Higher Education Under Generative AI: Biographical Orientations of Democratic Learning and Teaching

Abstract

Generative artificial intelligence (AI) is reshaping higher education (HE) by reconfiguring how knowledge becomes visible, how judgment is exercised, and how recognition is distributed. These systems intervene in the pedagogical and democratic conditions under which plurality, critique, and participation can be sustained. This study examines how students and lecturers interpret and navigate these transformations and what they reveal about the possibilities of democratic education under algorithmic mediation. Drawing on n = 151 written articulations (122 students, 29 lecturers) to open-ended questions collected via LimeSurvey, analyzed through Grounded Theory in combination with biographical interpretation and oriented by education theory (Bildung) and democracy pedagogy, the research reconstructs five orientations that range from pragmatic coping to struggles over recognition. These orientations illuminate how systemic dynamics of acceleration, opacity, and infrastructural authority are refracted into everyday academic practice. They are further synthesized into three broader axes of temporal sovereignty, epistemic opacity and accountability, and recognition ecologies. The findings highlight how fragile orientations emerge as both risks and resources. The study contributes to HE didactics by outlining strategies to transform fragility into pedagogical occasions, emphasizing reflective delay, dialogical engagement with opacity, and diversification of recognition practices. It concludes that democratic education depends on cultivating spaces where algorithmic pressures become educable and fragile orientations can develop into dispositions of reflexivity, critique, and participation.

1. Introduction

In higher education (HE) today, few learning situations remain untouched by generative artificial intelligence (AI). A student preparing for a seminar may rely on ChatGPT, receive algorithmically generated feedback on an assignment, or follow a learning pathway curated by platform recommendations (Hummel & Donner, 2023). Such systems do not simply support learning; they intervene in what becomes visible, which arguments are recognized, and how academic discussions unfold (Bokelmann, 2023; Hummel, 2021). In this way, they reshape fundamental processes of orientation and judgment that are not only cognitive but also pedagogical and democratic, as they determine how plurality, critique, and participation can still be cultivated under digital conditions (Richter, 2023).

The societal implications of generative AI are already the subject of extensive debate. Recent assessments warn of the distortion of public discourse, the proliferation of synthetic media, and the concentration of power in global technology providers (Nentwich et al., 2025). Critical research in education and media studies underscores that algorithmic infrastructures do not merely reflect existing practices but actively generate categories and conditions of recognition (Amoore, 2020; Couldry & Mejias, 2019; Selwyn, 2019). These dynamics, often analyzed in the domains of politics and governance, find concrete parallels in HE: narrowing of debate, opaque structuring of learning trajectories, and growing institutional dependencies on external providers (Knox, 2020; Williamson, 2017). At the same time, these dynamics intersect with the fragile, biographically embedded transitions of students who negotiate entry into study programs, shifting orientations, and the uncertainties of longer educational trajectories (Alheit, 1992, 1993; Marotzki, 1990). HE is therefore not only an institutional field but also a biographical space where processes of formation are structured by algorithmic conditions. Universities thus occupy a hinge position: they are simultaneously exposed to the systemic risks of generative AI and tasked with cultivating the reflexive and democratic capacities needed to confront them. This dual role becomes particularly evident in teaching contexts where AI systems no longer remain in the background but co-determine what is taught, discussed, and assessed. The challenge is therefore not only to integrate new technologies but to ask how HE can continue to serve as a space in which political thinking, deliberation, and socially reflected action are possible under algorithmic conditions. The idea that democracy requires education has a long tradition in political and educational thought. Negt (2004) emphasizes that democracy is the only form of government that must be actively learned and appropriated by its citizens, since it depends on their capacity to participate, to judge, and to act responsibly. In this sense, democracy is not a self-sustaining societal order but a living practice that must be continuously renewed through education and collective experience. Himmelmann (2007) builds on this view by stressing that democratic life relies on people who, through educational engagement and shared practice, sustain and reshape its foundations in ever-changing conditions. These demands become even more pressing in view of diversity and intersectionality. Algorithmic infrastructures do not affect all students equally but interact with patterns of national-ethno-cultural belonging, gendered experiences, and social inequality (Bourdieu, 1987; Fraser, 2003; Mecheril, 2003; Mecheril & Seukwa, 2006). Recognition and visibility are unevenly distributed, and digital mediation can both reinforce and destabilize these asymmetries. Such dynamics recall the problem of epistemic injustice, where credibility and interpretive authority are unequally assigned, leaving some voices systematically marginalized or unheard (Fricker, 2007). Addressing democratic education in the digital condition therefore calls for an approach that links institutional and pedagogical analysis with biographical orientations and questions of justice. Universities are thus challenged to sustain spaces of democratic formation where algorithmic mediation becomes a subject of critical reflection rather than an unseen structuring force. This article responds to that challenge by combining an education-theoretical perspective with empirical analysis. Conceptually, it connects debates on AI and democracy with pedagogical traditions of reflexivity and subject formation, acknowledging the entanglement of democracy theory and democracy education (Berkemeyer & May, 2023). Empirically, it examines how students and lecturers interpret and navigate algorithmic mediation in their everyday practices.

The central analytical focus of this study is to examine how the interpretations and orientations articulated by students and lecturers reflect the ambivalences of algorithmic mediation while at the same time indicating how democratic education can be sustained under digitally reconfigured conditions of learning and teaching. Against this background, the study is guided by the following research question: How do students and lecturers in HE interpret and navigate the presence of generative AI in their academic practices, and what do these orientations reveal about the conditions and possibilities for sustaining democratic education under algorithmic mediation? To address this question, the next section develops the theoretical framework that situates Bildung, biography, democratic education, and algorithmic mediation as the conceptual foundations of the study.

2. Theoretical Background—Education-Theoretical Perspectives on Democracy Under Generative AI

Democratic education under digital conditions requires a theoretical framework that reaches beyond the transmission of competencies. At its core, education can be understood as a reflexive process of Bildung1, in which subjective orientations, institutional practices, and societal structures interact. Dewey’s conception of democracy as a way of life remains a crucial reference point. He emphasizes that education is not only a means of acquiring knowledge but a collective practice through which citizens learn to negotiate shared problems (Dewey, 2011; Gutmann, 1999). In this tradition, democratic education is not conceived as the achievement of a fixed end state but unfolds as a dynamic and fragile process in which individual self-assurance, societal participation, and technological mediation intersect (Fauser, 2022; Habermas, 1992). Klafki’s concept of the capacity for responsible self-determination2 frames this process as the cultivation of orientation in contexts of uncertainty, while Biesta’s triad of qualification, socialization, and subjectification clarifies that education encompasses multiple and sometimes conflicting functions (Biesta, 2013). Under algorithmic conditions, these functions are reconfigured: qualification through automated assessment, socialization through platform-mediated participation, and subjectification through encounters with opaque infrastructures. Biesta (2013) also stresses that democracy involves exposure to the unpredictable and the unfamiliar. Rancière (1991) deepens this perspective by showing that democracy is enacted through disruption, contestation, and the insistence on equality. While Habermas grounds democratic education in communicative rationality and the pursuit of consensus, Rancière emphasizes disruption and radical equality. Read together, these perspectives reveal a field of tension between consensus and dissensus that becomes especially pronounced in digital settings, where infrastructures mediate both visibility and forms of participation. From this perspective, democratic education can be understood as cultivating epistemic friction, where encounters with opacity, disagreement, and disruption become productive moments for reflexivity and collective learning (Medina, 2013).

This democratic horizon also involves a biographical perspective. Educational trajectories are not linear but marked by transitions, ruptures, and contingencies. Alheit’s notion of Biographizität3 conceptualizes learning as the ongoing negotiation of self and world across discontinuities and turning points (Alheit, 1992, 1993, 2003b). Marotzki (1990) emphasizes that Bildung emerges in processes where individuals reconfigure their self-relations and interpretive horizons in response to societal challenges. Biographical research has repeatedly shown that educational processes unfold as dialogical negotiations between subjective meaning-making and institutional frameworks (Dausien & Kelle, 2005; Egger, 1995, 2006). Comparable insights appear in international adult learning theory, where Merriam et al. (2007) stress the embeddedness of learning in the broader life course, and in Marsick and Watkins’s (1992) work on informal learning, which highlights the contingent and unplanned character of formative experiences. Under algorithmic conditions, these negotiations are refracted through infrastructures that filter recognition, accelerate decision-making, and impose classificatory logics. What Alheit described as fragile and transitory processes connects closely with international debates on biographical learning (Tedder & Biesta, 2009; West et al., 2007) and can be related to Mezirow’s (2000) concept of transformative learning as a reorientation of meaning perspectives. These perspectives situate biographical research within a wider international discourse on formation and reorientation, highlighting how individuals negotiate continuity and change under disruptive conditions. Democratic education is also closely linked to diversity and intersectionality. (Mecheril, 2003; Mecheril & Seukwa, 2006) shows how natio-ethno-cultural belongings shape educational experiences, while Dausien (2004) highlights the entanglement of gender and biography. Fraser’s (2003) dual focus on redistribution and recognition indicates why democracy involves not only access but also symbolic orders that determine whose voices are heard. Bourdieu’s (1987) analysis of habitus and distinction illustrates how educational fields reproduce hierarchies that may be reinforced by digital mediation. Critical race theory provides further insights into how systemic inequities shape educational trajectories, while Ahmed (2012) demonstrates how institutional cultures of diversity can simultaneously recognize and marginalize minority voices. Together, these perspectives describe HE as a site where biographies, inequalities, and identities intersect and where algorithmic infrastructures may stabilize or unsettle these dynamics.

