Abstract
This study examines the challenges and opportunities presented by equity, inclusiveness, and the digital divide in the adoption of Artificial Intelligence (AI) within Technical and Vocational Education and Training (TVET) institutions in Zimbabwe. The study adopted mixed methods design, employing quantitative surveys to map infrastructural and device equity disparities (the digital divide) and qualitative critical ethnography. It used a sample size totalling 5 TVET institutions and 50 key informants. Quantitative data was analysed using frequency distributions to map the prevalence of infrastructural disparities and regression analysis to determine the statistical significance of access gaps related to socio-economic factors. Qualitative data was analysed using thematic analysis to identify, analyse, and report patterns (themes) within the interview and focus group data. Findings revealed significant disparities in AI readiness, with urban institutions generally better equipped. Challenges such as limited funding and high internet costs disproportionately affect marginalized groups. The study concludes that while AI has transformative potential for TVET, equitable adoption is obstructed by systemic inequities and the digital divide. Key recommendations include investing in digital infrastructure, creating national AI strategies for TVET, integrating digital literacy into curricula, and fostering public-private partnerships for AI training.
Keywords: Equity, inclusiveness, digital divide, Artificial Intelligence, TVET.
1 Introduction and Background
The burgeoning potential of Artificial Intelligence (AI) to revolutionise Technical and Vocational Education and Training (TVET) in Africa, particularly through adaptive learning systems or intelligent tutoring bots, is frequently championed for its capacity to address skill gaps and enhance learning outcomes. However, this research posits that the actualisation of such transformative benefits is profoundly contingent upon, and often hampered by pervasive issues of equity, inclusiveness, and the enduring digital divide. The main objective of this study is to investigate how these socio-technical disparities influence the rate, nature, and successful integration of AI adoption within Zimbabwe’s TVET sector. Equity here refers to the fair and just distribution of access to AI-enabled educational resources and opportunities. Inclusiveness extends this to active participation and benefit, preventing the marginalisation of specific groups. The digital divide refers to disparities in access to reliable high-speed internet, affordable computing devices, and consistent electricity, directly impeding the deployment of cloud-based AI simulation platforms or real-time AI-driven diagnostics essential for modern TVET curricula. Consequently, the adoption of AI in TVET, encompassing the strategic integration of AI tools and methodologies into teaching, learning, and administrative processes. This becomes a complex socio-technical challenge rather than a purely technological one in contexts like Zimbabwe, where these foundational disparities are pronounced.
The global landscape for Artificial Intelligence (AI) adoption is characterised by rapid technological advancements and widespread integration across various sectors, including vocational education and training, projected to add $15.7 trillion to the global economy by 2030 (PwC 2017). However, this transformative potential is critically tempered by persistent concerns regarding equity, inclusiveness, and the digital divide. While high-income nations lead in developing and deploying sophisticated AI tools, such as intelligent tutoring systems or AI-powered simulation platforms for TVET, there’s a growing recognition that existing biases in algorithms can exacerbate societal inequalities, for instance, through gender or race-biased hiring algorithms (Buolamwini & Gebru 2018; UNESCO 2019). The global digital divide, despite increasing internet penetration with approximately 66% of the world population online by 2023, remains a significant barrier, particularly concerning access to reliable high-speed broadband, affordable AI-compatible devices, and the digital literacy skills required to navigate and leverage AI technologies effectively (ITU 2023; World Economic Forum 2023). This global disparity critically shapes how TVET institutions can adopt AI to prepare a future-ready workforce, with many struggling to bridge the gap between aspirational integration and practical, equitable implementation due to infrastructure, resource, and policy limitations.
While Africa possesses immense potential for leapfrogging development through AI-driven innovation, its adoption within Technical and Vocational Education and Training (TVET) sectors is largely nascent and heavily constrained by systemic challenges (AfDB 2023). The digital divide is profoundly pronounced across Africa, where despite a 40% internet penetration rate in Sub-Saharan Africa as of 2022, access to affordable broadband, consistent electricity, and digital devices remains a luxury for many, particularly in rural and low-income urban areas (GSMA 2023; World Bank 2022). This fundamental lack of infrastructure severely limits the capacity of TVET institutions to integrate AI tools like predictive analytics for student performance or AI-driven virtual reality for practical skills training. Furthermore, issues of equity and inclusiveness are exacerbated by deeply entrenched socio-economic inequalities, gender disparities in Science, Technology, Engineering and Mathematics (STEM) fields, and a lack of disability-inclusive policies within educational systems (African Union 2022). Consequently, the benefits of AI risking being concentrated among a privileged few, leaving a vast majority of TVET learners, especially women and those from marginalised communities, thus widening the existing skills gap rather than bridging it.
