Work

When social interaction becomes an algorithm: AI-mediated education, intercultural competence, and the risks to democratic schooling

Year: 2026
Type: article
Abstract: The rapid integration of generative Artificial Intelligence (AI) into educational settings has been increasingly framed as a transformative opportunity for learning, especially in relation to Social-Emotional Learning (SEL) and Intercultural Competence (IC). Generative AI systems and adaptive learning platforms, including conversational interfaces and adaptive feedback mechanisms, are increasingly positioned as tools for mediating educational interactions (Riordan Alfredo et al., 2024;Holmes & Tuomi, 2022). However, this assumption obscures a critical issue. Research on Artificial Intelligence in Education (AIED) highlights that educational AI encompasses diverse technological applications and involves not only technological design but also questions concerning adoption, impacts, challenges, and the ways in which such systems shape teaching and learning processes (Wang et al., 2024). From a critical perspective, generative AI systems and adaptive learning platforms can therefore be understood not as neutral tools, but as sociotechnical infrastructures that may reflect and reproduce certain epistemological, cultural, and political assumptions about what counts as information, knowledge, communication, appropriate social interaction, and educational practice (Williamson & Eynon, 2020).In this sense, AI-mediated education refers not simply to the use of educational systems, particularly generative and adaptive learning systems, in classrooms, but to the reconfiguration of educational relationships through algorithmic systems that intervene between teachers, learners, knowledge, and peers (Kuhail et al., 2023). The convergence of generative AI systems and adaptive learning platforms may increasingly shape educational processes through prediction, optimisation, and algorithmic mediation (Turvey & Pachler, 2026;Riordan Alfredo et al., 2024). Instead of direct human-to-human interaction, educational processes increasingly take the form of mediated exchanges, such as student-AI-teacher or student-AI-student interactions, enabled particularly by generative AI systems and adaptive learning systems, where algorithms may shape, filter, or mediate elements of dialogue, collaboration, feedback, and critical thinking (Bayne, 2024;Williamson et al., 2020).It is important to note that AI-mediated education does not refer to a homogeneous category of technologies. Contemporary educational AI includes generative AI systems (i.e., large language model-based tutors and conversational agents), adaptive learning platforms, automated assessment tools, machine translation technologies, and other data-driven educational systems that operate through different technical architectures and pedagogical logics (Riordan Alfredo et al., 2024;Holmes et al., 2019). Recent scholarship has highlighted substantial variation in the degree of human control, automation, stakeholder involvement, and educational mediation across systems (Riordan Alfredo et al., 2024). The present opinion paper focuses primarily on generative AI systems and adaptive learning environments, because these technologies can increasingly mediate educational interactions, shape learning pathways, and influence how learners may encounter knowledge, feedback, and social relations. Consequently, the arguments outlined here should not be interpreted as applying equally to all forms of educational AI. While some applications, such as machine translation and accessibility tools, may facilitate communication and participation without fundamentally restructuring educational relationships, generative AI systems and adaptive learning systems may raise more substantial questions regarding algorithmic mediation, epistemic authority, and the democratic purposes of education. Accordingly, the potential consequences discussed in the following sections are considered primarily in relation to generative AI systems and adaptive learning environments, rather than being generalised across all educational AI technologies.This transformation is particularly significant for domains such as SEL and IC, both of which depend on exposure to difference, relational engagement, and negotiation of meaning (Deardorff, 2006;Cipriano & McCarthy, 2023). SEL in this opinion paper is understood in its main formulation as a framework for developing various competencies, such as self-awareness and social awareness, self-management, and relationship skills (CASEL, 2021). However, SEL is not just a set of individual competencies, but a socially, culturally, and relationally embedded process, through which learners understand emotions, norms, and relationships, negotiate social meanings, encounter different perspectives, and develop the capacity to engage constructively with disagreement. From this perspective, SEL depends not only on emotional regulation but also on participation in shared spaces of dialogue and collective meaning. SEL is consequently a situated process that intersects with power, identity and communication within educational contexts.The connection between SEL and IC can be understood more explicitly through their shared relational and dialogical foundations. In Deardorff's (2006) conceptualisation, intercultural competence is not limited to the acquisition of knowledge about cultural groups but involves attitudes, skills, and dispositions that are developed through engagement with cultural difference, including openness, respect, perspective-taking, and the ability to negotiate meaning. Similarly, contemporary approaches to SEL emphasise that social-emotional capacities develop through relationships, communication, perspective-taking, and the interpretation of social situations rather than through isolated individual skill acquisition (CASEL, 2021;Cipriano & McCarthy, 2023). From a relational and dialogical perspective, these processes necessarily involve encounters with