With the rise of generative AI, these dynamics acquire additional layers of complexity that extend beyond technical questions. Algorithmically mediated platforms structure visibility, prioritize information, and channel participation in ways that are not neutral but embedded in social and cultural contexts (Amoore, 2020; Helsper, 2021; Selwyn, 2019). Knox (2020) shows that HE is increasingly shaped by algorithmic infrastructures that mediate access to knowledge while preconfiguring futures of learning, often privileging efficiency and prediction over reflexivity and dialogue. In practice, personalization, feedback systems, and gamification introduce subtle yet pervasive forms of guidance that shape how students orient themselves in learning environments (Hummel et al., in press). Many of these processes remain opaque, producing what Busch (2016) describes as a normalization of algorithmic influence. Pasquale’s (2015) analysis of the ‘black box society’ highlights how algorithmic decision-making often eludes scrutiny, raising questions of accountability that affect educational legitimacy. At a societal level, these dynamics mirror broader risks such as discursive narrowing and dependency on global infrastructures (Nentwich et al., 2025). As Couldry and Mejias (2019) argue, such developments culminate in a form of data extractivism that transforms participation itself into a resource to be appropriated and monetized. What appears as engagement and inclusion thus simultaneously reproduces relations of dependency, revealing how the logics of data capitalism extend into the very practices through which knowledge and education are produced. This perspective highlights how algorithmic mediation not only organizes knowledge flows but also implicates universities in larger structures of power and inequality.

From a pedagogical perspective, these insights align with longstanding debates on democracy education. Himmelmann (2001) conceives democratic learning as fostering the capacity for judgment and the ability to engage constructively with plurality and conflict. Kenner and Lange (2020) further argue that such learning depends on institutional and pedagogical conditions that render participation both possible and meaningful as a lived and reflective practice. In HE, algorithmic infrastructures intervene in both dimensions by reshaping the conditions under which judgment and participation can unfold. They influence how recognition, authority, and accountability are distributed, thereby transforming the spaces in which democratic dispositions are formed and sustained. The pedagogical challenge lies in rendering algorithmic infrastructures visible and discussable, enabling students and lecturers to critically interrogate their operations. This connects to traditions of critical pedagogy (Brookfield, 1987; Freire, 1970; Giroux, 2011; Hooks, 1994), which highlight the importance of addressing power, participation, and reflexivity in educational practice.

These discourses form a shared horizon rather than a catalogue of separate traditions. Democratic education under generative AI is not reducible to technical adaptation or the delivery of competencies but appears as a fragile interplay of Bildung, biography, and mediation, in which Bildung theory emphasizes orientation in situations of uncertainty where judgment remains provisional and contested, biographical perspectives show that such orientations are embedded in life histories marked by transitions, inequalities, and intersecting identities, and analyses of algorithmic mediation expose how infrastructures intervene by structuring recognition, redistributing authority, and delimiting participation. Seen together, these strands clarify how democratic education in HE unfolds as cultivating dispositions of reflexivity, critique, and participation—dispositions that are biographically situated, attentive to diversity, and responsive to the conditions of digital mediation. To situate these perspectives within an integrated conceptual lens, this article draws on the DEA Framework (Democratic Education under Algorithmic Conditions) previously introduced by the author (Hummel, 2025). The DEA model conceptualizes HE as a triadic field constituted by Education, Democracy, and Algorithmic Conditions. At the first corner, Education (in the sense of Bildung) denotes processes of qualification, socialisation, and subjectification in Biesta’s sense, with a particular emphasis on the cultivation of judgment, orientation, and responsible self-determination under conditions of uncertainty. At the second corner, Democracy refers to participation, recognition, and contestation, drawing on Dewey’s idea of democracy as a way of life, Habermasian notions of deliberation, and Rancièrian emphases on disruption and equality. At the third corner, Algorithmic Conditions capture the socio-technical infrastructures that filter visibility, classify subjects, and preconfigure futures of learning, including platform architectures, generative AI systems, and data-driven governance. The DEA framework (see Figure 1) does not treat these three corners as separate domains, but as a relational field in which subject-formation is continuously shaped by competing claims of efficiency, legitimacy, and recognition. Within this triangular space, pedagogical practices, democratic aspirations, and algorithmic operations intersect and sometimes collide. Recognition can be redistributed from dialogical encounters into infrastructures; responsibility can shift from professional judgment toward automated procedures; legitimacy can be negotiated between institutional norms, civic expectations, and opaque system outputs. In this sense, the DEA model provides a mid-level theoretical structure that makes visible how Bildung, democracy, and digital governance are co-implicated in HE.

In the present study, the DEA framework serves as the conceptual architecture that links the theoretical background to the empirical reconstruction of orientations. The three corners of the model orient the analysis toward educational-theoretical questions of formation and judgment, democratic questions of participation and recognition, and structural questions of algorithmic mediation and governance. The five reconstructed orientations are thus not interpreted in isolation but situated within this triangular field, as empirical condensations of how students and lecturers navigate tensions between Bildung, democracy, and algorithmic conditions in their academic lives. In subsequent sections, the DEA framework informs both the coding of articulations and the synthesis of findings into three analytical axes that structure the study’s educational and pedagogical implications. In this way, the model does not stand alongside the empirical analysis but underpins it, providing a coherent lens through which the biographical, democratic, and infrastructural dimensions of generative AI in HE can be read together.

3. Methodology

3.1. Research Design and Methodological Positioning

This study is based on a qualitative and exploratory design, drawing on Grounded Theory (Charmaz, 2014; Strauss & Corbin, 1996). The purpose was to reconstruct how students and lecturers make sense of generative AI and how they orient themselves in relation to democratic education. Grounded Theory is especially useful in areas where practices are new and theoretical frameworks are still emerging. It makes it possible to generate categories inductively while staying in dialogue with existing concepts. The focus was not on statistical representativeness but on analytic generalization and transferability, i.e., whether the findings can illuminate comparable contexts beyond the study sample. The study follows an interpretive paradigm that foregrounds meaning-making from the participants’ perspective. In this logic, written responses to open prompts were not treated as survey data in a narrow sense but as articulated self-positionings through which participants locate themselves within AI-mediated educational trajectories. At the same time, the analysis was not ‘theory-free’: sensitizing concepts from education theory, democracy pedagogy, and governance research (Blumer, 1954; Sander, 2023) served as orientation points. This balance was necessary because researching AI in education involves a paradox: the phenomenon is new and must be reconstructed from lived experiences, yet these experiences cannot be fully understood without reference to established pedagogical and normative traditions.

3.2. Sampling Strategy and Data Collection

The empirical material consists of n = 151 written articulations generated through a LimeSurvey-based online questionnaire at two German-speaking universities. Of these, 122 responses were submitted by students (72 women, 50 men) and 29 by lecturers (16 women, 13 men). The questionnaire combined closed items to contextualize participants’ institutional position and digital practices with open-ended prompts designed to invite reflective and biographically embedded self-positionings regarding the role of generative AI in academic learning and teaching. Participants were recruited through institutional mailing lists, course-wide announcements, and postings in the learning management systems of the two universities. No incentives were offered. The two universities were selected because they differ in size, disciplinary profile, and digitalization strategies, which provided contrasting institutional contexts for reconstructing orientations toward generative AI. Sampling followed the logic of theoretical variation rather than representativeness. The aim was to access contrasting constellations of disciplinary background, academic role, digital affinity and orientations toward democratic education. Students were included as the primary actors addressed by digitalization and AI governance strategies in HE, while lecturers contributed perspectives on pedagogical responsibility and recognition practices under algorithmic conditions. The gender distribution was not operationalized as a variable for comparison but acknowledged as part of the heterogeneous institutional and biographical positions from which orientations toward AI are articulated. No exclusion criteria were applied beyond affiliation with one of the participating institutions.

To ensure transparency and replicability, the questionnaire included two parallel sets of open-ended prompts—one for students and one for lecturers—each designed to elicit biographically embedded articulations while remaining comparable across groups. The choice of open-ended, biographically oriented prompts follows the logic of reconstructive and education-theoretical research. Such questions are necessary to elicit narrative accounts of transitional moments, interpretive shifts, and recognition experiences that cannot be captured through predefined categories. This approach aligns with Marotzki’s (1990) conceptualisation of Bildung as reflexive self-world relations and with Alheit’s (1992, 1993) emphasis on Biographizität as the unfolding of orientation across critical situations and discontinuities. It thereby ensures that the data reflect situated processes of meaning-making rather than opinions detached from lived experience.