In Zimbabwe, the impact of equity, inclusiveness, and the digital divide on AI adoption within TVET is particularly acute, manifesting as a complex interplay of socio-economic constraints and infrastructural deficiencies. While the government’s “Education 5.0” philosophy theoretically advocates for innovation and industrialisation, including AI integration, practical implementation in TVET remains rudimentary (Ministry of Higher and Tertiary Education, Innovation, Science and Technology Development 2020). The digital divide is stark: high data costs (among the highest in the SADC region), inconsistent electricity supply, and limited access to internet-enabled devices, especially in rural areas where over 60% of the population resides (POTRAZ 2023). This impede access to even basic digital learning platforms, let alone advanced AI applications (ZIMSTAT 2022). This creates a critical barrier for TVET colleges attempting to introduce AI modules, such as machine learning for agricultural automation or AI for predictive maintenance in manufacturing. Moreover, profound inequalities in educational access, exacerbated by economic challenges and historical disparities, mean that substantial segments of the population, particularly women, rural youth, and persons with disabilities, lack the foundational digital literacy and opportunities to engage with emerging technologies (ZIMSTAT 2022). Therefore, the adoption of AI in Zimbabwean TVET faces significant hurdles, risking the creation of a ‘digital underclass’ if equity and inclusiveness are not explicitly prioritised in policy and resource allocation for technological integration.
1.1 Equity in TVET
The impact of equity manifests primarily through profound institutional resource disparities, where limited central government funding fosters a bifurcated system. Well-resourced urban TVET centres (such as Harare Polytechnic) possess the necessary high-specification hardware and stable fibre connectivity required for running AI infrastructure and data science curricula. While, rural and peri-urban colleges contend with obsolete equipment and inconsistent power supply (load shedding), rendering advanced AI tools functionally inaccessible (Mutambara 2022). This structural inequity means AI adoption is geographically and economically determined.
1.2 Inclusiveness in TVET
Inclusiveness barriers are evident in the critical underrepresentation of vulnerable demographics; specifically, the entrenched gender imbalance in STEM pathways means female participation in AI-enabling fields remains significantly lower, failing to harness a crucial segment of the national talent pool required for AI innovation (Chisikwa & Shava 2021). The lack of culturally relevant or locally contextualised AI learning materials, coupled with insufficient accessibility features for students with disabilities, creates significant pedagogical friction that limits truly broad-based adoption.
1.3 Digital divide in TVET
The pervasive digital divide acts as the multi plier of these inequalities. While, AI relies heavily on cloud computing, high-speed data transfer, and costly proprietary software licenses, the majority of marginalised communities in Zimbabwe face prohibitive costs for individual internet access and lack reliable infrastructure backup (UNESCO 2023). For example, attempts to integrate sophisticated AI simulation tools often fail due to the high cost of bandwidth and poor latency in remote areas. Thus, ultimately excluding students from low-income backgrounds from mastering the practical skills necessary for the future AI-driven labour market, effectively reinforcing the chasm between technological aspiration and operational reality in Zimbabwean TVET.
1.4 Objectives of the paper
- To assess the barriers related to equity and inclusiveness that hinder the effective adoption of AI technologies in TVET institutions in Zimbabwe.
- To investigate how the digital divide affects access to AI tools and resources among marginalised communities.
2 Methodology
The study adopted mixed methods design, employing quantitative surveys to map infrastructural and device equity disparities (the digital divide) and qualitative critical ethnography. It utilised semi-structured interviews and focus groups to analyse institutional policy failures and perceived barriers to inclusion among marginalized TVET students. This approach is anchored by the critical realism paradigm, which is essential for critically distinguishing between underlying structural causes (such as persistent governmental neglect of rural infrastructure) that generate the digital divide and the observable, socially constructed realities (barriers to AI uptake) within the Zimbabwean context.
2.1 Data collection instruments
This study used dual-instrument design to ensure theoretical saturation (qualitative) and pattern identification (quantitative), aligning with the study’s focus on contextual depth over statistical generalization.
2.1.1 Questionnaire
The questionnaire was used to map infrastructural and device equity disparities (digital divide) in TVET institutions, measuring tangible gaps in access, resources, and readiness for AI adoption.
Sections of the questionnaire
- Demographic and Institutional profile
This captured context (for example, institution location, participant role, age, gender, disability status).