perspectives that are not fully predictable or identical to one's own. Ambiguity and disagreement therefore constitute not merely potential difficulties but important conditions for developing interpretive flexibility, perspective-taking, and the capacity to negotiate meaning. Democratic educational theory further extends this relational logic by positioning plurality, dialogue, and constructive disagreement as conditions of participation in a shared world (Biesta, 2022;Dewey, 1916Dewey, /2018;;Freire, 1970Freire, /2005)). Taken together, these perspectives suggest that SEL and IC share a common relational foundation: both are developed through encounters with others, negotiation of meaning, engagement with difference, and the capacity to remain open to perspectives that cannot be fully predetermined. Democratic schooling provides the broader collective context in which these capacities can be enacted, tested, and developed through participation and dialogue.From this perspective, SEL is sustained not only through emotional regulation but through ambiguity, co-construction, dialogue, and the negotiation of disagreement⎯conditions that may sit uneasily with high predictive and optimisation-oriented forms of algorithmic mediation (Mesquita et al., 2026). Empirical research on generative AI and AI-supported translation tools suggests that these technologies may support intercultural learning by facilitating communication, translation, and access to multiple perspectives (Akhrenova & Moradkhani, 2026). However, this form of mediation also raises theoretical and normative concerns about standardisation, cultural bias, and the possibility that AI may mediate rather than simply facilitate intercultural encounters (Ali et al., 2024;Holmes et al., 2019). Consequently, it also raises critical questions about the nature of social interaction, the role of disagreement, and the conditions under which intercultural and democratic learning take place (Andreotti, 2021;CoE, 2018).Therefore, this opinion paper aims to critically examine how AI-mediated educational settings reconfigure social interaction and to explore the conditions under which such mediation may either strengthen or constrain IC, SEL, and democratic schooling. Rather than treating generative AI and adaptive learning systems as inherently harmful or beneficial, the paper focuses on the pedagogical and relational conditions under which AI-mediated interaction may expand participation, access, and opportunities for dialogue, while also potentially reshaping the conditions of interpersonal and democratic engagement.Towards this endeavour, we first argue that AI-mediated education is not a neutral enhancement of learning, but a form of algorithmic mediation that may shape how learners encounter knowledge, others, and difference (Turvey & Pachler, 2026). Second, we argue that although AI may enhance accessibility and communication, it may also potentially reduce some forms of ambiguity, unpredictability, and productive tension that are theoretically important to both intercultural and social-emotional learning (Turvey & Pachler, 2026). Third, we argue that generative AI systems and adaptive learning systems may contribute to the individualisation of learning and the optimisation of interaction, potentially constraining collective meaning-making processes that are essential to democratic education (Biesta, 2022;Selwyn, 2019). Finally, we argue that generative AI and adaptive learning systems should be understood as supplementary technologies within democratically governed educational frameworks rather than as substitutes for human relationality and dialogical engagement (Wieczorek, 2025;Jin & Wang, 2025). This aligns with calls for democratically governed approaches to educational AI that foreground collective oversight, ethical accountability, and participatory design in the development and deployment of AI systems in education (Kerssens & van Es, 2025).The theoretical contribution of this opinion paper lies not in introducing algorithmic sociality as a new concept, but in extending its analytical relevance to educational interaction. We argue that algorithmic sociality helps explain a dimension that is less directly captured by accounts of algorithmic governance, datafication, platformisation, or technological mediation: how changes in the mode of mediation can reconfigure the relational conditions of educational encounter. Drawing on broader scholarship on algorithmic sociality and governance, we theoretically infer that, in AImediated education, prediction, personalisation, filtering, and automated feedback may do more than redistribute knowledge or organise learning processes; they may reshape learners' opportunities to encounter ambiguity, negotiate difference, engage with disagreement, and participate in collective meaning-making. This provides the basis for connecting algorithmic sociality to the relational of SEL and IC in to the conditions of democratic opinion paper a rather than a The on of including scholarship on generative AI and adaptive learning in algorithmic mediation, algorithmic governance, and algorithmic relational and dialogical approaches to and democratic education and schooling. The also research AI-supported communication, language learning, participation, and theoretical and critical scholarship epistemic authority, cultural bias, standardisation, personalisation, and the of educational technologies. Rather than to the or of AI-mediated education, the of this is to these of into dialogue in to potential through which AI-mediated may strengthen or constrain and democratic of education. The therefore in relation to the of the Empirical are to where the directly examine the technologies or educational processes under broader theoretical and critical scholarship is to develop theoretically and normative Accordingly, the are not as of generative AI or adaptive learning Empirical are considered and are not interpreted as AI systems and adaptive learning platforms increasingly new forms of