, Volume 7: The Modern Social Sciences, edited by Theodore M. Porter and Dorothy Ross. Cambridge: Cambridge University Press.

Wagner, Peter, Carol H. Weiss, Björn Wittrock, and Hellmut Wollmann, eds. 1991. Social Sciences and Modern States: National Experiences and Theoretical Crossroads. Cambridge: Cambridge University Press.