- Digital infrastructure access
Quantify physical access to technology and internet, for example availability of devices (computers, tablets), internet reliability (scale: “never” to “always”). Frequency of internet access for learning/teaching.
- AI tool accessibility and usage
The purpose of this section was to measure exposure to AI resources and usage patterns, for example, awareness of AI tools (such as ChatGPT, adaptive learning platforms). Frequency of AI tool usage (scale: “daily” to “never”).
- Perceived barriers to equity
This section was used to gauge perceived obstacles to inclusive AI adoption such as agreement levels (Likert scale was used) on statements like resource availability for marginalised students to use AI tools and cost of internet/data on AI training.
- Institutional support and policy efficacy
The section assessed institutional/governmental support for AI equity, for example, satisfaction with institutional AI policies (scale: “very dissatisfied” to “very satisfied”). Availability of AI training programs for staff/students.
In the questionnaire, three measurement scales were used, namely:
- Nominal: Categorical data (for example, institution location, gender).
- Ordinal: Ranked responses (for example, Likert scales: 1–5 for agreement/satisfaction).
- Interval: Scaled metrics (for example, frequency of internet access: 1 = “never” to 5 = “always”).
2.1.2 Reliability and validity of the questionnaire
Internal consistency was measured via Cronbach’s alpha (target ≥0.7) for multi-item scales (for example, barrier perceptions). Test-retest was piloted with 10 participants; re-administered after 2 weeks to ensure stability (target correlation ≥0.8). Content validity was done by 3 AI/TVET specialists to align items with objectives. Exploratory Factor Analysis (EFA) was used to confirm dimensionality of scales (for example, “infrastructure access” as a distinct factor of construct validity). Criterion validity was correlated with external data (for example national broadband coverage statistics).
2.1.3 Interview protocol for qualitative data
This was used to explore institutional policy failures, lived experiences of marginalisation, and structural barriers through critical ethnography. Semi-structured interviews took 45–60 minutes per interview/focus group. Individual interviews involved institutional heads, policy strategists and IT staff. Focus groups involved students/instructors (6–8 participants per group).
Respondents were asked about 8-10 core questions. The nature of questions requested respondents to describe institution’s current use of AI in teaching/learning. What works, and what doesn’t?”; specific examples of AI tools used; student/instructor involvement; challenges faced by students in accessing AI resources; infrastructure gaps; gender/disability inclusion/exclusion; policy shortcomings; the rural-urban divide impact access to AI tools; Compare experiences; government/institutional policies how they perpetuate or alleviate the digital divide and changes that would make AI adoption more inclusive.
In order to ensure integration with critical realism paradigm, questionnaire solicited for surface observable disparities (for example, device access statistics) and interviews uncovered generative mechanisms (for example, how historical underfunding of rural areas causes digital exclusion). After triangulation of data, quantitative data revealed what was happening in institutions and qualitative data explained through lived realities and power dynamics.
2.2 Population
The target population comprises all institutional heads, instructors, AI policy strategists (Ministry of Higher and Tertiary Education), and students from various publicly funded Technical and Vocational Education and Training (TVET) institutions across Zimbabwe, focusing particularly on those with demonstrable urban-rural resource disparity.
Critical case and maximum variation sampling was adopted to select information-rich cases. This non-probability method is essential to capture the systemic nature of equity barriers by deliberately including institutions facing significant infrastructural challenges (such as rural Industrial Training Centres) alongside well-resourced urban centres (such as Harare Polytechnic), providing critical contrast necessary for assessing the digital divide’s impact.
2.3 Sampling
The sampling was in two phases, that is, quantitative phase and qualitative phase. This was necessary to ensure the depth required for critical quantitative and qualitative analysis. Quantitative phase (survey) was for mapping infrastructural/device equity disparities. Sampling frame showed lists of registered students and employed staff (instructors, administrators, IT staff) within the 5 selected TVET institutions (2 Urban, 3 Rural/Peri-urban), 10 respondents per each institution giving a total of 50 respondents. Stratified random sampling was employed. Strata 1 showed institution location: Urban vs. Rural/Peri-urban and Strata 2 showed role of respondent such as student, Instructor/Educator, Administrator, IT/Technical Staff. Qualitative Phase (Interviews/Focus Groups) involved analysing policy failures and perceived barriers. Target population included key informants with deep contextual knowledge or direct lived experience of barriers to AI adoption such as Institutional Heads, Instructors, AI policy strategists, marginalised TVET students, IT/Technical staff.