Students responded to the following items:

• When you look back on the course of your studies so far: Were there transitions, uncertainties, or phases of strain in which you used generative AI to gain orientation or to cope with demands? Please describe a concrete situation and how you interpret it in retrospect.
• How has your use of generative AI developed over the course of your studies? Do you use AI in a consciously reflective way, or is it more embedded in everyday routines—possibly even without knowing exactly how the systems operate in the background? Please describe a typical situation and how you assess its significance for your studies over time.
• Were there situations in which you considered suggestions or feedback from AI more strongly than feedback from lecturers or fellow students? Please describe what this meant for your feeling of recognition, trust, or uncertainty, and how you interpret this situation today.
• Over the course of your studies, have you encountered contradictory expectations or rules regarding the use of AI (e.g., strict prohibition in one course, implicit expectation of use in another)? How did you orient yourself in such situations? Please describe whether you felt included, equally involved, or overlooked—and how you interpret these experiences today for your studies.

Lecturers responded to a parallel but role-specific set of questions:

• Looking back on your professional experience as a lecturer: Were there transitions, periods of strain, or moments of pedagogical uncertainty in which you used generative AI to support decision-making or to regain orientation? Please describe a concrete situation and how you interpret it in retrospect.
• How has your use of generative AI developed throughout your teaching practice? Do you use AI in a consciously reflective way for designing your teaching, or is it more embedded in routinised processes—possibly without detailed knowledge of the underlying systems? Please describe a typical situation and how you assess its significance for your professional practice over time.
• Were there situations in which students relied more strongly on AI-generated feedback or assessments than on your own? Please describe what this meant for your experience of pedagogical authority, recognition, or resonance, and how you interpret this experience today.
• During your teaching experience, have you encountered contradictory institutional expectations or regulations regarding the use of AI (e.g., inconsistent guidelines across modules or study programmes)? How did you orient yourself in such situations? Please describe whether you experienced students as strengthened, unsettled, or irritated—and how you pedagogically interpret such dynamics today.

The open prompts were formulated to encourage more than descriptive statements by inviting respondents to situate their practices in relation to transitions, uncertainties and biographical turning points. They encouraged participants to situate their experiences in relation to transitions, uncertainties, and biographical turning points, thereby generating narratives through which orientations become empirically accessible. The design follows biographical research traditions that conceptualize educational orientations as developing across longer trajectories of Bildung as individuals negotiate institutional structures, expectations, and recognition regimes (Alheit, 1992, 1993; Marotzki, 1990).

3.3. Data Analysis and Category Development

Analysis followed the iterative logic of Grounded Theory coding, proceeding through open, axial and selective phases (Charmaz, 2014; Strauss & Corbin, 1996). In the open phase, the written responses were examined line by line, generating in vivo codes that remained close to the formulations used by participants. During axial coding, these codes were related to one another by examining conditions, practices and perceived consequences. In the selective phase, the most robust conceptual clusters were consolidated into core categories that expressed the central tension between algorithmic structuring and democratic judgment in academic life.

The analytic process was abductive rather than linear. Categories were continually compared with the material, and deviations were treated as productive irritations that enabled further conceptual refinement (Reichertz, 2003). Throughout the process, memos, comparative readings and schematic mappings were used to document interpretive decisions and to relate empirical reconstruction to theoretical reflection. Beyond procedural coding, the analysis employed interpretative condensations. Interpretative condensations refer to the analytic synthesis of recurrent meaning patterns across individual articulations into higher-order meaning structures. Rather than summarizing statements, this procedure integrates conditions, practices and perceived consequences into coherent conceptual units that connect biographical articulations with pedagogical and institutional contexts. These condensations served as a bridge between line-by-line coding and the development of the five orientations. This procedure aligns with reconstructive approaches in qualitative educational research where condensations are used to identify higher-order meaning structures across cases. Drawing on Marotzki’s (1990) structural theory of Bildung and Alheit’s (1993, 2003a) concept of Biographizität, the written articulations were not interpreted as isolated comments but as positionings situated within longer educational trajectories. Particular attention was given to passages in which respondents described moments of uncertainty, disruption or reorientation in their engagement with AI. These were treated as indicators of biographical sedimentation, revealing how orientations toward AI are linked to broader processes of negotiating competence, belonging and recognition in academic contexts. This approach resonates with international debates on biographical learning (Tedder & Biesta, 2009; West et al., 2007) and aligns with Mezirow’s (2000) conceptualization of transformative learning as the reframing of taken-for-granted meaning perspectives.

Theoretical saturation was monitored systematically. After approximately two thirds of the material had been coded, no additional responses yielded novel properties, relationships, or conceptual dimensions beyond those already captured in the emerging analytic frame. Subsequent cases served primarily to deepen and nuance the existing categories. This stabilization was taken as an indication that theoretical saturation had been reached in accordance with established criteria in Grounded Theory. Credibility was strengthened through iterative memoing and critical dialogue with the sensitizing concepts drawn from education theory, democracy pedagogy and governance analysis, ensuring that empirical reconstruction remained theoretically grounded without being predetermined by it.

3.4. Research Ethics and Quality Assurance

The study followed the ethical guidelines of the German Sociological Association (2017) and international standards such as the Declaration of Helsinki. At the beginning of the questionnaire, participants were informed about the objectives of the study, the use of the data and their rights regarding withdrawal and anonymization. Consent was given digitally before accessing the response fields, which is consistent with ethical standards for online qualitative research. All written responses were anonymized and stored securely in accordance with data protection regulations. Because biographically oriented articulations can touch on sensitive experiences such as inequality or discrimination, particular care was taken to prevent re-identification while preserving analytical richness. Quality assurance relied on several strategies. Coding decisions were documented transparently throughout the analysis, and a selection of written responses was cross-coded by a second researcher to enhance interpretive robustness. Rather than relying on communicative validation in a narrow sense, analytical claims were subjected to critical peer reflection to ensure that category development remained accountable to the material rather than predetermined by theoretical assumptions.

Reflexivity was maintained throughout the process by interrogating the researcher’s positionality in relation to the field, including disciplinary background, institutional embeddedness and normative assumptions. As the researcher was situated within the academic contexts examined in the study, methodological vigilance was required to avoid reproducing institutional logics in the interpretation. Ethical responsibility was understood in an expanded sense. In line with current debates on generative AI and democracy in education, the task was not only to protect participants but also to render algorithmic mediation itself visible as a matter of pedagogical concern. This aligns with the democratic and educational commitments of the study, which treats ethics not as procedural compliance but as an enabling condition for reflexivity, participation and shared deliberation in HE.

3.5. Methodological Contribution

The methodological approach makes two contributions. First, it demonstrates how Grounded Theory can be combined with biographical interpretation to reconstruct empirical orientations toward generative AI from written self-positionings, capturing them both as situated practices and as elements of longer-term educational trajectories. Second, it shows that researching AI in HE requires methodological reflexivity regarding recognition, power and democratic participation when orientations are articulated not in dialogical encounters but in digitally mediated formats. This perspective expands reconstructive qualitative research by treating written articulations in online environments as spaces of subjectivation rather than as reduced survey data. It thereby contributes not only to biographical learning and education theory but also to emerging debates on algorithmic governance in education (Knox, 2020; Williamson, 2017), positioning digital articulation itself as part of the empirical field. In this way, the study secures empirical rigor while opening a methodological horizon for the analysis of educational orientations, algorithmic mediation and democratic learning under digital conditions.

4. Results of the Empirical Study

Through the combination of Grounded Theory coding and biographical interpretation, the analysis reconstructed not just thematic clusters of statements but deeper orientations that shape how students and lecturers engage with generative AI. In line with biographical research (Alheit, 1992; Marotzki, 1990), orientations are understood as patterned ways of coping, interpreting and positioning that are embedded in longer educational trajectories rather than isolated practices. These orientations crystallized across the written articulations as responses to the institutional contradictions of HE and are marked by distinct biographical, pedagogical and democratic implications. In the following, five such orientations are presented, each grounded in excerpts from the written responses and elaborated through dialogue with education theory and democratic pedagogy. They are not conceived as fixed empirical themes but as interpretative condensations, developed through open, axial and selective coding. The orientations capture both immediate practices and broader processes of subjectivation under algorithmic conditions. They span different levels, from pragmatic strategies of coping with workload to struggles over recognition and authority, reflecting the multi-layered ways in which academic learning is reconfigured. Each orientation is empirically reconstructed and analytically enriched through educational-theoretical interpretation, ensuring that the results remain faithful to participants’ articulations while highlighting their wider significance for HE.

4.1. Pragmatic Orientation: Coping with Workload

The first orientation reconstructed from the written responses centers on how students and lecturers pragmatically cope with the workload and accelerating demands of university life, using generative AI primarily as a means to intensify existing strategies of dealing with pressure. One student respondent (S114) wrote: “Even before AI, I often skipped readings or just skimmed them. Now I put the text into ChatGPT and get a summary, it feels like cheating time.” This articulation shows how familiar coping practices persist but take on new shape through technological mediation. At the same time, it captures the vulnerability of early study phases, when students seek stability and belonging while navigating new academic demands (von Felden, 2006). Another participant (S22) admitted: “I used to double-check facts myself, but with ChatGPT I just take the answer. Sometimes I notice I rely on it too much.” Here dependency does not appear as fascination with AI but as reliance on its usefulness in coping with overload. Such articulations can indicate biographical turning points, where routines are momentarily destabilized and the possibility of reflexive reorientation emerges (Dausien, 2004; Schütze, 1984). A lecturer respondent (L9) confirmed this tendency from a teaching perspective, noting that students increasingly turn directly to AI tools, which creates habitual patterns in which reflection becomes secondary (Hummel & Donner, 2023).