mediation into the of social interaction within educational approaches to education educational processes as relational in which meaning is through dialogue, disagreement and shared (Biesta, 1970Freire, /2005)). with educational technologies primarily to translation, or access, generative AI systems and adaptive learning systems can more directly mediate learning processes through conversational interaction, feedback, mechanisms, and of learning (Riordan Alfredo et al., 2024;Holmes et al., 2019). Consequently, the integration of generative AI systems and adaptive learning platforms a mediating that may reshape educational interactions through processes of prediction, optimisation, and algorithmic mediation (Turvey & Pachler, this we on of which refers to forms of social interaction that are increasingly and mediated through algorithmic While to such as algorithmic governance, datafication, platformisation, and algorithmic mediation, algorithmic sociality on how algorithmic processes reconfigure the conditions of social to education, this a dimension that is less directly captured by accounts primarily on governance, or technological mediation: the transformation of the relational conditions through which learners encounter others, negotiate difference, disagreement, and meaning AI-mediated education can be understood not merely as the of new technologies, but as a potential transformation of the relational through which learning and social interaction take on broader critical scholarship on algorithms and datafication, we theoretically infer that, under forms of design and such may potentially reduce some forms of and that are important of human Generative AI systems and adaptive learning platforms under forms of design and of knowledge and potentially the of perspectives to From this perspective, certain epistemic may more or than others, potentially the of This is with broader of algorithmic as a form of epistemic rather than neutral but these should be understood as theoretically rather than as across educational AI educational this theoretical possibility concerns not only what is but also how knowledge is Drawing on broader scholarship on prediction, optimisation, and algorithmic we theoretically infer that the of and reduce opportunities for negotiation and ambiguity, which are important conditions for learners to develop interpretive and critical (Williamson & Eynon, the algorithmic mediation may operate as a form of through the and of information, potentially which perspectives more or suggest that logics may shape not only learning processes but also the conditions under which knowledge and social are et al., these concerns should not be understood as by AI. standardisation, cultural bias, and approaches to SEL within educational systems and technologies generative AI. Generative AI and adaptive learning systems may or more such by within increasingly and automated forms of educational interaction. The analytical is not AI these but how forms of algorithmic mediation may or potentially educational This technological while that AI may the and relational consequences of educational suggests that the educational of generative AI and adaptive learning systems cannot be fully understood by only on their pedagogical or technological is particularly important is the through which AI and forms of algorithmic mediation that may reshape educational interaction and the relational and epistemic conditions of education. In this AI refer to by the such as conversational interaction, adaptive feedback, and refer to technical or through which these are such as personalisation, prediction, and or mediating refer to the processes through which and interaction are or refer to how these are into educational practice by and learners, while educational and conditions refer to changes in participation, negotiation of meaning, plurality, IC, and conditions refer to the and through which these systems are and on this we a framework that a potential rather than a AI and may operate through forms of algorithmic mediation, which may in educational interaction and the conditions under which ambiguity, plurality, encounters with difference, and negotiation of meaning are conditions are to SEL, IC, and democratic participation, while pedagogical and human mediation, governance, educational and context may shape how these processes The framework therefore relational without or The framework AI-mediated education as a and potentially process rather than as a how AI and may operate through forms of algorithmic mediation that shape educational interaction in the relational and epistemic conditions under which ambiguity, plurality, encounters with difference, and negotiation of meaning are conditions are to SEL, IC, and democratic participation, which are as rather than of AI and critical negotiation may influence how AI-mediated interactions while pedagogical and human mediation, governance, educational and context may these processes across the The theoretically relationships and should not be interpreted as of or across generative AI systems and adaptive learning environments, but the under which they may operate to the and pedagogical of feedback, or are as that learners can and with others, they may expand access to relational these are as or they may opportunities for learners to explain meanings, and negotiate directly with The is therefore the degree to which AI-mediated remain open to interpersonal and Ambiguity and interpretive AI systems or situated into and they may reduce the of interpretive and questions are for AI-mediated may support rather than interpretive The is not the of but is or for collective or foreground some perspectives while or others, they may constrain learners' opportunities to encounter cultural and epistemic the and of are open to and perspectives remain AI may learners access and that be less to The conditions the of perspectives, learners' opportunities to their and the possibility of or to are to a shared learning for and collective they may support