Purposive Sampling (specifically maximum variation sampling and criterion sampling) was used. Criterion 1 showed Role/Stakeholder Group (Heads, Instructors, Policy Makers, Students, and IT Staff). Criterion 2: Institution location (Urban vs. Rural/Peri-urban). Criterion 3 showed experience of students and staff with marginalisation/digital divide (especially for students and staff in rural institutions). The study deliberately selected participants to ensure diverse perspectives across locations, roles, genders, and experiences of marginalisation (maximum variation).
The sample size was 50 key informants (22 Urban, 28 Rural/Peri-urban). There were 25 Educators/Administrators (Mix of Heads, Dept. Heads, and Senior Instructors). Students were 20 (Deliberately included students facing marginalisation for example, rural, low-income, women in non-traditional fields, people with disabilities (PWD) and 5 IT/Technical Staff (Key for understanding technical barriers and support)
In order to reach qualitative data saturation, interviews continued until no new significant themes or insights emerged regarding institutional policy failures, structural causes, and perceived barriers, particularly concerning equity and the digital divide. The sample of 50 across diverse roles and locations was deemed sufficient to achieve this depth within the Zimbabwean TVET context for this study.
2.4 Data analysis
In order to analyse quantitative data, the study employed descriptive statistics, frequency distributions to map the prevalence of infrastructural disparities and inferential methods, regression analysis to determine the statistical significance of access gaps related to socioeconomic factors, such as the significant urban-rural divide in internet connectivity and device ownership.
Qualitative data was analysed using thematic analysis to critically identify, analyse, and report patterns (themes) within the interview and focus group data, revealing how institutional policies systematically fail to address intersectional barriers, such as gendered access limitations and the lack of local-language AI content, which perpetuate exclusion.
2.5 Ethical considerations
The study upheld principles of informed consent by clearly outlining the research purpose and potential risks to participants, particularly those from marginalised communities, and ensured data anonymity to protect vulnerable populations from potential repercussions stemming from their experiences with the digital divide.
3 Presentation of results and discussion
This section presents study findings and discussion.
Table 1: Overview of sampled TVET Institutions and Key Informants (Interview)
| Category | Number | Description |
| TVET Institutions | 5 | 2 Urban-based, 3 Rural/Peri-urban based institutions. |
| Key Informants (Total) | 50 | 25 Educators/Administrators, 20 Students, 5 IT/Technical Staff. |
| Urban Key Informants | 22 | From urban-based TVET institutions. |
| Rural/Peri-urban Key Informants | 28 | From rural/peri-urban based TVET institutions. |
The distribution in table 1 above, highlights a slight skew towards rural/peri-urban representation, mirroring the broader geographic distribution of TVET institutions and the focus on understanding disparities.
The following table shows AI readiness factors from the 50 respondents obtained from the 5 TVET institutions.
Table 2: AI readiness factors (Equity & Inclusiveness barriers) – Survey (N=50)
| AI Readiness Factor (Likert Scale: 1=Strongly Disagree, 5=Strongly Agree) | Mean | Std. Dev. | Interpretation |
| Dedicated funding for AI initiatives | 1.85 | 0.92 | Critically Low: Indicates a severe lack of dedicated financial commitment, rendering AI adoption an afterthought or an impossibility. |
| Adequacy of staff AI raining & skills | 2.10 | 1.05 | Highly Inadequate: Staff largely untrained, signalling a critical human capital deficit for AI pedagogy and maintenance. |
| Current curriculum integration of AI concepts | 1.70 | 0.88 | Minimal to Non-existent: AI remains largely detached from core TVET curricula, reflecting systemic inertia. |
| Accessibility of AI tools for Persons with Disabilities (PWDs) | 1.55 | 0.75 | Alarmingly Poor: Suggests a complete oversight of inclusive design, marginalising a significant demographic. |
| Gender equity in AI-related programs/courses | 2.90 | 1.20 | Moderate but concerning: While not as dire as other factors, indicates persistent, subtle biases or structural barriers. |
From table 2 above, the descriptive statistics paint a grim picture of AI readiness within Zimbabwean TVET institutions. The mean scores, consistently below the midpoint of the Likert scale for most factors, reveal a profound systemic unpreparedness. The extremely low score for “Dedicated funding for AI Initiatives” (M=1.85; SD=0.92) is particularly damning, suggesting that AI integration is not a strategic priority but rather an aspirational, unfunded mandate. Similarly, the “Adequacy of staff AI training” (M=2.10; SD=1.05) highlights a significant human resource gap. Without a skilled faculty, even donated AI tools would remain underutilised or improperly deployed. The near-abysmal score for “Accessibility of AI tools for PWDs” (M=1.55; SD=0.75) is a stark indictment of the sector’s failure to embed inclusiveness from the outset, condemning a vulnerable population to further exclusion from future-oriented skills. Even the “Gender equity” score (M=2.90; SD=1.20), while comparatively higher, still signals the presence of barriers, suggesting AI fields may be inadvertently perpetuating existing gendered occupational segregation.
