Research Article
Generative artificial intelligence use among lecturers in a resource-constrained higher education institution in Somaliland
Synthesis: A qualitative single case study of 14 lecturers at a resource-constrained medical university in Borama, Somaliland, interviewed about why they use generative AI, what it costs them, and what they understand about its ethics, with TPACK as an interpretive lens. All 14 taught in one of the university's seven health-professions programs and had used GenAI directly; interviews averaged 30 minutes, eleven on campus and three over WhatsApp. ChatGPT and DeepSeek dominated, and the reported motivations were efficiency, ease of use, simplified explanations and — distinctively — access to material in Somali. Uses spread across teaching, assessment, research and capacity building. Constraints were economic and institutional rather than pedagogical: a $20 monthly subscription the university did not cover, misleading generated content, uneven ethical knowledge (seven of the 14 said they did not know GenAI ethical policies, four said they did, three partly), and no institutional AI policy. The study extends TPACK's contextual application rather than modifying the framework, treats ethical awareness as a contextual dimension rather than a knowledge domain, and — being a single case of 14 self-reporting interviews in one discipline — establishes no causal relationship and no student outcome.
Key Findings
- A small, saturated sample across seven programs. 14 lecturers were interviewed — two from each of the university's seven undergraduate programs — until no substantially new themes emerged.
- Two tools carry most of the use. ChatGPT and DeepSeek were the most commonly reported, with lecturers typically using several tools rather than one.
- Motivation is efficiency plus language access. Time saving, ease of use, simplifying complex concepts and Somali-language translation were the four reasons lecturers gave for their daily use.
- Use spans four functions. Teaching preparation, assessment design, research support and capacity building, the last including English language development for lecturers teaching in a second language.
- A $20 monthly subscription is the access boundary. The university did not cover it, and lecturers on free versions reported more errors and restricted functionality.
- Ethical knowledge is partial and policy is absent. Seven of 14 lecturers did not know GenAI ethical policies, four did, three partly, and no institutional AI policy existed.
- TPACK is extended, not revised. Institutional support, ethical awareness and language accessibility are treated as contextual factors shaping how the framework is enacted in practice.
Motivation is efficiency and language access
Lecturers described a pragmatic, workload-driven relationship with the tools. The dominant motive was time and effort saved on assignments, research and teaching materials, and the interviews return to three affordances: it is easy, it is fast, and it explains difficult content simply. What distinguishes this case is a fourth motive the study treats as central rather than incidental: the tools produce and translate material into Somali, the lecturers' mother tongue, in a system where academic resources are predominantly in English. GenAI functions there as a linguistic bridge as much as a productivity tool, which makes the finding relevant to multilingual learning and to comprehension rather than speed alone.
The pattern of use is correspondingly broad and unspecialized: one accessible tool is made to serve lecture preparation, examination writing, literature searching, paraphrasing, coding and language development. In TPACK terms the author reads this as technological knowledge in use, shading into pedagogical and content knowledge where the tools support planning and assessment, while noting that the number of tools a lecturer used varied widely — some reported five or six, others one.
The subscription is the access boundary
Cost is the study's sharpest finding, and it is sharper than a generic infrastructure complaint. Most GenAI tools require a paid subscription; one lecturer named the figure directly — $20 monthly, described as a challenge in the Somaliland context — and the university did not reimburse it. Lecturers who could not pay used free versions, and several reported that free or open-access versions generate more errors and expose fewer capabilities.
The author's framing is the useful part: the problem is not access to GenAI but access to its more advanced capabilities, producing a divide in tool quality rather than in tool availability. That refines the standard digital divide account, which typically separates having access from using it well, by adding a tier between them — technically using AI while remaining unable to benefit from the paid features. For a Global South institution this is an equity question with a concrete price attached, and it is one an institution can answer with a budget line in a way that infrastructure cannot.
Ethics by personal judgment, without a policy
Awareness of GenAI ethics was uneven and largely informal. Seven of the 14 lecturers said they did not know GenAI ethical policies, four said they did, and three claimed partial knowledge. The more interesting observation is what sat underneath that distribution: five of the seven who admitted no policy knowledge nonetheless described practices they considered ethical — acknowledging AI use, verifying output, withholding confidential information — while at least one lecturer reported not disclosing GenAI use in research at all. The author reads this as ethical awareness concentrated on output reliability and stopping short of authorship, transparency and academic integrity, and warns that self-reported awareness does not translate into ethical practice.
The institutional explanation is straightforward: no national AI strategy, no institutional AI policy, no formal professional development on responsible use. Lecturers fell back on general academic norms and personal judgment, which the study links to the governance gap rather than to individual deficiency — and which is why it recommends AI ethics modules, AI literacy training and explicit privacy guidance as the first institutional moves.
What this means for practice
- Instructors. Verify before you use, and disclose when you do. Lecturers in this case reported fabricated references, invented data and repeating questions from the tools, and one described using GenAI for research content without acknowledging it — the two practices the study treats as the floor of professional responsibility.
- Faculty developers. Build ethics into staff training rather than assuming it arrives with tool familiarity: seven of 14 lecturers did not know GenAI ethical policies even while using the tools daily for assessment and research.
- Administrators. Price the subscription. A $20 monthly fee that the institution does not cover pushed lecturers onto free tiers that they reported as more error-prone, so a modest budget line changes the quality of tool in use, not merely the number of users.
- Researchers. Treat single-case, single-discipline, self-reported ethical knowledge as a hypothesis about practice, not evidence of it — the study's own design supports contextual interpretation only.
Limitations
- One university in Borama, Somaliland, drawn from a single institution chosen as an information-rich case; the author states the findings are not statistically generalizable.
- All 14 interviewees taught in health professions programs — the university's seven programs are Medicine and Surgery, Optometry, Dentistry, Medical Laboratory, Nursing and Midwifery, Public Health, and Human Nutrition — so the disciplinary homogeneity may have shaped both tool use and ethical reasoning.
- Twelve of the 14 participants were male and two female, reflecting the institution's 33 lecturers (26 male, 7 female); the author notes the range of perspectives captured may not represent female lecturers.
- Interviews averaged approximately 30 minutes, with three conducted over WhatsApp rather than in person, and analysis was manual and inductive by a single researcher.
- The TPACK framing is interpretive: the author states explicitly that associations are contextual, that the framework was not modified, and that the design cannot establish causal relationships.
Citation
Negash, K. H. (2026). Generative artificial intelligence use among lecturers in a resource-constrained higher education institution in Somaliland. Discover Education.