Digital borders, algorithmic governance and civic digital literacy: examining the educational implications of AI-driven immigration management in post-apartheid South Africa

Synthesis: South Africa is assembling an AI-mediated border regime — biometric screening, drone surveillance, electronic travel authorisation — and this qualitative thematic synthesis of 53 sources asks what that regime demands of citizens as learners. Its answer is a civic form of AI literacy: the capacity to interrogate algorithmic risk scores, know one's data rights, recognize reproduced racial bias, contest automated decisions, and join the AI Governance of the systems themselves. Working through critical digital pedagogy, algorithmic accountability theory and postcolonial education studies, the author argues that frameworks such as DigComp 2.1 and South Africa's CAPS teach functional skills and leave algorithmic accountability untouched, and that the communities most exposed to automated border decisions are also the least connected — a Global South equity problem rather than a skills deficit. Geography and Social Sciences education, with teacher education behind it, is proposed as the vehicle. This is a synthesis of documents, not evidence of classroom effects.

Key Findings

1. The corpus comprised 53 sources — peer-reviewed literature, policy and governance documents, and gray literature — screened through an adapted PRISMA-style procedure and analyzed thematically.
2. The synthesis produced five descriptive and three analytical themes spanning the border architecture, its literacy demands, framework inadequacy, and educational responses.
3. The architecture has four integrated systems: biometric screening at ports of entry, drone surveillance of four priority border segments, Electronic Travel Authorisation risk screening, and One-Stop Border Posts at six land ports including Beitbridge.
4. Parliamentary replies report that between 1 April 2025 and 31 March 2026 the Border Management Authority deported 37,356 people and prevented 10,512 inadmissible or prohibited foreigners from entering at ports.
5. The analysis identifies five civic digital literacy demands: algorithmic transparency, data rights, bias recognition, rights contestation, and participatory governance, each tied to a specific legal or technical gap.
6. Frameworks fall short — DigComp 2.1 omits algorithmic accountability, CAPS omits algorithmic governance, and nearly 16,000 South African schools lack dedicated computer laboratories.

Summary

The paper treats border control as an educational problem. South Africa's Border Management Authority, established under the Border Management Authority Act 2 of 2020, is deploying biometric verification at ports of entry, aerial surveillance over priority border segments, and an Electronic Travel Authorisation system that the Department of Home Affairs calls the single entry point for all immigration processes. Each shift moves discretion from an official who can be questioned to a risk score that cannot. The author asks who is equipped to notice, interpret, and contest that shift, and answers: almost no one, because the communities whose mobility is most directly regulated are the least likely to hold the vocabulary needed to challenge an automated decision. The contribution is to name that gap as civic rather than technical, and to argue that critical pedagogy is the appropriate response.

The Border Architecture and Its Literacy Demands

The four-system account is the paper's empirical core. Biometric enrollment and verification against the Automated Biometric Identification System sit in the first phase of the ETA rollout; drone surveillance targets four priority vulnerable segments; the ETA applies a risk formula before a traveller reaches an officer; and the One-Stop Border Post model, piloted at six commercial land ports including Beitbridge, ties physical infrastructure to algorithmic processing. From that architecture the coding derived five demands: knowing that consequential decisions are algorithmic (transparency), knowing what data are collected and what rights attach (data rights, including section 33 limits), recognising reproduced bias, knowing appeal routes under the Promotion of Administrative Justice Act 3 of 2000 and section 29 of the BMA Act, and participating in oversight.

Why Existing Frameworks Fall Short

The mismatch argument is precise about what it criticises. Gilster and Eshet-Alkalai defined digital literacy as finding, evaluating, and communicating digital information; DigComp 2.1 adds data literacy, digital safety, and problem solving but leaves algorithmic accountability outside its competence areas. South Africa's Curriculum and Assessment Policy Statements for Information Technology, Computer Applications Technology, and Social Sciences carry little digital literacy content, and the geography syllabus contains nothing on AI or algorithmic systems. The White Paper on e-Education and the draft Coding and Robotics curriculum emphasize access, ICT capability, and coding, with algorithm used in its programming rather than governance sense. The result is a curriculum that can produce competent users of digital tools who cannot interrogate a decision made about them.

Geography and Social Sciences as the Vehicle

The proposal is disciplinary rather than generic. Geography already teaches power, governance, and inequality, and its treatment of borders, migration, and surveillance offers case sites — Beitbridge, the Nkomazi interface with Kruger National Park, the Limpopo River crossing — that can be read as algorithmically regulated spaces with histories of racialised movement control. Social Sciences carries citizenship, human rights, and democratic participation, which lets algorithmic literacy be framed as a democratic competence rather than a technical one. Life Orientation supplies the human rights content where digital rights can be anchored to constitutional rights. The weak link is teacher education: the author reports that civic AI literacy is absent from pre-service preparation, including at universities serving the communities most affected by border governance, and recommends phased integration through existing methodology modules using text-based, low-bandwidth activities.

What this means for practice

• Instructors. Use border, migration, and surveillance cases already in the Geography and Social Sciences syllabus to teach how algorithmic decisions are made, rather than adding a separate AI unit.
• Faculty developers. Prepare pre-service teachers in critical perspectives on algorithmic governance through existing methodology modules, since the paper reports it is absent from initial preparation.
• Administrators. Plan low-bandwidth delivery — document analysis of ETA and BMA texts, mapping of border technologies — given that nearly 16,000 schools lack computer laboratories.
• Researchers. Test the five proposed demands against what border communities and teachers actually describe, since the paper derives them from documents rather than from affected people.

Limitations

• The design is documentary: there is no primary data from border communities, immigration officials, educators, or students, which the author names as the main methodological limitation.
• Parliamentary replies do not state the algorithmic methods or criteria used, so classifying the ETA and surveillance systems as AI-enabled is the author's judgment rather than a reported finding.
• Coding was interpretive, done by a single researcher with no independent second coder, and the PRISMA-style diagram reports no counts at intermediate stages.
• The ETA was still being rolled out at the time of writing, so its design and algorithmic logic may differ from what public documents describe.

Citation

Nkalanga, S. D. (2026). Digital borders, algorithmic governance and civic digital literacy: examining the educational implications of AI-driven immigration management in post-apartheid South Africa. Frontiers in Education.