AI,Education,and Academic Publishing in between Knowledge Capitalism and Socialism

Abstract

This paper explores the intersection of artificial intelligence (AI), higher education, and academic publishing within the framework of knowledge socialism, a concept promoting a truly free and equitable exchange of ideas based on collective ownership of the means of knowledge creation and equality in the social relations of knowledge production. Knowledge socialism, rooted in the Marxist critique of capitalism, is to be understood vis-à-vis knowledge capitalism featuring privatised regulation of the production and circulation of knowledge and the transformation of knowledge into exchange values in the marketplace. While the two systems appear to represent mutually exclusive orientations, they are both fluid, always in motion, and meet each other in a reality of in-betweenness. With an understanding of the knowledge capitalism/socialism problematics, this paper then discusses how AI technologies disrupt existing paradigms of academic publishing and scholarship, enabling decentralised knowledge creation, open publishing, and democratised learning. Meanwhile, challenges such as algorithmic bias, corporate acquisition, and ethical issues threaten to undermine the progressive potential of AI. This paper argues that AI's capacity to tip the balance between knowledge socialism and capitalism is hardly questionable. However, such power needs to be ethically guided and directed by consensual efforts prioritising equity, transparency, and openness if it is to catalyse knowledge socialism.

Introduction

The twenty-first century has witnessed unprecedented technological developments, with artificial intelligence (AI) emerging as a transformative force in education and academia. The Internet, advancements in digital technologies, and the worldwide pluralisation of science capacity have enabled a space of global science (Marginson, 2022a). On one side, neoliberal globalisation has created an almost seamless global market terrain where knowledge is commodified and traded across national borders, and nation-states, universities, and individual researchers are caught up in competition over profitability and efficiency (Olssen & Peters, 2005). On the other hand, the concept of knowledge socialism, a critique of capitalist knowledge monopolies, has been gaining momentum, arguing for an open, shared, collaborative, and democratised intellectual landscape. In 2012, Peters, a pioneer scholar of knowledge socialism, noted that:

Whereas knowledge capitalism focuses on the economics of knowledge, emphasising human capital development, intellectual property regimes, and efficiency and profit maximization, knowledge socialism shifts emphasis towards recognition that knowledge and its value are ultimately rooted in social relations (Peters & Besley, 2006). Knowledge socialism promotes the sociality of knowledge by providing mechanisms for a truly free exchange of ideas. Unlike knowledge capitalism, which relies on exclusivity — and thus scarcity — to drive innovation, the socialist alternative recognises that exclusivity can also greatly limit innovation possibilities (Peters et al., 2012, p.88)

The tension between knowledge capitalism and socialism reflects deeper ontological, epistemological, moral and social issues. This paper takes academic journal publishing as a window to explore the knowledge capitalism/socialism problematics based on the understanding that journal articles are now the basic unit of research communication and are thereby close to the heart of science. “There are no boundaries, no walls, between the doing of science and the communication of it; communicating is the doing of science” (Montgomery, 2003, p.1). The following discussion addresses how AI reshapes higher education and academic publishing, evaluating its potential to tip the scales between knowledge capitalism and knowledge socialism.

Understanding Knowledge Capitalism and Knowledge Socialism

The concept of “knowledge economy”, influenced by Romer (1986) and Lucas’s (1988) endogenous growth theory and formulated and popularised by OECD's (1996) influential report, has become a powerful discourse shaping science and education policies and more broadly national systems of innovations and developments worldwide. Central to the knowledge economy is the emphasis on knowledge as a driving force for economic growth and the importance of investments in human talent. Theorists interested in the knowledge economy have drawn our attention to a crucial change within capitalism from material resources and physical labour as chief factors of production to brain power and knowledge as input and output, pointing to the substitution of a new historical stage for traditional industrial capitalism. Informed by post-industrialism that is characterised by the growing importance of immaterial labour and the service industries, Stehr (1994) noted that “While these (Marxian) features of labour and property certainly have not disappeared entirely, a new principle, ‘knowledge’, has been added, which, to an extent, challenges as well as transforms property and labor as the constitutive mechanism of society” (p.7). The new sociological significance of knowledge has also been picked up by Bell (1973), who claimed that “what comes to the fore is ‘intellectual technology’ (based on mathematics and linguistics), which uses algorithms (decision rules), programming (software), models and simulations, in the running of the new ‘high technology’ (p. xvii). The coming of the knowledge economy has been enabled by two parallel trends, namely, an upsurge in knowledge-related investments and activities, and technological progress that greatly facilitated the production and transfer of knowledge and information. As Foray (2004) suggested,

The collision between these two phenomena has spawned a unique economy, characterized essentially by (1) the accelerating (and unprecedented) speed at which knowledge is created and accumulated and, in all likelihood, at which it depreciates in terms of economic relevance and value as well as (2) a substantial decrease in the costs of codification, transmission, and acquisition of knowledge. This creates the potential for a massive growth of knowledge flows and externalities (p. x).

If the industrial society was largely reliant on unskilled and semi-skilled labour for manufacturing and on transportation—railways, highways, seaways—to make the exchange of materials and goods possible, society in the age of knowledge economy relies primarily on intellectual labour as the source of invention and value-creation and on communication infrastructure for the travelling of knowledge and information.

As the knowledge economy has been gaining popularity since its emergence in the 1990s in both the scholarly and policy fields, some critical voices have also been raised. One notable example is Keven Smith, who criticised OECD's (1996) vision of knowledge economies as “those which are directly based on the production, distribution and use of knowledge and information” (p.7) for being too broad and too vague. Smith identified four major perspectives on the changed implications of knowledge, contending that existing approaches to the knowledge economy in the literature suffered from a reluctance to “consider what knowledge is in epistemological or cognitive terms” (2002, p. 7). Smith acknowledged that our societies were increasingly structured around knowledge-related activities such as formal education and intellectual properties, but “it is not clear that they are either new, or that they represent some new role for knowledge” (2002, p.11). Although knowledge has taken on a new significance alongside the advancement of information and communication technologies (ICT), this does not “justify talking about a new mode of economic or social functioning” (2002, p.12). To some extent, Smith's view makes sense: from agriculturalism to industrialism and post-industrialism and across different social systems, the evolvement of human societies has consistently rested on the precondition that people know something and are able to put that knowledge into action (Simmel, 1906). Defined in this sense, knowledge is arguably an anthropological constant. Meanwhile, Quah (2001) showed that examples of the economic significance of knowledge can be found prior to the arrival of the knowledge economy, such as the use of machines to improve economic performance during the Industrial Revolution. The knowledge economy developed in the 1990s, as Quah (2001) saw it, is helpful in understanding the contemporary economic resonance of knowledge, but the idea can be more accurately and clearly framed as the “weightless economy” consisting of four defining features: ICT, especially the Internet, intellectual assets, digital databases, and biotechnology. These critical reflections represent efforts to carefully assess claims for the knowledge economy, and the concept is very much still contentious. That said, most Western countries and later other countries have been closely monitoring and exploring the economic potential of knowledge since the 1990s (Peters, 2001), and there is substantial momentum in the academic discourse and empirical observations to seriously consider the knowledge economy as a working hypothesis, if not a given reality.