Across these articulations, three dimensions can be distinguished. First, in moments of heightened stress such as exam preparation or assignment deadlines, AI functions as a situational resource for saving time. Second, students describe epistemic shortcuts, where system outputs are accepted without verification and reliance shifts from interpretive judgment to statistical regularities. Third, repeated use normalizes this reliance, embedding AI as an invisible part of academic practice. These tendencies continue earlier strategies of coping with institutional pressure, such as delegating tasks or routinizing learning steps (Alheit & Hoerning, 1989). Outsourcing entire readings through AI-generated summaries reflects such sedimentation of shortcuts into normalized routines. By contrast, articulations that express unease illustrate how routines can be interrupted and turned into occasions for self-reflexivity. These differences reveal how coping is embedded in biographical trajectories shaped by study entry, disciplinary socialization and future projects (Alheit, 1993, 2003a; Marotzki, 1990). They also intersect with habitus and cultural capital: while some students mobilize AI strategically as an additional resource, others risk dependency when their biographical and disciplinary resources are less robust (Alheit, 2008; Bourdieu, 1987). Lecturers’ observations further show how disciplinary traditions mediate the reception of digital tools, with hermeneutic cultures of justification being particularly vulnerable to algorithmic acceleration. A temporal dimension sharpens this interpretation. The acceleration enabled by AI-generated outputs reflects coping practices situated within broader societal dynamics of acceleration, where students attempt to stabilize themselves within compressed schedules (Rosa, 2024), while resonance is simultaneously undermined as efficiency displaces dialogical engagement with knowledge. From a pedagogical perspective, this orientation reveals structural tensions. Articulations that describe ‘cheating time’ and concerns about displaced reflection highlight the risk of narrowing academic learning to instrumental rationality. What is lost is not only reflection but also the relational work at the core of teaching, namely the pedagogical labor of building dialogical relationships between teachers and students (Szczyrba, 2009; Wedekind, 1986). Pragmatic coping may bring temporary relief by allowing students to redistribute energies, yet it simultaneously contracts the horizon of Bildung, which entails cultivating the capacity for responsible self-determination (Klafki, 1996) and responding to the educational call into responsibility (Biesta, 2006). Still, moments of hesitation articulated in the responses demonstrate how irritation can serve as a transformative occasion that destabilizes routines and reopens reflexive engagement (Koller, 2012). The democratic implications extend beyond the classroom. When pragmatic coping becomes normalized, dialogical spaces for contestation and plural interpretation shrink as instantaneous answers replace deliberation and reduce opportunities for participation. At the same time, coping is not equally distributed. Students with greater cultural capital or stronger biographical resources may use AI strategically, while others risk dependency. This inequality reflects struggles over recognition and redistribution, where infrastructures silently privilege some groups while disadvantaging others (Fraser, 2003; Honneth, 2012). Pragmatic coping should therefore not be dismissed as a mere deficiency but understood as a crystallization of biographical fragility, institutional contradiction and systemic acceleration. The pedagogical task lies in treating such orientations as occasions for reflexive engagement, where the tensions between efficiency and reflection, delegation and autonomy and algorithmic mediation and Bildung become visible and discussable. In this sense, coping practices are not only risks for democratic education but also potential entry points for cultivating judgment, sustaining relational engagement and maintaining spaces of participation in HE.

4.2. Adaptive Orientation: Learning Under Opacity

A second educational orientation reconstructed from the written responses can be described as an adaptive orientation toward opacity. Respondents repeatedly emphasized their limited insight into how generative AI systems operate and expressed little willingness to engage with this complexity. One student (S27) wrote: “Already in school I never wanted to go too deep into how technologies work. Now at university it is the same. I honestly do not want to know what happens in the background of ChatGPT. It is too complicated, and as long as it works for me, that is enough.” This articulation illustrates a biographical continuity of orientation: what was rejected in earlier educational phases is continued in HE. Rather than signalling indifference, this stance reflects a protective strategy that secures stability by sidestepping complexity. Another student response (S41) stated: “In my studies I have often learned to live with uncertainty. With AI it is similar. I have the feeling that it influences me more than I notice. But you cannot really see it, so you just continue using it.” Here opacity is acknowledged, yet the statement reflects a broader adaptation process in which opacity is normalized as part of everyday study routines.

Lecturer responses (L5, L16) expressed similar attitudes, noting that AI had been integrated into teaching practices even though its internal logic remained unclear. For some, adaptation took the form of pragmatic withdrawal, described as focusing on other tasks rather than questioning the system in depth. These articulations suggest that adaptive orientations are not confined to students. Faculty members, confronted with competing institutional demands, may also develop strategies of disengagement that function as forms of protective non-knowledge. The recurrent expression that it is “enough” if the system works illustrates how educational orientation can contract into a purely functional stance. In Klafki’s (1996) terms, the demand for responsible self-determination becomes muted when confrontation with societal challenges is bypassed. Biesta’s (2006) notion of being summoned into responsibility provides a useful lens here, yet in these responses this summons appears softened or deferred by the pragmatic need for stability. Some articulations also expressed unease, such as the awareness of being influenced by AI “more than I notice.” Such moments indicate irritation in the sense described by Koller (2012), where habitual routines lose their reliability and the potential for reflexive reorientation emerges. When lecturers noted that they had “stopped asking how AI works” the analysis shows how decision-making authority subtly migrates into infrastructures. Fricker’s (2007) concept of epistemic injustice clarifies how such withdrawal risks producing uneven participation, as the credibility of human judgment is displaced by opaque system outputs. Honneth’s (1996) analysis of recognition further explains why this shift is not merely technical but normative: recognition is undermined when dialogical justification is replaced by infrastructural authority. The responses also highlighted that opacity extends beyond the technical operations of AI to include the implicit structures of academic life. Jackson’s analysis of the hidden curriculum (Jackson, 1968) helps make visible how unspoken expectations govern engagement with AI: what respondents withdraw from is less machine learning logic and more the contradictory institutional environment in which AI is at once prohibited and encouraged, demanded and disavowed. This produces a mode of non-knowledge that brings short-term relief while reproducing uncertainty.

From the perspective of tacit knowledge, the responses offer an additional nuance. Polanyi (1966) described tacit knowledge as the background resource that supports skilled action, yet what becomes visible here is not tacit mastery but a form of tacit non-knowledge. Rather than drawing on stable interpretive schemes, respondents rely on functional routines that secure temporary orientation without deeper understanding. This can be read as a fragile coping strategy in light of institutional contradictions. Alheit and Dausien’s work on biographical knowledge (Alheit, 1993, 2003a; Dausien, 2004) clarifies how such fragile routines may be biographically motivated attempts to maintain continuity. Egger’s analysis of the tension between institutional expectations and individual meaning-making (Egger, 1995, 2006) helps interpret these orientations as situated between personal biography and systemic pressure. These dynamics are also linked to social differentiation. Drawing on Bourdieu’s concept of habitus (Bourdieu, 1987), the responses suggest that students with more robust institutional confidence experience opacity as unproblematic, while others with more fragile biographical resources perceive disengagement as their only viable strategy. These articulations reveal how recognition, belonging and legitimacy are unevenly distributed when the hidden curriculum of AI use remains contradictory and opaque.

From the standpoint of HE didactics, the findings reflect Reinmann’s analysis of converging institutional imperatives of efficiency, digitalization, competence orientation and research excellence (Reinmann, 2025). These conditions generate pressures in which disengagement becomes an attractive form of relief. Reinmann distinguishes between educational quality as the normative horizon of Bildung and learning effectiveness as the measurable output of instructional design (Reinmann, 2016). When respondents state that it is sufficient for the system to function, effectiveness is implicitly privileged over quality, narrowing education to immediate usability while sidelining dialogical and critical dimensions. Building on education theory, opacity must not be treated only as a deficit but also as a pedagogical challenge. Klafki emphasizes that Bildung requires confronting societal contradictions rather than bypassing them, while Biesta insists that education involves being summoned into responsibility. Moments of irritation, as articulated in the responses, indicate that this call can re-emerge even within adaptive routines. From a democratic perspective, the adaptive orientation reveals that opacity produces uneven conditions for participation. Only some respondents articulate strategies to navigate complexity, while others withdraw. Fricker’s concept of epistemic injustice clarifies how this withdrawal threatens equal participation, while Fraser’s reminder that recognition and redistribution are interlinked highlights that opacity redistributes symbolic legitimacy as much as access to knowledge. The adaptive orientation therefore reveals opacity not only as a coping strategy but as a pedagogical site that demands new forms of educational response. HE cannot restrict itself to providing factual knowledge about AI but must cultivate reflexive capacities that enable students and lecturers to treat opacity as a shared condition of academic life. Following Reinmann’s call for a humane university, the task lies in acknowledging adaptive orientations without dismissing them as deficiencies while transforming them into occasions for critical judgment, reflexivity and democratic participation.