democratic participation by more learners to and contribute their the as a the perspectives or for they may constrain opportunities for disagreement, and collective meaning-making. The is AI-mediated and support remain and collective generative the primarily conversational interaction, feedback, and the of in adaptive learning environments, they pathways, feedback, prediction, and The therefore refer to relational that may operate across these systems rather than to of AI as a together, the a but framework for how algorithmic mediation may reshape educational between the accessibility of participation and the interpretive conditions under which participation the of perspectives, and the of to collective as conditions that may shape AI-mediated In they the of algorithmic sociality by how algorithmic mediation may influence the relational processes through which learners negotiate meaning, encounter cultural difference, and participate in democratic The framework therefore SEL, IC, and democratic participation as of a relational educational rather than as of AI relationships remain theoretically and across educational contexts.The following sections this framework to competence and social-emotional examine its for democratic involves the ability to engage with difference, ambiguity, and negotiate meaning across cultural frameworks (Deardorff, et al., 2019). Generative AI systems and AI-supported translation tools may support these processes through translation, communication and access to multiple Recent research on AI-supported translation and language learning suggests that generative AI and AI-supported translation tools can enhance intercultural competence by facilitating meaning-making across and more participation in communication et al., 2026). research further suggests that communication and translation tools may participation for learners and facilitate engagement across and cultural & 2023). In this sense, generative AI systems and AI-supported translation tools may support more forms of intercultural participation, particularly for learners to and research further suggests that conversational generative AI may support the conditions through which such participation can et for that different conversational generative AI with changes in learners' to language and Similarly, et that AI-supported practice with and to and language further suggests that such may the AI-mediated interaction. that feedback with and and learners also to engage in communication including with and these situated within a broader learning and rather than treating as a for human interaction. these do not directly intercultural their suggest that AI-mediated interaction may opportunities for communication and may strengthen learners' and to participate in communication potential are particularly AI-mediated interaction access to communication for learners may remain the their educational is on how such technologies are and to human interaction. The present is therefore not AI can expand access to intercultural interactions, but how algorithmic mediation may shape the ways in which cultural difference is and within educational this as a while generative AI-mediated interaction can expand learners' exposure to diverse perspectives, it may also some of the ways in which this is Drawing on broader scholarship on datafication, prediction, and algorithmic we theoretically infer that generative AI systems and adaptive learning platforms may organise and through of prediction, and personalisation, potentially cultural difference more and to process, but also some of its ambiguity, and analytical to SEL is framed as a set of competencies, it is in the norms, and relationships are understood and (Mesquita et al., et al., From a theoretical perspective, feedback, or assessment are mediated through generative AI systems, may be or through algorithmic including of emotional and potentially to rather than relational cultural bias, generative AI systems and adaptive learning systems may reshape the conditions under which intercultural and SEL potentially some forms of encounter and opportunities for a may opportunities for disagreement and emotional that are important to relational more the of interaction. The AI may to participation while also what can and what and may facilitate communication while certain situated forms of disagreement, or are in more Research on machine translation has difficulties in cultural and while research on large language has cultural in that may et al., et al., 2024). do not that AI necessarily cultural difference in educational interaction. they the to examine accessibility is by interpretive a interaction in which different and cultural use AI-supported translation to what it to with may are less in the language of to contribute to the However, situated of respect, authority, or disagreement are into more in meaning may less by to the with their what they to and which of meaning are to In this AI participation without being as the on what concerns use a to for a about should cultural or The may and to its the as a set of or it may or perspectives that are less The therefore to the with other and perspectives are and the as a contribution rather than The educational is AI-supported or to for to a common AI can expand the conditions for participation while the interpretive conditions under which participation The is therefore not simply between AI use and human interaction, but between forms of mediation that meanings, perspectives, and open to and that or to conditions learners can or perspectives remain rather than being into a and are into collective where their assumptions and can be conditions a more basis for the theoretical developed education depends on plurality, participation, and the role of constructive disagreement in collective meaning-making (Biesta, 2022;Dewey, 1916Dewey, is important to that adaptive learning systems may educational particularly by support and in where educational are support through AI systems under appropriate pedagogical