Scientists and influential international agencies such as OECD and the World Bank have all highlighted the significance of higher education in the new knowledge economy for improving the cognitive abilities and adaptabilities of knowledge workers, producing research and scientific knowledge and driving innovation and development. “Old struggles of industrial labour versus capital give way to conceptualizations of a class of highly educated scientists that characterize an increasingly highly qualified university-based service-oriented economy” (Peters, 2022, p.9). At about the same time as the knowledge economy was being discussed and realised, neoliberalism, a grand narrative advocating for free markets, deregulation, privatisation, and individualism, has tapped into knowledge as a path to economic growth, captured the policy agendas of many countries, and fundamentally changed higher education.

At various levels and across many aspects of higher education, the market has become the ruling paradigm. In the name of individual choice and autonomy, neoliberalism frames individuals as entrepreneurial selves responsible for investing in their skills and abilities to remain employable, turning knowledge into a private asset that underlies positional advantage in the labour market (Peters, 2016). There is a new market in knowledge as items for sale, from formal schooling to higher education and lifelong learning, and intellectual property regimes incentivise knowledge producers with market-priced rewards. Students are now “clients” or “buyers”, while teachers are “service providers”. Much like traditional capitalistic companies, profitability now drives the modus operandi of neoliberalised academia (Mintz, 2021). Customer satisfaction and market share have become popular indicators of teaching and learning performance, and universities have followed corporate models to extract revenues from tuition fees and prioritise financially profitable research (e.g., patents and business partnerships). Moreover, metricised measurements regulate competitive distribution systems that assess, compare, and valorise research output (Lynch, 2015). Nation states are in an arms race in innovation (often measured by scientific papers and citation metrics), universities compete for funding and ranked positionings, and individual researchers struggle for job security and academic reputation. Alongside developments in ICT that shrank the world into a global village, the neoliberal knowledge economy has become globalised since the early 2000s and is closely monitored by government decision-makers, university managers, and academics to figure out where oneself, one's own institution, or one's own country is positioned in relation to other local, national or international competitors (Olssen, 2016).

Perhaps unsurprisingly, intense competition among neoliberal agents over resources, prestige, and power has led to ethically questionable behaviours. Kulczycki (2023) classified individual behaviours in an economised and metricised global science landscape into three categories: outright fraud or misconduct, gaming in order to maximise one's gains, and playing the game simply to survive the system and maintain the status quo. The first is legally and ethically unacceptable. The second is legal but often violates scientific ethos, whereas the third is relatively difficult to evaluate in moral terms. Despite their differences, all three categories deviate from the classical liberal account of knowledge as neutral and disinterested and truth as something worth pursuing in its own right—knowledge and science are subjected to exterior and often utilitarian purposes.

The neoliberal knowledge economy can be seen as a particular political economy, yet its power goes deeper into the way in which the assumptions of individuality and self-interest have influenced people's understandings of the world, knowledge, and the ethics of human action. Agents of the prevailing neoliberalism tend to downplay the role of macro-level regulation, instead believing that an interplay between individual choice, free competition, and market capitalist regulation will lead to an optimal collective outcome. The individualistic ontology is well exemplified by Neilson's (2020) observation that “They (neoliberal subjects) view society simply as the spontaneous aggregate effect of private owners/individuals going about their own business” (p.139). In the realm of education, neoliberal ideology displaced the neoclassical understanding that government policies and initiatives were central to education investment and infrastructure building, which can then boost economic development (Clarke, 2021). Epistemologically, the neoliberal knowledge economy promotes private ownership of the production and circulation of knowledge. Digital infrastructures and services that now support university teaching, learning and research have been mostly acquired and integrated into data analytics and management companies (Posada & Chen, 2018). The massively big publishers hold the majority of academic journals that collect and distribute scientific knowledge via the Internet (Larivière et al., 2015). Elsevier-Scopus and Clarivate Analytics-Web of Science are so widely used and recognised that they de facto define the global knowledge base (Waltman, 2016). Meanwhile, there has been growth in other companies that provide platforms of algorithmic engines and are looking to expedite the innovation cycle for institutions and nations. Knowledge governed and regulated by these commercial bodies is mostly put behind subscription paywalls and intellectual property protections, with limited sharing with the public or among the companies themselves due to market competition. Ethically, a neoliberal agent's rationality is ultimately centred on oneself. Collaboration and teamwork exist, but they are often driven by an aggregate of private individual interests rather than mutual solicitude and empathy (Lynch & Kalaitzake, 2020). My interests, my desires, and my profits outweigh communal concern and welfare.

Given what has been discussed so far, the integration of the knowledge economy and neoliberalism with its abstract and universalist beliefs in rationality, individuality and self-interest has arguably led to a form of capitalism focused on immaterial capital and labour and the role of knowledge in the accumulation and distribution of wealth. Knowledge capitalism as such has demonstrated three defining features: the economisation of knowledge (knowledge as both input and output is closely associated with economic growth), the privatisation of knowledge (private ownership of the means of knowledge production as well as knowledge products), and the commodification of knowledge (the transformation of knowledge into exchange values or items for sale in the marketplace).

While knowledge has largely been absorbed into a neoliberal capitalist system, it does not seem to quite follow the classic rules that have defined capitalistic property rights and exchange since Adam Smith, namely, rivalry, excludability, and transparency (Harris, 2001). As Peters (2020) neatly summarised,

1) knowledge is non-rivalrous: the stock of knowledge is not depleted by use, and it this sense knowledge is not consumable; sharing with others, use, reuse and modification may indeed add rather than deplete value; 2) knowledge is barely excludable: it is difficult to exclude users and to force them to become buyers; it is difficult, if not impossible, to restrict distribution of goods that can be reproduced with no or little cost; 3) knowledge is not transparent: knowledge requires some experience of it before one discovers whether it is worthwhile, relevant or suited to a particular purpose (p.2).

Knowledge does not behave like typical goods and services of capitalistic logic; in some sense, it wants to be free and public. Meanwhile, one of the noteworthy features of contemporary knowledge capitalism is the way knowledge production and exchange have been assimilated into the global expansion of cloud-based services and the development of many social networks that highlight mass participation and collaboration. There is a well-established argument for the democratising capacity of ICT to enable a new model of production and distribution based on community, collaboration, and sharing instead of hierarchy, competition and control (Atton, 2004; Benkler, 2006; Tapscott & Williams, 2007). The Internet now seemingly provides free access to countless knowledge and information services, while social media and other online platforms bring people together in a borderless, experiential and collective space that is changing the conditions of imagining, producing and circulating creative and intellectual work.