4.3. Relational Orientation: Authority and Resonance

A third educational orientation reconstructed from the written responses can be described as relational, as it revolves around how evaluative authority and affective resonance shift when generative AI becomes part of academic routines. What emerges is not a simple replacement of teacher judgment with automated scoring but a redistribution of credibility, responsibility and recognition across institutional and infrastructural contexts, often intensified by the accelerating pressures of workload in HE (Rosa, 2024).

One lecturer (L12) stated: “In the past I always insisted on discussing feedback with students, but now I notice that some trust the feedback they receive from the system more than mine. They even cite it as if it were unquestionable.” Placed against a longer teaching biography shaped by hermeneutic traditions, this articulation illustrates how algorithmic promises of objectivity can erode dialogical justification. What is experienced here as a loss of resonance in pedagogical interaction points to a redistribution of evaluative authority under institutional conditions where speed and efficiency dominate (Reinmann, 2016). The tendency to treat system-generated feedback as unquestionable signals a transfer of legitimacy into infrastructures, exemplifying what may be called transferred authority, in which the space for dialogical negotiation narrows. Another lecturer (L7) articulated a similar tension: “I sometimes rely on automated scoring because it helps me appear consistent and objective, especially under pressure. But at the same time, I feel I have surrendered part of my judgment.” Rather than signalling individual weakness, this response makes visible the structural contradictions of HE, where institutional imperatives of fairness and efficiency push lecturers toward risk-averse strategies. This reflects what Egger (1995, 2006) describes as securing objectivity through procedural stability, even when it narrows the formative horizon of assessment (Wildt, 2013). Viewed biographically, the feeling of having ‘surrendered judgment’ resembles what Schütze (1984) conceptualizes as a biographical turning point, momentarily unsettling routines and opening space for reflexive reorientation (Dausien, 2004).

From the student perspective, one response (S18) noted: “In school I always depended on teachers’ evaluations. Now with AI it is similar. If the system says my essay is fine, I assume it is fine. I do not question it further.” This reliance on algorithmic validation exemplifies a credibility allocation dynamic, in which epistemic legitimacy is more readily conferred on algorithmic outputs than on human judgment. Biographically, this extends earlier experiences of externalized authority from school to digital infrastructures in HE (Alheit & Hoerning, 1989). Yet the explicit remark “I do not question it further” signals irritation that unsettles habitual patterns, pointing to reflexive potential. A lecturer with a migration-marked biography (L3) added: “When I was a student, I often noticed that my contributions were questioned more quickly. Now, as a lecturer, I sometimes see that students accept algorithmic feedback more readily than my comments. It feels as if the system’s voice counts as more neutral than mine.” This articulation exposes how algorithmic authority intersects with existing hierarchies of recognition. What appears as neutrality can reproduce established asymmetries, as recognition is redistributed along gendered and migration-related lines. This resonates with Honneth’s (1996) argument that recognition depends on dialogical justification, Fraser’s (2003) insistence on the inseparability of recognition and redistribution, and Besand’s (2016) observation that ideologies of inequality persist where some voices remain less legitimate.

Viewed across responses, these articulations do not represent isolated decisions but continuities sedimented along educational trajectories. Feelings of lost judgment, uncritical reliance or diminished credibility illustrate how agency and structure remain intertwined (Krüger & Marotzki, 2006), how biographies draw reliability from institutional frameworks (Kohli, 1985, 1986) and how fragility becomes visible in transitional phases (von Felden, 2006). Algorithmic evaluation inscribes itself into these trajectories by producing new textualities of recognition (Schulze, 1999, 2006) and narrowing opportunities for self-thematization and reflexive positioning (Hahn, 1987, 1988).

Pedagogically, the relational orientation reveals an ambivalence that is simultaneously adaptive and precarious. On one hand, automated systems provide relief under conditions of acceleration, enabling lecturers and students to redistribute their energies (Williamson, 2017). On the other hand, this relief narrows education to procedural fairness and instrumental rationality, sidelining relational work and dialogical justification. Bildung presupposes the cultivation of responsible self-determination (Klafki, 1996) and responsiveness to the call into responsibility (Biesta, 2006). Moments of disruption, such as articulations of lost judgment or diminished credibility, can serve as transformative occasions for renegotiating recognition (Koller, 2012). Didactic responsibility, as Apitzsch (2018) argues, lies in creating pedagogical contexts where evaluative authority is critically reappropriated rather than ceded to infrastructures.

From a democratic perspective, the implications are significant. Foucault’s (2006) concept of governmentality highlights how infrastructures guide conduct while redistributing authority in barely perceptible ways. When credibility is channelled into algorithmic systems, dialogical justification becomes weakened and symbolic legitimacy is restructured. These dynamics reinforce existing inequalities, particularly those linked to gender and migration. By reordering recognition under the guise of neutrality, algorithmic systems reshape the conditions of democratic participation (Besand, 2016; Fraser, 2003; Honneth, 1996). Rather than functioning as external tools, they become infrastructural media through which acceleration, workload and institutional contradiction are rendered effective. The relational educational orientation therefore shows how evaluative authority and affective resonance are reorganized when algorithmic mediation becomes embedded in HE. The task for pedagogy is not only to expose these shifts but to create spaces in which their ambivalences can be critically interrogated. Only then can Bildung and democratic participation be sustained under conditions where authority silently migrates into infrastructures.

4.4. Ambiguous Orientation: Improvisation and Fragility

A fourth educational orientation revolves around the pervasive experience of in-between spaces in which generative AI is situated within HE. What participants described was not primarily fascination with technology but rather the constant confrontation with contradictory expectations: in some courses AI was strictly prohibited, in others its use was implicitly taken for granted. Instead of clarity, students and lecturers reported a persistent condition of liminality, a space in which orientation had to be improvised from one day to the next.

One lecturer summarized this uncertainty pointedly: “During my career I have seen many reforms, and with AI it feels similar: there are official guidelines, but nobody can tell us what really happens in the background. It feels like we are improvising all the time” (L1). Her remark gains meaning when viewed biographically: her long career had already been marked by continuous reforms and shifting demands, so the introduction of AI became another moment of institutional fragility. The sense of “constant improvisation” reflects what Kohli described as the erosion of the institutionalized life course, in which stable structures normally provide reliability and orientation (Kohli, 1985, 1986). Where such structures are contradictory or opaque, academic practice no longer offers security but forces biographical navigation on unstable ground. Egger’s account of the tension between institutional expectations and individual meaning-making helps illuminate this dynamic, showing how lecturers like L1 are pushed into improvisation because institutions fail to provide clear frameworks (Egger, 1995, 2006). From the learner’s side, the picture is similar. One student explained: “In some seminars we are told not to use AI at all, in others it is almost expected that we use it for preparation. You never know what is actually allowed, and every decision feels risky” (S9). Her irritation makes visible how prohibition and expectation coexist in unstable ways, producing what Wimmer conceptualized as boundary-making, where institutions redraw the lines of legitimacy in contradictory forms (Wimmer, 2008). Instead of providing orientation, such shifting boundaries place students in zones of uncertainty where every decision entails the risk of being sanctioned. S9’s account reflects Bourdieu’s notion of habitus, as her ability to navigate these contradictions is shaped by her cultural capital and prior educational experience. For some, improvisation may be navigable; for others, it deepens vulnerability and exclusion. The fragility of these in-between spaces becomes especially clear in the account of S11, a first-generation student: “At the beginning of my studies I already had to figure out everything on my own, and now with AI it feels the same again, every teacher says something different, and I am never sure if I belong here” (S11). Her statement highlights how contradictions around AI re-activate earlier insecurities of transition and belonging. von Felden (2006) has shown that transitional phases are particularly vulnerable moments in which belonging and competence need to be reestablished. For S11, institutional ambiguity does not mitigate this fragility but deepens it, illustrating Turner’s (1969) notion of liminality as a state of suspension between stable roles. What she describes can be understood as a biographical turning point in Schütze’s sense, where irritation about ‘something not being right’ produces both dislocation and the potential for reflexive reorientation (Dausien, 2004; Schütze, 1984). Her oscillation between belonging and exclusion mirrors Breinig et al.’s notion of transdifference, in which individuals inhabit hybrid spaces that are neither fully inside nor fully outside (Breinig et al., 2002).

