Beyond the algorithm: academic developers as digital mediators in Global South higher education

Synthesis: Sithole interviews twelve academic developers and learning designers — six each at two South African Historically Disadvantaged Institutions — to ask what AI integration looks like when it arrives at universities still shaped by apartheid-era underfunding. Read through Critical Digital Pedagogy and decolonial theory, their accounts describe work that is not technical implementation but mediation across three registers: pedagogical, ethical and institutional. Institutional AI rhetoric was "performance more than practice," policy loosely coupled to infrastructure, while developers filtered the hype through questions about students who struggle with Wi-Fi, absorbed the affective cost of projecting expertise they were still acquiring, and built peer spaces to test and deliberately break tools rather than master them. The paper's sharpest analytical move is to separate digital inequality, a distributive problem answerable through access, from algorithmic coloniality, an epistemic problem that survives full access because it inheres in what the systems encode. The claim that follows is that AI adoption in Global South institutions is a negotiation over pedagogy, institutional purpose and epistemic justice, not a technology rollout.

Key Findings

1. AI discourse ran ahead of the material conditions it assumed. Across both institutions, strategy documents and management communications framed AI as central to innovation and global relevance, but participants described the rhetoric as structurally disconnected from everyday teaching. "There's a lot of excitement from management, emails, slogans about digital transformation, but no infrastructure or policy to support it. It becomes performance more than practice" (P2). One developer's standing question was "what does this mean for teaching here, for our students from poor schools?" (P9) — read by the author as an epistemic intervention that privileges situated knowledge over universalising technological claims.
2. Professional identity shifted toward being a bridge, including as a brake. Participants moved from facilitating reflective teaching toward acting as technological intermediaries and problem-solvers on demand. One described translating between "what management wants, what the technology can do, and what lecturers are actually worried about," and named the deliberate slowing-down — raising plagiarism, student dependency, outsourcing of thinking — as part of the job: "It is not just technical support; it is also negotiation" (P3).
3. Affective labour and professional vulnerability are constitutive, not incidental. Developers were expected to embody institutional confidence in tools whose implications they were still working out. "I feel like an impostor. I am learning AI as I go, but the institution expects expertise" (P7). A second dimension is ethical: "You want to support lecturers, but sometimes you are not convinced the tool is appropriate. That creates discomfort; you are facilitating something you are still questioning" (P4). The paper names this an under-theorised dimension that affect theory might address.
4. Judgement, not skill, was the goal of professional learning. Peer spaces were used to test tools together and to try to "break" them — asking for generated feedback and then reasoning about whether it would help or mislead students. The participant's own framing: "It becomes less about skill and more about judgement" (P10). Concrete questions followed: what happens to writing development if students draft with AI, do students still feel seen when lecturers use AI for Feedback, and who is accountable when AI makes errors.
5. Reflexivity and context acted as the criteria for adoption. Academic developers voiced concerns about student dependence, Academic Integrity and homogenised thinking, and created reflective spaces rather than prohibitions. "I am not anti-AI; I am pro-context. If we automate feedback without understanding students' struggles, we lose the human connection teaching depends on" (P1) — a position the author reads as treating context as an epistemic principle rather than a design preference.
6. Infrastructural scarcity produced both constraint and low-resource invention. Developers worked around limited budgets, outdated devices and low digital literacy ("some lecturers still struggle to upload materials to the LMS. AI feels far away from that reality," P6), while still generating case examples with Gemini and ChatGPT for under-resourced modules and reviewing them critically with lecturers. The summing metaphor is P11's: "We are asked to build a digital future on analogue foundations."

Digital inequality and algorithmic coloniality are distinct problems

The paper's main theoretical contribution is analytical rather than empirical: it insists these two frequently conflated phenomena operate at different registers and demand different responses. Digital inequality is distributive and concerns access to the material conditions of AI use — devices, connectivity, budgets, digital literacy. Algorithmic coloniality, Mohamed et al.'s term, is epistemic and concerns the knowledges systems privilege and erase; it persists even under conditions of full access because it is built into what the systems encode.

P11's "analogue foundations" image is the hinge between them. The analogue foundations name the material register; the digital future built on them is imported, "arriving pre-loaded with the epistemic assumptions of the contexts that designed it." The Historically Disadvantaged Institution is therefore doubly positioned: excluded from the technology by scarcity, and, when included, subordinated by systems that encode other knowledges and norms. Recognising that doubling is what the author calls the digital mediator's epistemic labour, and it aligns the study with epistemologies of the South arguments that legitimate knowledge emerges from situated contexts rather than dominant Northern epistemologies. The paper also adopts Adams's warning that calls to decolonise AI can be misappropriated as metaphor unless they confront race and the historically embedded power relations inside the systems themselves.