The potential of public goods inherent in knowledge and the tendency for decentralised peer-to-peer communication, peer-with-peer cooperation and sharing have inspired a group of scholars represented by Michael A. Peters to consider the possibility of an alternative intellectual landscape to knowledge capitalism. In 2004, Peters envisioned knowledge socialism as an international knowledge system characterised by collective ownership of the means of knowledge production and equal social relations in which knowledge-related activities are rooted. However, soon after knowledge socialism was proposed, the big companies were quick to exploit the ICT-enabled connectivity and made considerable profits from online platforms utilising social networking and user creation. Expansions of the new forms of participation and interactivity did not always translate into open and communal sharing, and equalised access to knowledge and information services (provided there has been equalisation to a degree) did not wipe out or even mitigate social inequalities of various kinds. Acknowledging that technological progress itself does not guarantee any particular social outcome, Peters and Jandrić (2018) argued that the coming online of knowledge socialism driven by communal steering of ICT entails a corresponding change in philosophical understanding of human nature from homo economicus, which is crucial for the development of “the self-interested individual of liberal political economy established by Adam Smith and David Ricardo” (p.82) to the figure of homo collaborans:

HC (human capital) based on the figure of homo economicus is committed to three assumptions that tend to run counter to the collective learning processes that characterize the digital environment. The assumption of individuality is counter posed by collective intelligence (Lévy, 2015; Peters, 2017; Peters et al., 2016), that can take different forms from collective awareness and consciousness to collective intelligence, responsibility and action. The assumption of rationality is contradicted in a networked environment as the ontological basis is contained in the relations between entities rather than any one self-sufficient entity that is rationally aware and transparent to itself. The network is a very different kind of epistemic set of relations rather than the individual knowing agent. Finally, the assumption of self-interest again tends to be offset or decentred by forms of collective responsibility (pp.342–343).

Peters’ knowledge socialism is opposite to knowledge capitalism on conceptual, material, organisational and technological fronts, but the two are not necessarily black-and-white in reality. On the one hand, Peters and Besley (2013) did not deny that knowledge capitalism can contribute to social development of some kind, but they were unsure whether it will always lead to just, democratic outcomes; and the authors referred to knowledge socialism to acknowledge that there are alternative ways of conceptualising, organising, and utilising knowledge. In other words, knowledge socialism and capitalism are tools rather than ends or dogmas. This hints at pragmatism as a philosophical bridge, prioritising practical outcomes over ideological purity. For instance, capitalistic patents may be used to attract private investments in costly innovations such as the development of medicines, while taxpayer-funded socialist mandates can compensate the developers and ensure stable and affordable public access during health crisis. The survival and prosperity of mankind as a knowledge species now depend on its capacity to handle ongoing global issues (e.g., climate change), and it thus behoves us to be oriented toward the effectiveness of collective problem solving instead of ideological skirmishes (Locatelli & Marginson, 2023).

On the other hand, it has been observed that both knowledge socialism and capitalism operate simultaneously at three levels and often overlap. Globally, there is tension between profit-driven globalisation (e.g., multinational corporations, global rankings, and intellectual property regimes) and networks of international collaborations, open-access and global public goods (Marginson, 2007). Nationally, governments walk a tightrope between economic competitiveness and social equity (Buckner, 2017). Locally, universities navigate market and financial pressures while upholding the ideal of community-engaged, open knowledge sharing (Macfarlane, 2024). To say that knowledge systems are neither purely capitalist nor socialist is nothing new, yet there have been relatively few efforts to systematically examine how knowledge capitalism and socialism are in motion and in interaction with each other across the three levels and the multilayered and contested landscape that follows. Marginson and Rhoades’s (2002; also see Marginson, 2022b) glonacal theory, which has been widely used to study higher education and science, offers a particularly apt and nuanced framework to analyse the hybridity and tension between knowledge capitalism and socialism at and across different scales. The discussion below draws on the glonacal framework to innovatively shed light on the various shades of grey in between knowledge socialism and capitalism that we now live in.

A Glonacal Perspective on Knowledge Capitalism and Socialism

The glonacal framework highlights that the global, national, and local scales are not isolated but interconnected, constructing and reconstructing one another. In the context of knowledge capitalism and socialism, global forces can shape national strategies. Global markets, world university rankings, and transnational intellectual property regimes pressure nations to prioritise competitive, market-driven research and education, but international agreements and global crises can also push nations to open to and collaborate with each other. For instance, during COVID-19, many nations temporarily waived vaccine patents and other medical secrecies for the greater good of mankind. National governments often juggle global capitalist demands with socialist commitments, and they are subject to but not determined by the global systems. Powerful nations such as the US, the UK, and China can project their models of higher education and research onto global norms. World university rankings that measure institutional performance and competitiveness, many of which favour the elite, Anglo-American research university, now define “excellence” worldwide, albeit with controversy (Ordorika & Lloyd, 2015). Nations may swim with the tide of global capitalism; alternatively, they may resist it by creating regional alliances (e.g., the European Universities initiative, the ASEAN University Network) and advocating open science, creating opportunities for collective human flourishing.

In terms of national-local interactions, national funding mechanisms can force universities to follow national economic goals, potentially sidelining local community voices. Conversely, nations may introduce socialist policies such as free tuition schemes to support local access and development, though this is contingent on national fiscal capacity. While the national scale often precedes the local scale in administrative terms, local agents can exert their agency on national agendas. Grassroots efforts may help governments to revise existing policies, and universities may lobby national authorities to protect academic autonomy and secure funding for open education, open access, open archive and so on.

Thirdly, as to the global-local dynamics, universities compete globally for talents, resources and prestige, following corporate styles to maximise profit while struggling to fulfil local commitments and public-service missions. Freely available and user-centred platforms like MOOCs ¹ act as the socialist counterparts to global capitalist pressures on universities and individual researchers, enabling local actors to participate in knowledge-making and sharing and contribute to a global science collective. Similar to the global-national and national-local dyads, the global-local relationship is a two-way street. Open-source software and community-led platforms challenge global capitalist norms by creating decentralised, non-proprietary models, and indigenous activism can escalate into socialist-leaning global movements.