These three accounts together indicate that ambiguity in HE is not a marginal phenomenon but a structural condition in which improvisation becomes the dominant mode of orientation. Responsibility shifts away from reliable institutional frameworks and is relocated onto individuals, leaving them simultaneously inventive and vulnerable. L1’s continuous improvisation, S9’s navigation of contradictory boundaries, and S11’s fragile belonging each illustrate how opacity in AI governance reflects the hidden curriculum of academic life (Jackson, 1968). What remains concealed are not only the technical processes of AI but also the tacit institutional rules that structure legitimacy and orientation. As Polanyi (1966) noted, tacit knowledge normally provides stability by guiding action without explicit articulation, yet under algorithmic conditions this stabilizing function becomes fragile. Yet in these answers, what surfaces is not strong tacit knowledge but tacit non-knowledge: students and lecturers are left without reliable background resources, forced to improvise orientation where institutional guidance dissolves. Theoretically, this orientation demonstrates how ambiguity destabilizes both biographical and pedagogical processes. L1’s improvisation shows how the erosion of institutional reliability undermines the security of the academic life course (Kohli, 1985, 1986). S11’s fragile belonging exemplifies how transitions expose learners to heightened vulnerability (von Felden, 2006). S9’s oscillation between prohibition and expectation makes clear how institutions redraw legitimacy boundaries in contradictory ways (Wimmer, 2008). Egger’s tension between expectation and meaning-making (Egger, 1995, 2006) runs through all three cases, showing how individuals continuously balance institutional contradictions against their own sense of orientation. Schulze’s notion of biography as text (Schulze, 1999, 2006) and Hahn’s work on self-thematization (Hahn, 1987, 1988) help explain how algorithmic authority inscribes itself into life stories, limiting opportunities for reflexive positioning.

From the perspective of Bildung, these contradictions point to the fragile conditions under which responsible self-determination can develop. Klafki emphasizes that Bildung requires frameworks enabling learners to critically engage with central societal challenges (Klafki, 1996). Where frameworks dissolve into contradictions, responsibility shifts onto individuals who improvise without dialogical support. Yet as Koller argues, contradictions can also become transformative occasions that provoke reflection on agency within contingent environments (Koller, 2012). Benner’s view of Bildung as entangled with contingency further clarifies that institutional ambiguity makes this contingency sharply visible (Benner, 1991). The dimension of inequality sharpens this interpretation. Several female and migrant lecturers reported that algorithmic authority was sometimes accepted as more neutral than their own judgment. What presented itself as impartiality thus reinforced long-standing asymmetries of recognition. Ambiguity, in this sense, is not experienced uniformly: those with stronger biographical and cultural resources can navigate in-between spaces with greater security, while others remain more vulnerable. This asymmetry can be read through Honneth’s idea that recognition depends on dialogical justification (1996), Fraser’s argument that recognition and redistribution are mutually constitutive (Fraser, 2003), and Besand’s insight that ideologies of inequality endure where certain voices continue to lack legitimacy (Besand, 2016).

The democratic implications become visible in several ways. Dewey emphasized that institutions structure opportunities for participation (Dewey, 2000). Ambiguity weakens this structuring role, leaving participation distributed unevenly: those with greater cultural capital or stronger biographical resources can improvise effectively, while others risk exclusion. Rancière reminds us that democracy involves the enactment of equality, yet inconsistent frameworks reproduce asymmetries by privileging those already skilled in navigating contradictions (1999). Foucault’s concept of governmentality sharpens this perspective by showing how ambiguity itself can function as a mode of governance, compelling individuals to self-regulate in the absence of transparent rules (2006). The in-between orientation indicates that HE is not merely challenged by generative AI but also implicated in reproducing contradictions through ambiguous frameworks. Ambiguity appears less as a temporary deficit than as a structural condition that both restricts and provokes educational practice. Read through the vulnerability of transitions (von Felden, 2006), the tension between institutional expectations and meaning-making (Egger, 1995, 2006), and the liminality of in-between spaces (Turner, 1969), it becomes evident that ambiguity can simultaneously undermine and deepen Bildung. The task for universities is therefore twofold: to establish reliable frameworks that sustain responsible self-determination, and to cultivate dialogical spaces in which contradictions are addressed collectively as shared conditions of democratic education rather than endured as private burdens.

4.5. Recognition Orientation: Voice and Visibility

The final educational orientation reconstructed from the written responses concerns recognition and the micropolitics of visibility. Generative AI was described not only as a technical aid but as an active force that shapes whose voices are amplified, which contributions gain legitimacy, and whose perspectives remain marginalized. Students and lecturers repeatedly pointed out that visibility and credibility no longer emerge solely from dialogical interaction but are increasingly distributed by algorithmic infrastructures.

One student respondent (S63) explained: “On the platform, some posts get highlighted automatically and receive many comments, while others disappear without notice. I have seen that even thoughtful contributions sometimes vanish, and then nobody reacts. It changes what the group actually talks about.” Her irritation shows how infrastructures silently reorder attention, much like what Jackson described as hidden curricula, where unspoken structures guide participation and influence what counts as legitimate learning (Jackson, 1968). What disappears without explanation, as in her case, can gradually sediment into orientations of misrecognition that shape not only momentary interaction but also the long-term disposition to participate. Such dynamics illustrate Honneth’s account of recognition struggles, where the withdrawal of credibility erodes engagement and weakens the foundations of self-confidence (2012). Another student (S71) described a similar experience: “You can tell that the system privileges certain styles of writing. Short, polished, and formal. If you do not match that, you stay invisible, no matter how much effort you put in.” Her observation points to how algorithmic infrastructures set tacit standards that sort contributions into visible and invisible. Polanyi’s notion that much of practice relies on tacit knowledge (Polanyi, 1966) helps clarify the contrast observed here: tacit routines that usually provide security are replaced by what might be described as tacit non-knowledge, since the student lacks the background resources that would allow adaptive orientation.

The resulting invisibility becomes a formative experience, reviving earlier insecurities within her learning trajectory. Schütze’s concept of biographical turning points offers a useful lens for this process, showing how moments of misrecognition can unsettle habitual orientations and initiate new configurations of meaning (Dausien, 2004; Schütze, 1984). Whether such experiences deepen withdrawal or open paths of reflexive reorientation depends on the resources available within a biography. The lecturers’ perspective confirmed similar shifts. One respondent (L24) explained: “In discussions, students sometimes cite the automated feedback as more reliable than peer comments or even my remarks. It alters the hierarchy of voices in the classroom.” Her remark shows how algorithmic mediation reshapes legitimacy by positioning automated outputs as epistemically superior. Egger’s analysis of the tension between institutional expectations and individual meaning-making helps to situate this: what changes here is not simply personal authority but broader conditions of evaluation that redistribute credibility into infrastructures (Egger, 1995, 2006). Kohli’s account of institutional life courses adds another layer: when recognition frameworks become contradictory, uncertainty intensifies precisely at moments when orientation is needed (Kohli, 1986).

The structural dimension of these struggles was articulated even more sharply by L28, a female lecturer with a migration background: “When I was a student, I often noticed that my contributions were questioned more quickly than those of others, and I felt that both my gender and my migrant background shaped this experience. Now, as a lecturer, I sometimes see that my students accept algorithmic feedback more readily than my comments. It feels as if the system’s voice counts as more neutral than mine, even though I bring in expertise and personal experience.” Her account makes visible how algorithmic neutrality does not erase but can reinforce asymmetries of recognition. Butler’s insight that recognition is always framed by cultural intelligibility (Butler, 1997) clarifies why some voices remain less audible, even in environments claiming neutrality. Fraser’s reminder that recognition and redistribution are inseparably linked highlights how L28’s marginalization reflects broader inequalities along intersecting lines of gender and migration (Fraser, 2003). Yet the written responses also revealed counter-movements. Several students explained that they deliberately amplified the posts of peers whose contributions had been ignored, thereby enacting solidarity. Such practices show that recognition is not completely captured by infrastructures but can also be collectively renegotiated. Taylor argued that mutual acknowledgment is constitutive of identity (Taylor, 1993), while Tully (2004) emphasized that struggles for recognition can open democratic spaces of dialogue. Acts of amplification may therefore function as biographical resources of resistance, enabling students to counter invisibility and expand the repertoire of orientations in academic life.

Experiences of invisibility and misrecognition also carried a liminal dimension. One first-generation student (S76) recalled: “At the beginning of my studies I already had to figure everything out on my own, and now with AI it feels the same again, every teacher says something different, and sometimes I feel invisible.” Her words recall Turner’s idea of liminality, a suspension between established and emerging roles (1969). Positioned between recognition and misrecognition, her account illustrates both the risk of destabilization and the possibility of reflexive reorientation. Alheit’s notion of Biographizität helps to explain why such moments can either sediment as withdrawal or be reinterpreted as occasions for new orientations, depending on the resources students are able to mobilize (Alheit, 1993, 2003a). From an educational perspective, these findings highlight the fragile conditions under which recognition unfolds in HE. Benner (1991) emphasized that Bildung always takes place in contexts of contingency that demand reflexive judgment. Prange’s (2005) analysis of education as making visible clarifies why the invisibility described by S63 and S71 is pedagogically significant: when automated systems obscure contributions, formative dimensions of showing and recognition are displaced. At the same time, Dausien’s (2004) work reminds us that reflexivity often arises when recognition collapses, opening new interpretive horizons. What appears as misrecognition can therefore hold transformative potential if it is thematized and supported pedagogically. From a democratic perspective, these struggles raise pressing concerns. Fricker’s concept of epistemic injustice highlights that exclusion from credibility is not merely individual but systemic, shaping who can participate meaningfully in knowledge production (Fricker, 2007). Young (2000) stressed that democracy depends not only on access but also on assurance that all voices count as equally valid. The accounts of S63, S71, S76 and L28 illustrate how algorithmic infrastructures privilege certain linguistic styles and cultural forms, reinforcing inequalities that universities should not ignore. Rancière’s insistence that democracy rests on the enactment of equality is relevant here: when infrastructures silently redistribute visibility, that enactment is undermined (1999). Honneth and Fraser’s debate on recognition and redistribution underscores that both dimensions must be addressed together if democratic education is to be sustained (Fraser, 2001; Honneth, 2011).