learners or to participate more fully in shared AI-mediated may also to explore or arguments collective may therefore refer either to a or to a pedagogical on the of is not inherently to democratic it may strengthen participation support to collective and The present is not with but with the possibility that educational interaction increasingly through optimisation logics that individual collective meaning-making and democratic Generative AI systems and adaptive learning platforms can some learning processes and optimisation, including to individual and (Turvey & Pachler, 2026). While such approaches may enhance and personalisation, we theoretically infer that, under some they may contribute to a of shared learning and potentially reduce opportunities for collective From this perspective, a potential of shared epistemic may learners' opportunities to engage with common and disagreement, which are important to democratic although generative AI systems can access to multiple perspectives on their and may also shape how or is and to Empirical research on generative AI has in including to perspectives and in the or of in the and context by et important but they also raise questions about how disagreement, and normative may be framed within AI-mediated educational In educational this may suggest the to algorithmic while to reduce may under some opportunities for sustained engagement with perspectives and disagreement. such is in of or it therefore raises the of and under what conditions educational spaces to support democratic encounter of interaction are mediated through and this algorithmic mediation may reshape not only the of knowledge but also the relational and emotional of interaction. learning is not simply a of regulation or skill acquisition but involves norms, and relational under conditions by difference et al., 2019). Consequently, under some forms of algorithmic mediation, SEL may being to individual rather than engagement with and disagreement. some forms of disagreement and emotional or under forms of algorithmic mediation, education may some of its capacity to as a of democratic (Biesta, integration of generative AI systems and adaptive learning platforms is new forms of educational mediation, with potential for IC, SEL, and democratic schooling et al., the basis of the theoretical algorithmic mediation may reshape of educational processes through the of perspectives, of some interactions, of epistemic authority, and changes in emotional and relational learning SEL and IC are relational and situated processes that are through opportunities for ambiguity, disagreement, interpretive openness, and engagement with difference (Mesquita et al., the generative AI and adaptive learning systems may educational and that AI can reduce communication to facilitate participation, and opportunities for and may be particularly AI as a for human interaction rather than as a for educational therefore depends on how they are embedded within and conditions that collective participation and human this perspective, the theoretical contribution of this is to the educational of generative AI and adaptive learning systems not primarily in of technological but in of the conditions under which algorithmic mediation participation, and democratic encounter. While the present has on generative systems and adaptive learning platforms that mediate educational interaction through personalisation, prediction, and automated feedback, their educational consequences are understood as rather than predetermined. research on has that such approaches can enhance communication, collaboration, social skills, and through participatory and learning approaches et al., 2026). The therefore conditions through which these relational consequences may across educational further of this is that it does not between educational research should therefore examine how AI mediation intersects with the and pedagogical conditions of and education. research examine research how AI-mediated translation and learners' of situated including ambiguity, and disagreement remain or are as more Second, research examine how the of the of perspectives, and opportunities for collective shape participation, perspective-taking, and the negotiation of meaning in interaction. Third, research across educational and explore how mediation and influence these including AI-supported to more collective or to on research the without that of AI use educational this perspective, the democratic is not should or generative AI and adaptive learning systems, but how AI-mediated can be that support rather than learners' opportunities for collective participation, dialogue, disagreement, and intercultural encounter. This calls for approaches that examine not only what AI systems can but their are embedded in their how their are and how their use is within educational In democratic of educational AI can be through participatory and that on automation, and technological and for educational the theoretical framework developed in this paper into and The first therefore broader theoretical the and the are not to reproduce the as they their into domains that can educational AI and may be through on and as appropriate to the and for AI with the relational and democratic purposes in this are not as but as of that educational the and of AI-mediated In should between that may be automated and that human oversight, and the possibility of the this includes the of shared teachers, and in about the adoption, and of AI the are to what are how they are and under what conditions they can be or with frameworks that emphasise accountability, and human in educational AI systems, particularly generative AI systems and adaptive platforms & of the 2022). the for educational AI is therefore not simply to what technologies can but to that their integration to the cultural, and democratic purposes of education.
Language: English
Subfield: Education
Open Access status: gold
Estimated APC paid: $3,801
OpenAlex ID: W7235395325
FWCI:
Cites: 34
Cited by:
Related to: 10