AfricArXiv preprint · protocol · Ghana · 2026
Evaluating AI's Role in Public Technology Literacy in Ghana
A mixed-methods research protocol and conceptual framework
- Researcher
- Fahad Mohammed Gibrine
- Status
- Ongoing research · Version 1.0 preprint (not peer reviewed)
- Affiliation
- Technology & Cybersecurity Consultant (Contract), ByteLine Technologies, Ghana; Independent Researcher
Ask most Ghanaians whether they use technology and the answer is almost certainly yes. Mobile money, WhatsApp, YouTube tutorials, chatbots on customer-service lines — from Accra’s markets to villages in the Upper West, the tools are already there. Ask a harder question and it gets murkier: can they tell when a message is an AI-generated scam, understand why an app wants their Ghana Card number, or check a claim before they share it?
That gap — between using technology and actually understanding it — is what this project is trying to measure. The question is simple to ask and hard to answer:
Does AI help Ghanaians become more capable, critical, safe, and independent technology users, or does it mainly make digital services easier to consume without improving understanding?
Why this, why now
Internet use in Ghana jumped from roughly 8% of the population in 2010 to about 69% in 2021, but it is still uneven: around 80% in urban areas versus 54% in rural ones (World Bank, 2023). For most people the phone is the internet. That is a good setting for chatbots and local-language voice tools. It is a weaker setting for the slower skills — privacy, document work, checking what a model just told you.
Ghana launched its first National AI Strategy on 24 April 2026. The Data Protection Act (2012) and the Right to Information Act (2019) already exist, but UNESCO still flags the absence of AI-specific law, procurement rules, and a working path for algorithmic transparency and redress (UNESCO, 2026). If people cannot tell they are talking to an automated system, or how to complain about a bad decision, those protections stay on paper.
The security side is not theoretical either. In September 2026 the Ghana Cybersecurity Industry Forum warned that criminals are already using AI to write cleaner phishing. Training that only teaches “open the chatbot” will not keep up.
How I am measuring literacy
I am not treating “AI literacy” as one vague score. The working paper splits it into five things you can actually look at:
- 1. Access. Can people actually reach the tools? Device, data cost, electricity, local-language interfaces, disability access, and who gets left out in rural areas.
- 2. Functional skills. Can they finish ordinary tasks: email, e-government forms, safer mobile money, or using AI to summarise something and then checking that summary?
- 3. Critical AI literacy. Do they treat output as automatically true, private, or fair? Can they ask who built the tool, whether it can be wrong, and when a human should decide instead?
- 4. Safety and ethics. Passwords, phishing, what never to type into a public model (Ghana Card, OTPs, health details), spotting synthetic media, and consent before uploading someone else’s voice or photo.
- 5. Empowerment. Not a quiz score. Whether people get more independent: fewer scams, more confidence with public services, a bit more say in work and civic life online.
Study design
Mixed methods, quasi-experimental, pre-test / post-test. One group gets AI-supported digital-literacy training. The comparison group gets ordinary digital-literacy training without the AI tools (or gets the AI programme later). That is the only honest way to credit AI rather than “training in general.”
Sites should not all look like Accra. The draft plan is urban (Greater Accra or Kumasi), peri-urban, rural, and at least one low-connectivity community. Participants would mix students, teachers, traders, public servants, older adults, people with disabilities, and local-language speakers. A sample in the 300–600 range is what I am planning for, if ethics and partners hold.
Data: a survey; a practical test (find an official service page, spot a phishing message, ask an AI tool something and then name two ways the answer could be wrong); and focus groups for the things surveys miss — trust, language, past scams, who actually owns the phone in the household.
Working definition I am using: someone is AI-literate here if they can name one useful and one risky use, complete a simple task with a tool, check the answer against another source, spot a privacy risk, recognise a likely AI-enabled scam, and know when to ask a person instead.
Materials
The GitHub repo is the study package: paper, protocol, draft instruments, sampling notes, and references. No participant-level data is published there. Instruments still need translation, piloting, and ethics approval before anyone takes them into the field.
- paper/ — current PDF
- protocol/ — design, ethics, flow
- instruments/ — survey, practical tasks, interviews
- methodology/ — sampling and analysis notes
Cite this page
Gibrine, F. M. (2026). Evaluating AI's role in public technology literacy in Ghana: A mixed-methods research protocol and conceptual framework [Preprint]. ByteLine Technologies. https://donfahd.com/research
Draft. Comments welcome at faculty12000@gmail.com.