The written responses suggest that recognition under algorithmic mediation is not reducible to questions of efficiency or access. What is at stake is how contributions become visible and legitimate, how recognition and misrecognition accumulate across biographies, and how these dynamics intersect with existing hierarchies. Recognition appears as biographically sedimented and democratically charged: subjects struggle for visibility, legitimacy and voice in algorithmically filtered spaces. The challenge for HE lies in making these struggles visible, negotiating them collectively and transforming them pedagogically into occasions for reflexive orientation and democratic participation.

5. Discussion: Bildung, Dispositions, and Democratic Orientations Under Generative AI

5.1. Re-Situating Empirical Categories in the Horizon of Bildung

The five reconstructed categories illustrate how students and lecturers interpret and navigate the presence of generative AI in HE. They express orientations that oscillate between relief and disorientation, trust and suspicion, delegation and resistance. From an education-theoretical perspective, these orientations can be read as fragile processes of subjectivation in which individuals confront societal key problems while negotiating autonomy and responsibility. The findings show that algorithmic infrastructures intervene in this horizon by shaping visibility, recognition, and evaluative authority. This illustrates how such processes unfold as ambivalent orientations marked by simultaneous relief and uncertainty, rather than as unequivocal gains or losses. These dynamics are biographically embedded. Alheit’s (1993, 2008) concept of Biographizität clarifies that learning orientations are sedimented across transitions and discontinuities, shaping how students engage with new demands at different stages of their studies. von Felden’s (2006) research on educational transitions demonstrates that study entry and progression are particularly fragile phases, in which experiences of recognition or misrecognition can significantly influence future orientations. Egger (1995, 2006) highlights the tension between institutional frameworks and subjective meaning-making, which is sharpened when infrastructures filter visibility or automate evaluation. From this perspective, Bildung under digital conditions emerges as refracted through constellations of authority, opacity, affect, and recognition. It is neither eclipsed by technology nor simply affirmed but transformed into a precarious process in which fragility becomes a necessary condition of educational experience.

5.2. Democratic Education Under Algorithmic Mediation

The reconstructed orientations closely intersect with key debates in democratic education. Dewey (2000) understood democracy as a way of life sustained through shared communication and collective inquiry, in which problem-solving itself becomes a civic practice. Young (2000) extends this view by arguing that democratic participation attains its meaning only when diverse voices are equally heard and recognized. Algorithmic mediation modifies precisely these conditions by filtering contributions, amplifying some while silencing others. Rancière (1999) underscores that democracy requires the disruption of established regimes of visibility and audibility, while Honneth (1996) stresses that recognition is a prerequisite for participation. The empirical findings demonstrate how these dynamics play out in HE, as students and lecturers experience shifting patterns of authority, opacity, and recognition.

Biesta (2006, 2013) reminds us that education always involves subjectification, understood as the experience of being called into responsibility by the encounter with the unfamiliar. The answers show that students’ hesitation, pragmatic reliance, or withdrawal under conditions of opacity dramatize precisely this fragile process. From a democracy-pedagogical perspective, HE cannot be confined to knowledge transmission but must sustain spaces where students are enabled to test their voices and engage with plurality. Print and Lange (2012) emphasize that civic education requires competences for active participation, anchored in curricular and pedagogical strategies. While their work focuses on schools, the HE context requires similar strategies, adapted to study entry, disciplinary socialization, and the longer biographical horizons of students. Universities thus function not only as academic institutions but also as arenas of civic formation, where algorithmic infrastructures must be critically examined so that deliberation, participation, and recognition remain possible.

5.3. Linking Micro-Level Orientations with Systemic Mechanisms

The empirical categories capture not only immediate practices but also illuminate systemic dynamics identified as risks of generative AI. From an education-theoretical perspective, they show how macro-level logics of algorithmic mediation are refracted into the pedagogical lifeworld of HE. Pragmatic reliance and epistemic blind spots exemplify the tension between convenience and opacity. This pattern reflects Biesta’s (2013) view of subjectification as an encounter with the unpredictable and accords with Beck’s (1992) notion of the risk society, where uncertainty and opacity have become normalized as structural conditions that guide everyday decision-making, including in educational environments shaped by algorithmic mediation. Delegated responsibility and institutional ambiguities point to questions of accountability and governance. Foucault (2006) clarifies how power is exercised through infrastructures that diffuse authority and compel self-regulation. Gutmann (1999) add that democratic education requires reciprocity and justification, both of which are undermined when accountability is displaced onto opaque systems. von Felden (2006) shows that such ambiguities are particularly critical in transitional phases of educational biographies, when students need reliable frameworks for orientation. Affective frictions and the micropolitics of recognition illustrate how emotions, legitimacy, and visibility intertwine. Honneth’s (1996) theory of recognition, Fraser’s (2003) distinction between redistribution and recognition, and Fricker’s (2007) notion of epistemic injustice converge here. They show that participation in HE depends not only on access but also on the symbolic orders through which voices are legitimized. Giddens (1995) helps explain how micro-level practices of reliance, withdrawal, or adaptation reproduce and reshape systemic structures. For example, when students normalize opacity as a fact of digital life, they not only cope individually but also reinforce systemic invisibility of algorithmic operations.

Taken together, the categories reveal that HE is a hinge institution where systemic dynamics are refracted into everyday practices. The orientations reconstructed from the written survey articulations cannot be reduced to private strategies but must be understood as socially structured and biographically sedimented. They illustrate how democratic risks associated with generative AI are lived, interpreted, and sometimes contested in academic practice.

5.4. Theoretical Consequences for Bildung and Democracy

The integrated analysis suggests three consequences for conceptualizing democratic education under algorithmic conditions.

Bildung as fragile orientation. Bildung unfolds not as stability or mastery but as the cultivation of orientation under uncertainty, opacity, and contested recognition. Marotzki’s (1990) structural perspective highlights the transformative potential of such fragility, while Alheit’s (2008) account of Biographizität clarifies how orientations are carried across educational trajectories. Fragility is not a deficit but a condition in which reflexivity becomes possible. Pedagogically, this means universities should not aim to eliminate fragility but design settings where moments of disorientation can be productively taken up and reflected.

Democracy as contested practice. Algorithmic infrastructures reveal that democracy in education cannot be conceived as harmonious deliberation. They reconfigure visibility, redistribute legitimacy, and accentuate contestation. Rancière (1999) stresses that democracy depends on disruption, while Print and Lange (2012) emphasize that civic education must provide competences for engaged participation. Himmelmann (2007) reminds us that democracy operates simultaneously as a form of government, a societal structure, and a way of life requiring pedagogical spaces where these dimensions intersect. HE thus bears responsibility for supporting students in practicing participation, especially in fragile transitions such as study entry and professional orientation.

HE didactics as critical mediation. Pedagogical responsibility is not displaced by algorithmic infrastructures but transformed. The task is to render these operations visible, to create spaces where students and lecturers critically interrogate shifting relations of authority, recognition, and orientation, and to support democratic dispositions across diverse study trajectories. Giddens’ (1995) structuration theory shows that such practices both reproduce and reshape institutional orders, making HE a decisive arena of societal democratization. Concretely, formats such as dialogical seminars, reflective writing, and peer feedback can counterbalance algorithmic authority by fostering justification, reciprocity, and recognition. In this constellation, universities emerge as spaces where macro-level risks of algorithmic governance are transformed into pedagogical occasions. They are exposed to systemic pressures yet simultaneously cultivate reflexive and democratic capacities to confront them. This concerns not only curriculum and teaching but also study entry, progression, and the negotiation of diversity. By embedding democratic education in pedagogical practices attentive to biographies, inequalities, and civic competences, HE contributes to democratization and societal participation.

6. Implications for HE Didactics Under Algorithmic Conditions

The empirical analysis has reconstructed five orientations through which students and lecturers navigate generative AI in HE. These orientations reveal fragilities in judgment, recognition, and responsibility that are not signs of deficit but manifestations of how Bildung unfolds under algorithmic mediation. For HE didactics, a central task is therefore not merely the transmission of competencies but the capacity to transform such fragilities into productive pedagogical occasions. What is ultimately at stake is the university’s ability to render algorithmic pressures educable and to sustain democratic forms of subject-formation when infrastructures of visibility, evaluation, and orientation are technologically mediated.

As recent research emphasizes, this requires moving beyond generic appeals to “digital maturity” or “AI literacy”. Watanabe (2023) shows that opacity and the conditioning effects of algorithmic operations are not peripheral but central pedagogical challenges, since they determine what can be perceived, questioned, and contested in educational contexts. Didactic responses should therefore embed structured opportunities for reflection, dialogical reasoning, and recognition practices that turn algorithmic ambivalences into shared pedagogical resources (Hummel et al., 2024). At the same time, two further dimensions are decisive: policy frameworks at the institutional level and teacher education at the professional level. Universities need guidelines that not only regulate the use of AI but explicitly conceptualize algorithmic infrastructures as educable phenomena. Likewise, didactic training must integrate reflexive approaches to AI use, enabling teachers to work productively with ambivalence rather than reduce it to technical skills. To strengthen transparency of the analytical process, the table includes representative quotes used during category development, illustrating how empirical articulations informed the reconstructed orientations.