The glonacal framework usefully sheds light on interdependence and negotiations across scales. On the one hand, no scale exists in isolation. Global trends are enacted nationally and locally, while local practices can scale up to restructure national/global systems. On the other hand, actors at all scales have agency. Depending on the context, certain scale/actor might appear more influential than others, but no one remains dominant forever. Overall, the glonacal theory reveals a complex, contested, and evolving landscape where knowledge capitalism and socialism collide and coexist across global, national, and local scales. Global knowledge capitalism prevails, but socialist initiatives and practices persist through local/national resistance and hybrid models. To realise knowledge socialism necessitates coordinated efforts across scales, and the goal is a balanced, prosperous glonacal system in which knowledge serves collective human welfare, not just market logic.

Academic Publishing and Knowledge

Academic publishing enables the sharing of new knowledge and sustains ongoing scholarly discussions about discoveries, issues and challenges. Amidst a variety of publication outlets under the umbrella term “academic publication”, academic journal articles have gained growing popularity and are often regarded and treated as the more “valuable” currency in the academic prestige economy for various reasons, including their credibility grounded in peer review, responsiveness to cutting-edge academic progress, and the relative easiness of rank ordering on the basis of journal selectivity and citation impact (Cope & Kalantzis, 2009; Kwiek, 2021).

A popular discussion about academic publishing relates to the power of English as the de facto language of global scientific communication. Although “the majority of the world's scholars do not possess English as their first language” (Flowerdew, 2008, p.77), empirical evidence points to a general trend toward Anglicisation in linguistic terms in academic publishing. Currently, publication in “mainstream international journals” (predominantly published in English) is the main currency used by most evaluation schemes to assess research performance in both Anglophone and non-Anglophone countries (Salager-Meyer, 2014; Lu, 2019a). Apart from meeting the requirement for research assessment, English-as-an-additional-language (EAL) researchers may also publish in English to enhance their research readership and impact (Huang, 2011) and to widen international networks (Curry & Lillis, 2010; Li, 2014). That said, important differences remain across research fields and disciplines. There is a certain consensus that the target audience, the kind and scope of research tend to vary from discipline to discipline (Sanz et al., 1995; Petersen & Shaw, 2002; Gao & Guo, 2023). More specifically, research on fundamental aspects of nature is considered more likely to appeal to an international readership, while humanities and social sciences (HSS) scholarship is generally more locally oriented (Zhang et al., 2021). It has been assumed, albeit without controversy, that researchers publishing in science and engineering “share the same knowledge, scientific interests and concerns all over the world”, whereas in HSS, “cultural, linguistic and historical features play an important role” (Bordons & Gómez, 2004, p.190). Additionally, there is a relatively stronger ethical consideration in HSS, “in that there is a duty to make research accessible to the communities studied as far as possible” (McGrath, 2014, p.13). Invoking the multiple layers of complexity as a result of disciplinary variations, several studies have justified the relevance of non-English languages in “local and culture-encumbered” (Ferguson, 2007, p.17) scientific domains in HSS (Swales, 2004; Burgess et al., 2014).

In non-English countries, the idea of a “good researcher” is increasingly tied to competence in academic English (Olsson & Sheridan, 2012), and Flowdrew used the term “stigma” to denote the feeling among many “EAL writers who have difficulty with producing written English at an acceptable level” (2008, p.79). Scholars noted that EAL researchers face extra investment in academic English learning (Uzuner, 2008; Liu & Buckingham, 2022) and may encounter linguistically-related editorial inhospitality from English-language journals (Belcher, 2007; Xu, 2023). The prestige of English-language publications has also been suspected of damaging the visibility of non-Anglophone research and undermining epistemic diversity on the international stage (Olsson & Sheridan, 2012). For instance, Chinese HSS researchers claimed that the Chinese-medium literature they cited in Anglophone publications was hardly re-cited (Feng et al., 2013), though to what extent this lack of re-citing is attributable to limited access to Chinese-medium sources remains unclear. Furthermore, there are concerns that the subtlety and richness of meaning in HSS writings are not easily transferable from local languages to English (Harrison, 2014). Scholars also warned about the marginalisation of local issues and loss of scientific vocabulary in languages other than English (Yuan, 2014; López-Navarro et al., 2015). Notwithstanding these difficulties and some English-as-an-additional-language researchers’ reported emotional, ideological and ethical commitment to local languages (Duszak & Lewkowicz, 2008; Li & Flowerdew, 2009; Bocanegra-Valle, 2014), many EAL researchers still reserve their best work for mainstream international journals (Li, 2014). There is widespread resignation to the dominance of English even among researchers who hold a critical attitude toward this language of publication (Burgess, 2014; McGrath, 2014).

However, it must be said that there are significant benefits to using a common language for science. This enables a common space for global conversations and allows mankind to effectively address planet-scale issues on an integrated basis. Though such usefulness is seriously reduced in terms of epistemic diversity and cognitive justice when intellectual exchange privileges knowledge in a single language, it is possible to strike a balance between commonality and diversity. Some scholars have argued for an equal status of all English users in academic writing, regardless of their linguistic backgrounds (de Swan, 2001; O'Connor, 2006), while others have called for translating local-language works into English (Marginson & Xu, 2023). The efforts to “de-anglicise” English echo Marginson's (2004) vision that English as a global language in higher education could, under certain conditions, develop into diversified use-forms instead of linguistic homogeneity.

Scholars have noted not just the English-language primacy but also the geopolitical relations shaping academic publishing, problematising differential opportunities and obstructions faced by knowledge agents in various parts of the world. The prestige of researchers, faculties, and universities is closely linked to their ability to publish in high-impact journals (Bogotch, 2012), and the majority of journals with outstanding impact metrics are based in Anglo-American countries (Albuquerque, 2021). Researchers associated with British and American higher education institutions are over-represented in both authors and editorial members of those journals (Paasi, 2015). Despite a significant rise in submissions from Asian countries to these journals, their acceptance rate falls far behind that of Anglo-American counterparts (Tierney, 2018). Moreover, content analysis of literature in geography (Yeung, 2001), educational leadership (Mertkan et al., 2017), and communication and media (Demeter, 2018) revealed that research topics are disproportionately situated in Anglo-American contexts, while citation analysis further adds quantitative evidence of the geographic imbalances (Danell, 2013; Collyer, 2014).

The asymmetrical landscape of academic publishing reflected in scientometrics is consistent with the centre-periphery model, derived from Wallerstein's (1974; 2006) world-systems theory. The model “functions as both a normative framing of the world and a description of the actually existing hierarchy of resources and influence, in which some countries or cultures are the global metropoles and others are successively positioned away from the centre with decreasing proximity, power and scope to affect the agenda” (Xu, 2020, p.158). Manifested in material and cognitive inequalities, the centre-periphery framework largely corresponds to other divisions, such as the Anglophone/non-Anglophone and West/non-West (Alatas, 2003). The two mechanisms of inequalities are coeval in time, reinforcing each other. Material inequalities refer to a concentration of leading scholars, universities, databases and publishers in centre countries ² (Altbach, 1998; Frenken et al., 2010), all of which constitute a nation's material capacity for scientific production (Marginson, 2022c). Within the centre-periphery framework, systems at the centre are considered to produce a significant amount of research output, control means of knowledge dissemination, and have a global reach of ideas. The asymmetry of the materiality of science makes it “difficult for works published in the periphery to circulate in the metropole, and to other parts of the periphery” (Connell, 2007, p.219).