Digital mediators as sociotechnical praxis

Mediation is specified in three dimensions. Pedagogical mediation translates AI from institutional discourse into concrete teaching, learning and Assessment practice, keeping adoption subordinate to pedagogical purpose. Ethical mediation treats technology as value-laden and negotiates automation against academic integrity, innovation against care, efficiency against relationality. Institutional mediation works the liminal space between managerialist imperatives and academic concerns, carrying global digital-transformation discourse into local infrastructural realities. Together the author frames academic development not as support service but as sociotechnical praxis: situated, justice-oriented engagement in which technological change is continually interpreted through context, equity and reflexive judgement. The framework extends earlier metaphors for the role — critical friend, broker of change, migrant worker, change architect — into AI-mediated environments.

Method and limits

Twelve purposively sampled participants (nine academic developers, three learning designers; six per institution) were interviewed for 45–60 minutes over eight weeks, face to face or on Microsoft Teams, with verbatim transcripts analysed through Braun and Clarke's reflexive six-phase thematic analysis in Atlas.ti, combining inductive coding with sensitising concepts from Critical Digital Pedagogy and decolonial theory. The two sites differed in orientation: Institution-A had begun folding AI into teaching and learning strategy with assessment-redesign and digital-literacy pilots, while Institution-B's engagement was fragmented and driven by individuals. Limitations are acknowledged directly: a single method with no documentary or observational data, twelve participants in a small and functionally specific professional group, and findings from two historically disadvantaged South African institutions that cannot be generalised to other contexts. Institutional and personal identification is masked, so the paper cannot be read as a comparative account of the two sites despite the strategic difference between them.

Connected Concepts

• Global South — the study's epistemic frame: situated knowledge, epistemic justice and imported technological prescriptions
• Equity — the equity commitments developers held against institutional AI enthusiasm
• Digital Divide — the distributive register: connectivity, devices, budgets and digital literacy
• Change Management — the mediation of top-down digital transformation narratives into grounded practice
• Educational AI Policy — institutional AI rhetoric loosely coupled to policy and infrastructure
• AI Governance — liminal institutional positioning between management imperatives and academic concerns
• Critical Pedagogy — Critical Digital Pedagogy as the study's theoretical anchor for interrogating technology
• Educational Development — the field itself: academic development as sociotechnical praxis rather than technical implementation
• Teaching — the shift from facilitators of reflective teaching to technological intermediaries
• Teacher AI Competency — expertise demanded ahead of expertise held, and judgement as the target of professional learning
• AI Literacy — developers' own emergent, collaborative and sceptical learning about AI
• Anxiety and Stress — impostor feelings, public failure and ethical discomfort in mediation work
• Academic Integrity — the plagiarism, dependency and cognitive-outsourcing concerns developers raised with lecturers
• Workplace Learning — peer-learning spaces as the vehicle for practitioner development
• Generative AI — the technology participants' accounts centred on
• Higher Education — South African Historically Disadvantaged Institutions as the setting

Connected Articles

• Institutional Structures, Digital Inequality, and AI Integration in Higher Education — Institutional capacity and structural inequality as predictors of African AI integration
• Perceptions Of Generative AI in the Global South: A Scoping Review — What the Global South GenAI evidence base does and does not cover
• From fear to innovation: A case study of transformative faculty development for ethical AI integration in higher education — Faculty development for ethical AI integration at a US institution
• When faculty ask, 'what's the point of teaching?': GenAI as identity crisis, not skills gap — Faculty identity disruption as generative AI enters teaching
• Situated AI ethics: a cultural-historical and ecological framework for education — Situated ethics for AI in higher education
• Reorienting Critical AI Literacy: A Community-Rooted Praxis of “Resisting AI” — Critical AI literacy as a reorientation of practice
• The University AI Didn''t Replace: Rethinking Universities in the AI Era — Rethinking the university's purpose under AI-driven transformation
• Anticipatory governance and leadership for AI implementation in higher education: A scoping review — Governance responses to AI in higher education institutions

Citation

Sithole, M. P. (2026). Beyond the algorithm: academic developers as digital mediators in Global South higher education. International Journal for Academic Development, advance online publication.