Table 1. Framework of Orientations, Biographical Embedding, Tensions, and Pedagogical-Democratic Implications.

Orientation	Illustrative Quote	Biographical Embedding	Pedagogical/Epistemic Tensions	Democratic-Educational Implications	Didactic Strategies
Pragmatic Orientation: Coping with Workload	“Even before AI, I often skipped readings or just skimmed them, now I put the text into ChatGPT and get a summary—it feels like cheating time” (S4).	Continuities of coping strategies across study phases; sedimentation of economizing practices.	Efficiency vs. reflexivity; delegation vs. autonomy.	Risk of narrowing learning to instrumental rationality; unequal distribution of coping resources.	Assignments where summaries must be critically compared with original texts (S4); guided self-checks to interrupt reliance (S2); dialogical spaces to discuss time-economization (L4).
Adaptive Orientation: Navigating Opacity	“I honestly do not want to know what happens in the background of ChatGPT. It is too complicated, and as long as it works for me, that is enough” (S7).	Normalization of not-knowing as protective orientation in fragile study phases.	Transparency vs. opacity; engagement vs. withdrawal.	Unequal participation when opacity is normalized; potential reinforcement of exclusion.	Collective inquiries into “what the system does” (S7, S3); reflective writing tasks on invisible influences; peer discussions where disengagement (L2) is addressed as shared uncertainty.
Relational Orientation: Authority and Resonance	“Students sometimes trust the feedback they receive from the system more than mine. They even cite it as unquestionable” (L4).	Reliance on external authority sedimented across trajectories; recognition shaping trust in AI vs. human feedback.	Dialogical justification vs. automated validation; formative vs. procedural evaluation.	Redistribution of evaluative credibility into infrastructures; weakening of dialogical legitimacy.	Hybrid feedback formats combining algorithmic scoring with dialogical discussion (L4, L2); peer-feedback tasks to rebalance credibility (S8); reflective assignments on trust in feedback sources.
Ambiguous Orientation: Improvisation and Fragility	“In some seminars we are told not to use AI at all, in others it is almost expected. You never know what is actually allowed, and every decision feels risky” (S9).	Fragile belonging in transitional phases; improvisation as response to contradictory institutional rules.	Reliable orientation vs. contradictory expectations; institutional guidance vs. individual improvisation.	Unequal participation when navigating contradictions depends on biographical and cultural resources.	Collective clarification of rules (S9’s irritation); course-specific agreements to reduce contradictory signals (L1); reflective spaces where students like S11 can voice belonging insecurities.
Recognition-Oriented Orientation: Voice and Visibility	“On the platform, some posts get highlighted automatically and receive many comments, while others disappear without notice. It changes what the group actually talks about” (S5).	Biographical sedimentation of recognition/misrecognition; visibility shaping participation.	Inclusion vs. exclusion; legitimacy vs. invisibility.	Unequal distribution of voice and visibility; reinforcement of inequalities through infrastructures.	Design of forums where hidden contributions (S5) are surfaced; exercises amplifying marginalized voices (student solidarity practices); reflective accounts of (in)visibility in assessment.

synthesizes the reconstructed orientations with their biographical embedding, pedagogical tensions, democratic-educational implications, and didactic strategies. It should not be read as an exhaustive set of categories or a normative model. Rather, it serves as a heuristic that connects empirical findings with HE didactics, highlighting how fragile orientations can become occasions for pedagogical work. The strategies are not recipes but proposals that make visible how algorithmic mediation can be engaged in teaching while remaining attentive to biographical trajectories and democratic conditions of recognition.

The contribution of this framework lies in its integration of educational theory, biographical research, and democracy-pedagogical perspectives into HE didactics. In contrast to approaches that emphasize AI literacy or digital citizenship as fixed skill sets, the focus here is on cultivating fragile orientations as educable conditions. Temporal sovereignty, recognition, and accountability cannot be conveyed as discrete competencies but must evolve within dialogical, curricular, and institutional practices. This perspective builds on Dewey’s (2000) notion of democracy as a lived practice, Rancière’s (1999) understanding of disruption as the enactment of equality, and Honneth’s (1996) view of recognition as the foundation of participation. It also refines Himmelmann’s (2007) dimensions of democracy by demonstrating how algorithmic infrastructures transform each of them—shaping autonomy, reorienting institutions, and redistributing the legitimacy of voices.

Policy frameworks and teacher education emerge as crucial mediating layers. Universities should establish guidelines that conceptualize algorithmic infrastructures as pedagogical phenomena, making opacity and fragility part of structured educational reflection rather than leaving them as hidden influences. At the same time, teacher education programs should integrate biographical and reflexive dimensions of AI engagement, equipping lecturers to transform algorithmic ambivalences into occasions for judgment, recognition, and democratic participation. By situating the five orientations within multi-level didactic strategies and embedding them in biographical perspectives, the framework shows how HE becomes a crucial arena for responding to the pressures of algorithmic mediation. Universities are simultaneously confronted with systemic risks of algorithmic governance and equipped to cultivate the reflexive and democratic capacities needed to address them. Acceleration can be reframed as an occasion for structured reflection, opacity as a shared field of inquiry, and asymmetries of recognition as a starting point for pluralization. The didactic task is to treat fragility as a productive condition, to translate algorithmic ambivalences into educable situations, and to maintain HE as a setting in which democratic subject-formation can be sustained. These didactic implications derive directly from the five reconstructed orientations and the biographical, pedagogical, and democratic tensions identified in the empirical analysis. They translate students’ and lecturers’ articulated patterns of coping, navigating opacity, renegotiating authority, improvising under ambiguity, and struggling for recognition into concrete strategies for fostering reflexivity and democratic participation under algorithmic mediation.

7. Conclusions and Outlook

The study has shown that generative AI is not simply an auxiliary tool in HE but a structuring force that influences how learning, judgment, and recognition take form. The five reconstructed orientations (pragmatic orientation, adaptive orientation, relational orientation, ambiguous orientation, and recognition orientation) illustrate how systemic pressures of acceleration, opacity, and infrastructural authority are refracted into everyday pedagogical practice. These orientations are not to be read as deficits but as fragile processes of Bildung, in which students and lecturers continuously negotiate autonomy, responsibility, and recognition under digital conditions.

Theoretically, the findings contribute to education-theoretical and democracy-pedagogical debates. They suggest that the capacity for responsible self-determination (Klafki, 1996) and subjectification (Biesta, 2006, 2013) acquires renewed importance when orientations are shaped by algorithmic infrastructures. Dewey’s conception of democracy as a way of life underlines that pedagogical spaces are vital for cultivating dispositions of trust, courage, and critique. Rancière (1999) emphasizes that democracy also depends on disruption and contestation, while Honneth (1996) and Fraser (2003) highlight that recognition is a precondition for participation. Himmelmann’s (2007) account of democracy as government, society, and way of life clarifies why universities play a decisive role: they are not only institutions of knowledge transmission but also arenas where democratic capacities are tested, challenged, and renewed. Print and Lange (2012) further point out that democratic education requires concrete pedagogical arrangements that enable students to practice participation in plural and sometimes contested contexts. Empirically, the analysis indicates that fragilities surface most clearly in transitional phases such as study entry, progression through programs, and preparation for professional life. These moments are marked by heightened uncertainty, and it is here that algorithmic mediation intersects with biographical orientation, recognition needs, and institutional frameworks. Rather than attempting to eliminate fragility, universities can work toward transforming it into an educable condition. Didactic strategies such as reflective assignments, dialogical inquiry into algorithmic feedback, and diversification of recognition practices illustrate how pedagogy can sustain plurality and reflexivity. Teacher education and professional development for lecturers are essential in this respect, as they can help establish the capacity to work productively with algorithmic ambivalences rather than avoid them. In this light, HE appears both vulnerable and formative. It is vulnerable because systemic risks of algorithmic governance—including acceleration, opacity, and asymmetries of recognition—permeate teaching and learning. Yet it is also formative because these very risks can be reframed pedagogically into occasions for judgment, reflexivity, and democratic participation. This duality positions universities at the center of societal democratization, where global technological pressures are translated into lived educational processes and democratic subject-formation remains possible.

Looking ahead, further research could broaden the comparative scope across disciplines and national contexts, examining how orientations vary with epistemic cultures and institutional traditions. Methodologically, the combination of reconstructive, biographical, and interpretive approaches offers a promising pathway for connecting empirical findings with educational theory and democracy pedagogy. On the policy level, universities would benefit from developing guidelines that not only regulate AI use but also render algorithmic infrastructures educable, turning opacity into a shared object of inquiry. The central didactic task is to design pedagogical arrangements that keep fragility educable, recognition plural, and democratic participation open. In doing so, universities can uphold their role as hinge institutions where Bildung, democracy, and digital transformation intersect in ways that remain reflexive, inclusive, and oriented toward the future of democratic education.