Cognitive inequalities, by contrast, relate to less visible aspects, such as values, theories, and methodologies. Critical theorists have taken issue with the historical and ongoing dominance of one particular civilisational source of knowledge (the Anglo-American) since the 1960s (Mignolo & Tlostanova, 2006). Whilst knowledge must emerge from a particular locality, major social theories developed in European and American contexts are “master narrative” (Lyotard, 1984) that acts to generate “universally” relevant and applicable findings, concepts, and statements (Keim, 2011). In the long-lasting Orientalist stance, the East has been reduced to a demi-Other to be known rather than a knowing subject in its own right (Kuus, 2004). Social thinking in non-Western societies has been objectified as something to learn about rather than intellectual sources to learn from, and what exploded was not scholarship from the periphery but about the periphery (Slater, 1998). Studies showed that editors and peer reviewers of international journals, who were predominantly from “centre countries”, acted as both facilitators to offer feedback and gatekeepers to regulate the relevance and publishability of published articles (Lillis & Curry, 2010). Swales (2004) referred to the phenomenon as “the skewing of international research agendas toward those most likely to pass the gatekeeping” (p.52). Researchers from so-called peripheral countries were shown to cater to the requirements of Western gatekeepers by addressing Western problems and modelling Western theories and methodologies (Deng, 2010; Chou, 2014), raising concerns over academic self-colonisation and the local relevance of research (Geerlings & Lundberg, 2018).

The centre-periphery theory explains the global inequalities in knowledge construction and knowledge traffic, but it features a deterministic analysis and fails to consider the openness and dynamics of global science and the agency outside the metropole (Marginson & Xu, 2023). The interpretive framework leads to an over-generalised and static vision of world science that reproduces the Eurocentrism it opposes. Globally distributed capacity and power relations in science are changing rapidly. On the material side, large-scale science infrastructure and output have been widely observed outside centre countries (Marginson, 2022d), and the growing “South-South” scientific collaborations are signs of alternative transnational knowledge circuits to the unidirectional centre-to-periphery flow (Fitzgerald et al., 2021). On the imaginary side, increasingly de-territorialised global flows create opportunities for reciprocal dialogues and multiplicities in understandings from every network node in both Western and non-Western societies. Strategies for countering the monoculture of global research are now a topical reference, mainly incorporating discussions on the theorisation of the effects of power/knowledge so as to redress power asymmetries and to facilitate the re-activation of subordinated local intellectual traditions (Shin, 2013; Geerlings & Lundberg, 2018).

To sum up, far from being a neutral conduit for knowledge, journal publications sit at the junction of language politics surrounding knowledge representation (the linguistic medium whereby knowledge is expressed, circulated, and understood), epistemic struggles over knowledge construction (what is considered knowledge and whose knowledge counts), and evaluation of knowledge impact in a metrified age (in what way knowledge is considered useful and meaningful). In this way, the study of academic publishing is not simply about academic publications but delves deeply into the sociology of knowledge, and understanding academic publishing helps us to capture the rhythm of science.

Academic Publishing in the Time of AI: Swinging Between Knowledge Capitalism and Socialism

Capitalistic privatisation and control coexist with socialist sharing and openness in academic publishing. The massively big publishers (Elsevier, Springer Nature, Sage, Taylor & Francis, Wiley) account for nearly half of the world's active peer-reviewed English-language academic journals (Johnson et al., 2018), while Elsevier-Scopus and Clarivate Analytics-Web of Science profit from their enormous capacity to monitor and regulate the world's published output (Pranckutė, 2021). The commercialisation of scientific intelligence and the commodification of published research have transformed the production and perceptions of knowledge, pulling knowledge creation, circulation and re-creation towards the privileged few who can afford the informational services and products. Movements such as open access and open source have been pushing for accessible and publicly funded scientific infrastructures, but such efforts face the daunting task of addressing surveillance publishing (Pooley, 2022) and, more broadly, the platform capitalism surrounding (Srnicek, 2017) science. Why does it matter? Knowledge capitalism may incentivise and facilitate innovation in the short term, but there is a good possibility that it stifles scientific progress in the long run. Science can only function properly in an environment of organised criticism where ideas, methods, and results are made openly available to the scientific community so that they can be tested, scrutinised, verified and further developed by other researchers (Merton, 1973). New studies build on what is already known from existing literature, some more so than others. The collaborative and cumulative nature of science means that science can be in its optimal condition only if new knowledge is incessantly and unreservedly made available to the scientific community.

The capitalistic privatisation and monopolisation of knowledge have been associated with geopolitical inequalities in scientific activities. Global research now incorporates diverse agents, languages, cultures, research agendas and paradigms, but it is far from a truly free science characterised by an open and equal dialogue among everyone engaged in the scientific endeavour (Xu, 2022). Most academic journals that are managed by big publishers and listed in Scopus or Web of Science Core Collection are based in Anglo-American countries and published in English (Asubiaro et al., 2024). Such journals are globally visible and recognised not just because of English as the lingua franca of science and the commercial giants’ extensive transnational business networks but also due to these journals’ integration as standard currency into world university rankings. Open-access journals are typically less influential in metric terms and are published outside the Anglo-American world. The linguistic, material and social resources necessary to get published in the highly sought-after, “global” journals are concentrated in Anglo-American circles, making it difficult for the well-established journals to develop a more diverse and equitable knowledge profile with a multiplicity of voices (Mertkan et al., 2017; Albuquerque, 2021; Sugiharto, 2024). While many Anglo-American academics strive to work with colleagues and institutions in other parts of the world in the spirit of mutual respect and benefit, a significant portion of knowledge is still kept outside the global knowledge pool.

Knowledge capitalism and socialism are in a dynamic struggle of mobilisation and re-mobilisation of people, resources, and ideas — neither seems able to overcome the other. Recently, however, the growing capability of AI puts a new complexion on the situation. Being one of the most hyped innovations of our times, AI has been identified as a disruptive development with the potential to revolutionise academia and scholarly publishing. The definition of AI has been evolving since the groundbreaking work of Alan Turing (Turin, 1936, 1950), as there have been substantial advancements in AI on both theoretical and practical fronts. A current approach considers AI as “computing systems that are able to engage in human-like processes such as learning, adapting, synthesising, self-correction and the use of data for complex processing tasks” (Popenici & Kerr, 2017, p.2). The extensive interests in AI from scholars in the fields of maths, psychology, linguistics, philosophy, neuroscience, education and so on who link AI to nomenclatures, perspectives, agendas in their own disciplines have led to definitional diversities. While a useful step towards comprehending AI is to see it as a General-Purpose Technology (GPT) that is widely used and capable of ongoing improvement and innovation in a wide range of sectors (Trajtenberg, 2018), it is important to place AI in specific contexts for analytical clarity. On that note, this study focuses on AI and its implications for academic publishing. Meanwhile, it has been noted that a disproportionate amount of literature on AI is dominated by ChatGPT for its eminent versatility and capacity, which represents but one among many AI models (Marr, 2025). In any case, supplementing knowledge of AI with contextual situatedness and richness without reducing AI to any single model or technology raises the crucial question about how to systematically and appropriately conceptualise AI. The work of Kate Crawford (2021) seems particularly pertinent here, providing a comprehensive and nuanced framework that understands AI as simultaneously consisting of technical approaches, social practices, and industrial infrastructures.

Firstly, technical approaches are central to AI, as AI is essentially about the simulation of intelligent behaviour in machines so as to enable machines to mimic human intellect on important matters such as language processing, reasoning, and problem solving (Chowdhary, 2020). Self-learning is a critical element of AI, allowing AI models to absorb new information and improve their judgments through past experiences and data-based calculations. Development of machine learning (ML) has been the most powerful driving force for AI progress over the past two decades (Lu, 2019b); and deep learning (DL) and artificial neural networks (ANN) have been two significant contributors to more recent ML advancement, enabling the possibility of a jump from regular to generative AI (LeCun, Bengio & Hinton, 2015). Peters et al. (2024) neatly summarised the technological transition as follows:

Regular AI uses supervised machine learning where humans train the machine to label correctly patterns in source binary data or unsupervised machine learning where the machine asks humans to label statistical regularities or irregularities in source data. This produces programmed responses—reliably, consistently, predicably. Generative AI also requires training, with massive datasets processed through a multitude of parameters. But not only are the results it generates unpredictable; they are always uniquely reconstituted digital artifacts; text, image, sound or in multimodal combination (p. 841).

Unlike earlier technologies such as the internet and social media platforms that provided networked environments that facilitate human communication, interaction and creative work, generative AI aims to simulate human cognitive processes themselves. The influences of this on academic publishing have been transformative. On the side of authors, AI signals a fundamental reimaging of how scientific knowledge is made. AI tools have revolutionised the reviewing of literature, as they do not just scan literature but deconstruct complex academic discourse into data points, provide researchers with concise summaries, and even propose novel research areas (Kang et al., 2020). As to data analytics, ML algorithms are now good at conducting tailored content analyses and finding patterns that may otherwise escape human awareness (Onyema et al., 2022). It is well-known that AI's capacity for automation relieves researchers of many time-consuming and often repetitive tasks that are essential to research but do not necessarily require much intellectual input; yet AI can also constitute an intellectual input into research by allowing for a more nuanced, thorough understanding of phenomena and enabling researchers to make discoveries that they might otherwise not have been capable of, thereby expanding the boundaries of what is known and what can be studied. Moreover, AI has now been built into established editorial systems to assist editors in identifying potential reviewers and evaluating their “suitability”, while the technological interventions can also be found in peer-review processes where AI are used to perform initial screening of submitted manuscripts and make suggestions on the quality of writing. For example, AI algorithms can integrate the traditionally small and invitation-based reviewer networks into an extensive, crowd-based pool of reviewers (List, 2017), and papers that fail to meet the baseline quality standards do not even need to consume human reviewers’ efforts and time (Checco et al., 2021).

AI's functionalities in academic publishing cannot be dismissed as merely technological but have profound implications (Xu et al., 2021). On the one hand, much research expertise is either embodied or taught behind closed doors; and despite science being now a borderless endeavour combining autonomous scientists in relational networks across national boundaries, institutional infrastructures and specialised personnel that make up scientific activities are largely housed within the borders of national societies (Sá & Sabzalieva, 2018). While science capacity is still very much differentially distributed across individual and national scales, AI's research-related service can be made immediately available for the asking via cloud-based platforms. As such, there is potential within AI to democratise science by opening what is in its toolbox to all its users. On the other hand, although the evaluation of research and regulation of publishable content have been historically in the hands of human actors (especially editors and reviewers), journal gatekeepers find it rather difficult to handle the tsunami of scientific output being produced every day without the help of automation. Notwithstanding concerns over the decline in human involvement in academic journal management, AI can be designed and implemented to make paper reviewing and editorial decision-making faster, fairer and more accurate (Lund et al., 2023). Theoretically, AI can lead to an egalitarian spread of science capacity and nudge journal gatekeeping towards communal organisations and more efficient operations.

The increasingly active role of AI in research and academic publishing has brought to the fore the questions about whether and to what extent AI tools can be considered knowledge producers in the same manner as human beings. One notable criticism relates to the “black box” quality of AI. Many present AI models, especially those approaching generative AI, are based on super-complex algorithms with thousands and even millions of parameters. There is limited transparency regarding AI's operational mechanism either because the developers themselves have yet to fully understand how their creation works or because the source codes are treated as privately owned, protected secrecy (Pasquale, 2015). This obstructs any investigation into AI-generated knowledge claims by going back to its sources and reasoning processes, meaning verification and justification of truth can only take place post-hoc in the presence of AI.

A different but related concern is about the originality of AI-generated content. Algorithmic opacity makes it very challenging to determine the degree to which sources from AI's training set may be quoted or otherwise used in the final output (Dehouche, 2021); and though AI may propose novel ideas, whether serialised production of homogenous content based on the same algorithm counts as original work is a contentious matter. Certainly, a lot depends on the agreement reached between AI users and developers, but this hardly seems a satisfactory solution in academia, where credit and accountability are key to the scientific ethos. Furthermore, whereas human knowledge derives its meaning and significance from social contexts involving material environments, sensory experiences, needs, motivations, and interests, AI can hardly develop a similar notion of empirical reality (Larson, 2021). The functioning of most AI models is to produce coherent and logical narratives—they are genre machines, gathering what they believe to be facts from textual sources but without being able to effectively verify their connections to the empirical world (Peters et al., 2024). Some AI tools have even been found to make up non-existent facts to complete a plausible statement (Amaro et al., 2023). Perhaps unsurprisingly, AI models have been described as “stochastic parrots” that regurgitate what they “hear”, often twisted with inexplicable randomness (Hutson, 2021). There is little conclusive evidence that AI necessarily augments human judgement and creativity (Dourish, 2016); and although AI has been included as co-authors in some academic papers, many publishers and academic institutions have limited or banned AI generative tools fearing dubious research outcomes (Lund & Naheem, 2024). Until AI overcomes its current technological limitations and recalibrates its position on openness, truthfulness, credibility and originality, the trouble it brings to academic publishing is likely to outweigh the benefits.

Secondly, the social practices of AI refer to the complex human activities surrounding the creation, deployment and application of AI, revolving around questions about who is pushing the agendas behind AI, for what purposes they are doing so, and who has the decision-making power. Ultimately, these considerations relate to political struggles over inclusion and representation (Crawford, 2021). There have been efforts to use AI to promote universal rights to knowledge and human collective learning processes. Open-source initiatives and community-driven AI platforms enable free access to AI and decentralised participation in AI development and usage, which may then lead to greater openness in science. For example, AI translation tools can effectively break language barriers, facilitating the production, publication and circulation of knowledge in multiple languages (Mohsen, 2024). Additionally, AI-based research search tools, many of which surface freely available versions of papers and prioritise open-access content in search results, have been used to deliver customised paper recommendations based on research interests and reading patterns (Wagner et al., 2022).

Here, the implications of AI for academic publishing are twofold. On one side, AI allows independent researchers and scholars in underfunded institutions, especially those from the Global South, to access cutting-edge findings at little cost. On the other hand, “knowledge is made as it circulates; it is never made completely in one place and then simply consumed elsewhere” (Agnew, 2007, p.146). AI does not inherently bypass existing paywalls, but it can elevate the role of open-access alternatives in scientific discoveries. There is an AI-powered prospect of universal science in the sense of knowledge for all, characterised by free and unrestrained travelling, sharing and merging of ideas originating from diverse personal, institutional, cultural and national backgrounds. Subscription paywalls and other efforts to privatise and monetise knowledge are likely to stay in at least the near future, but they are already in jeopardy in a world where science is increasingly driven by openly shared knowledge.

While many AI tools are freely available, commercial bodies, big corporations in particular, have been quick to privatise and monetise AI. The giant tech companies, often known by the acronyms GAFAM ³ , have become leading players in AI and secured a nearly monopolistic technological and commercial position (Statista, 2025). The big AI companies are predominantly headquartered in Western countries, and their AI systems are trained primarily on English-language datasets, with limited attention to non-English sources (Park, 2024). This suggests that not only has AI been absorbed into the privatisation and commercialisation of scientific knowledge and information but also that knowledge capitalism and its associated inequalities in global knowledge systems have persisted and adapted. There is evidence showing that the computational logic of AI can influence researchers in a fashion that reproduces patterns/biases already existing in AI's training bases, potentially leading to dangerous outcomes. For example, it has been consistently argued that the affinity between AI and quantifiable data can result in a skewed methodological trend that prioritises quantitative measurements over qualitative inquiry (Anis & French, 2023; Chubb, 2023). Given the clustering of top AI companies in the Western hemisphere and the concentration of AI training in the English medium, it is not unreasonable to assume that the AI-induced methodological bias can extend to phenomena which are closely related to social justice, and failure to adequately attend to such complexities can have serious scientific and societal implications. There is a risk that the unbalanced landscape of AI heralds the continuance of the centre-periphery global research order instead of the realisation of knowledge socialism.

AI has been rapidly evolving over recent years, but advances in safeguards against AI danger and harms by governmental and legal means have not been progressing at the same pace. This is because 1) the technological nature of AI and its social significance largely remains unknown and 2) AI governance has been a matter subject to weak oversight. Influenced by the Hayekian school of thought, neo-liberalists believe that governments should simply set protective boundaries that foster favourable conditions for wellbeing while staying out of the way to allow innovation to flourish of its own accord (Davies & Gane, 2021). Seeing individual liberty as the precondition for justice, intellectuals with a neoliberal bent argue for a laissez-faire regulatory framework. In their eyes, corporations and businesses rather than state agencies should take the lead in inventing and upgrading technologies, and entrepreneurs are most productive and creative when they are allowed in the absence of external regulations to freely explore and compete (Castro & McQuinn, 2015). AI companies and technologists may well lead us to a beautiful new era, but the complexity of real life and the stakes involved are too important to be put solely to their discretion. There have been calls to hold AI more strongly in check, and these efforts can be classified into four basic categories: mandating more transparency, regulation, external auditing, and enforcing accountability.

To begin with, governments have the authority to initiate watchdog bodies and commission transparency investigations. The goal is to monitor the input into AI training, the output produced, as well as how it might affect the public, upholding civic rights to explanations of the workings of algorithmically driven models by demanding disclosure about criteria for incorporating and excluding data from AI systems and the decision-making behind (Ferretti, 2022). It is also within the power of governments to establish regulatory frameworks via policies and legislations. Dataficaition, understood as the process of turning many aspects of the world into digital and often quantified data (Cukier & Mayer-Schoenberger, 2013), is central to AI, whose performance hinges on the training set data. AI's voracious appetite for data has attracted concerns over to what degree AI can extract information from people's lives and where it does so, what data are taken. Proponents of the free market are suspicious of government intervention, especially of a growing bureaucratic structure, and they insist that AI is in a prime developmental condition where there is minimal regulatory interference. On the contrary, advocates of regulation claim that AI technologies shall be met with as much hope as scepticism, arguing for the necessity of preventative procedures to hold back data misuse and defend individual privacy. Pro-regulation voices have been reflected in the European Union, where legal rules were already in place to regulate algorithmic behaviours regarding personal information (e.g., the 2018 General Data Protection Regulation). Shortly after, similar legislations followed in many Asian and Latin American countries (UNCTAD, 2025). The overall trend is clear, but operational details such as the responsibilities of operational personnel, costs, and implementation plans are largely unsettled. At the same time, pressures for third-party AI auditing have been gaining momentum (Raji et al., 2022). The purpose is to check for any bias, discrimination and unfairness in AI development and usage. The success of such initiatives depends on how “abnormality” is defined, what counts as “acceptable” algorithmic behaviour, as well as who appoints and sponsors the inspectors. It remains to be seen how AI auditing can be translated into practice and what outcomes await. Furthermore, as much as AI is recasting our societies in technocratic terms, it is not exempt from one of the foundational tenets of democratic politics—accountability. Worthy plans to enable public scrutiny of AI and to keep AI legally bound have been discussed above, but ultimately it is the creators and curators of AI that shall be held responsible for its use and impact (New & Castro, 2018). To do so, however, is a daunting challenge that entails much political will; and now it seems rather easy for the target actors to claim accountability without consequences by paying lip service to ethical campaigns and carrying out nominal self-reform. Generally speaking, effective mechanisms for ensuring ethical, fair, and responsible social practices of AI are lacking, and coordinated attempts at AI governance are just starting to restore check and balance.

Lastly, the industrial infrastructures of AI incorporate datasets, computing power, and human talent that make AI algorithms and any subsequent algorithmic product possible in the first place. The interplay between data and digital platforms is of paramount importance in the age of AI. Digital platforms are shared virtual spaces where different kinds of actors (e.g., consumers, sellers, producers, suppliers) engage and interact with each other via given digital infrastructures (Srnicek, 2017). Such platforms are conveniently placed as data brokers thanks to their prerogative to capture, store, and manage the data generated by activities on platforms (Crain, 2018). The surveillance academic publishing (Pooley, 2022) and more broadly the platform capitalism (Srnicek, 2017) dominated by several big conglomerates mean that a handful of companies now control vast amounts of data. The giant AI corporations have either partnered or merged with the big data brokers at high prices prohibitive for many others in the AI market, thus gaining an unmatched advantage in AI training (Dyer-Witheford, Kjøsen & Steinhoff, 2019). Likewise, only the major AI corporations have the resources to purchase and continually upgrade computing facilities on a large scale, which are needed to perform and develop ML and DL. The big AI companies themselves can also follow the platform capitalism model by renting out their computing facilities while maintaining control over what happens via the infrastructures (Srnicek, 2019).

Since 2015, the leading tech companies have intensified the hunt for AI talent, acquiring AI start-ups often in the face of limited resistance from blockchain firms and the military. The escalating competition for AI talent creates a divide between algorithmic developers who are hired and handsomely paid by giant tech companies (Poseliuzhna, 2023), and the rest of computer scientists and other AI-associated social groups that are doing poorly rewarded work and even free labour (Altenried, 2020). Some selflessly dedicate themselves to AI research and development in a communal and egalitarian spirit, but this is hardly sustainable in the long run, not least in financial terms. All in all, we are now in a situation “where there are only a few AI behemoths, who own the expensive computational infrastructure, have access to vast amounts of data to train ML and DL models and can attract the highly skilled AI talent to develop new systems and services” (Verdegem, 2024, p.732).

As AI progresses, we are already in uncharted waters in terms of science and academic publishing where supercomputing, data-intensive algorithms and human-like chatbots construct an unfamiliar and unpredictable landscape. This is a global matter whose uncertainties and dangers demand collective awareness and action and whose benefits are to be harnessed by all and for all. Yet, given the concentration of AI industrial infrastructures in the hands of several Western-based Big Techs alongside the rising transformation of AI, especially the highly advanced models, into commodities available only to those that can afford them, chances are that AI entrenches existing hierarchies and closures in global science. The effects will be comprehensive and far-reaching, characterised by a combination of technological narrowing to restricted AI access, relational narrowing to fewer participants in AI-augmented science and scholarly publishing, and epistemic narrowing to homogenous intellectual contributions. Despite global inequalities in the distribution of AI resources and services, AI algorithms, as a particular form of knowledge, are very much a de-territorialising force that neither strictly corresponds to national borders nor can be contained within national boundaries. While some countries have asserted technological sovereignty for national security reasons, it is time that they see each other as fellow travellers on the AI journey and unite in setting out some beneficial ways that AI can be developed and used.

In the final analysis, Crawford's three-dimensional framework allows us to move beyond seeing AI as simply codes or algorithms, capturing instead the real-world human context that makes AI and its social significance possible. AI is poised to profoundly influence the dynamics between knowledge capitalism — a system where profit-driven agents privatise and control knowledge creation and dissemination — and knowledge socialism, which promotes open, collective knowledge sharing. Zooming in on academic publishing, this study has demonstrated the multi-scalar potential within AI to contribute to knowledge socialism: locally, an egalitarian spread of research capacity and open access to published knowledge among researchers; nationally, the cross-fertilisation of ideas across not just the physical national borders but also cultural and linguistic differences; and globally, a decentralised, communal and efficient working mechanism of academic publishing that brings together authors, reviewers and editors from different parts of the world and shares knowledge with the world. However, it must be acknowledged that knowledge capitalism largely retains its firm grip on knowledge and our societies, and AI as a relatively new invention is problematic in many technological, sociopolitical, and ethical aspects. AI—even if equipped with democratic affordances—is too easily subsumed into capitalistic markets and commercial infrastructures. AI in its present form is far from realising its full potential of promoting knowledge socialism, and the ultimate outcome depends on governance frameworks, infrastructure ownership patterns, and whether collective ethical considerations can supersede individual commercial interests.

Toward an Equitable Future

This paper draws on academic publishing as a window to explore the dynamics between knowledge socialism and capitalism, two opposing meta-frameworks that differ in political, economic, social and philosophical terms. The landscape of academic publishing now is characterised by a mixture of knowledge capitalism and socialism. The two are in a fluid state of competition and co-existence, and it is hard to tell precisely where knowledge capitalism ends and where knowledge socialism begins. AI's penetration of academia is a game changer, marking not just a groundbreaking technological progress but, more importantly, a significant methodological, ethical, and social turning point. AI is not a neutral tool but reflects the values, interests, and agendas of those who design and employ it. If left unchecked, AI can be subjected to capitalistic appropriation to serve privatised production and circulation of scientific knowledge and the commodification of published output.

To ensure that AI tilts the scales in favour of knowledge socialism, policies and collective actions promoting equity, transparency, and openness are needed. State authorities can negotiate transnational deals to support collaboration on AI research and the sharing of relevant resources and intelligence. Governments can increase investment in public funding for AI development and infrastructure and mandate that AI models be trained on diverse, accessible datasets. At the same time, the academic community must believe in the ideal of open science. Universities can push back against proprietary tools and metrics, while grassroots movements could pressure commercial bodies to follow ethical and sustainable practices. Public-private partnerships might combine free, open, and essential infrastructures with commercial add-on services, but this ought to be treated with caution. The key lies in the conceptualisation and operationalisation of AI and knowledge as public goods, guided by transparency and openness. If aligned with progressive principles, AI can act as a bridge to knowledge socialism rather than a weapon of exclusion. The aim of this paper is not to exhaust AI's potentialities for academic publishing or to draw a single true conclusion but to contribute to a nuanced viewpoint that extends beyond the technological aspects of AI to the foundational issues around which scientific ethos and social zeitgeist revolve. The goal is to open the way to collective exploration and, where possible, critical questioning of the ongoing understandings of AI in science. This paper is, first and foremost, an academic act, but it is also hoped that the discussion helps to equip the audience with the appropriate literacy to engage in any upcoming integration of AI into academic research and our societies as a whole.