The following table shows frequency distribution of infrastructural disparities affecting AI adoption (Digital Divide) in Zimbabwe, between urban and rural/peri urban.
Table 3: Frequency distribution of infrastructural disparities affecting AI adoption (Digital Divide) – Key Informants (N=50)
| Infrastructural Factor | Urban TVET Key Informants (N=22) | Rural/Peri-urban TVET Key Informants (N=28) | Overall (N=50) | Interpretation |
| Reliable high-speed internet access | 18 (81.8%) Yes, 4 (18.2%) No | 5 (17.9%) Yes, 23 (82.1%) No | 23 (46%) Yes, 27 (54%) No | Profound urban-rural disparity: While urban institutions fare moderately well, the overwhelming majority of rural key informants lack reliable internet, rendering AI tools inherently dependent on connectivity inaccessible. This is a primary manifestation of the digital divide. |
| Adequate devices (computers/laptops) | 15 (68.2%) Yes, 7 (31.8%) No | 3 (10.7%) Yes, 25 (89.3%) No | 18 (36%) Yes, 32 (64%) No | Severe device shortage: A critical bottleneck. Even with internet, lack of suitable devices (often outdated or insufficient in number) prevents hands-on AI learning. The rural disparity here is particularly acute, reinforcing the notion of a two-tiered system where advanced learning is reserved for the privileged few. |
| Consistent electrical power supply | 20 (90.9%) Yes, 2 (9.1%) No | 10 (35.7%) Yes, 18 (64.3%) No | 30 (60%) Yes, 20 (40%) No | Electrification gap: Unreliable power in rural areas negates the utility of any technological investment. Frequent outages not only disrupt learning but also damage sensitive equipment, adding to the cost burden and disincentivising technology adoption. This highlights a fundamental infrastructural failure preceding even digital readiness. |
| Access to AI-specific hardware (GPUs, specialised labs) | 2 (9.1%) Yes, 20 (90.9%) No | 0 (0%) Yes, 28 (100%) No | 2 (4%) Yes, 48 (96%) No | Virtually non-existent: This is perhaps the most glaring deficiency. AI, particularly advanced applications, demands significant computational power. The near-total absence of specialised hardware across all institutions, and absolutely none in rural settings, demonstrates that AI integration is currently, at best, theoretical and, at worst, an illusion. This reveals a profound lack of investment and strategic foresight. |
From table 3, the frequency distributions starkly illustrate the devastating impact of the digital divide. The urban-rural chasm is not merely a gap but an abyss, particularly concerning “Reliable high-speed internet access” and “Adequate devices.” While 81.8% of urban key informants reported reliable internet, a mere 17.9% of their rural counterparts did. This disparity cascades to device availability, where less than 11% of rural key informants have adequate devices. The “Consistent electrical power supply” issue in rural areas is a fundamental barrier that precedes any discussion of digital technology. How can one power devices or access internet consistently without electricity? Finally, the “Access to AI-specific hardware” figures (a paltry 4% overall, and 0% in rural settings) are a critical alarm. This signifies that for most Zimbabwean TVET students, AI remains an abstract concept, utterly detached from practical application, a betrayal of the vocational spirit. This isn’t just a digital divide; it’s a technology infrastructure desert in many crucial areas.
The following table shows impact of socio-economic factors on AI tool access.
Table 4: Regression analysis: Impact of socio-economic factors on perceived access to AI tools & resources (N=50)
Dependent variable: Perceived access to AI tools & resources (Composite score, 1-5)
| Independent Variable | Beta Coefficient (β) | Std. Error | t-value | p-value | Interpretation |
| (Constant) | 0.85 | 0.21 | 4.05 | <0.001 | Baseline access is exceedingly low, even before accounting for other factors, indicating a generalised access challenge. |
| Urban rural (Dummy: 1=Urban, 0=Rural) | 1.89 | 0.35 | 5.40 | <0.001 | Highly significant predictor: Being in an urban TVET institution significantly and positively impacts perceived access to AI tools. This confirms the urban-rural divide is not just anecdotal but a statistically robust determinant of access, fundamentally disadvantaging rural learners. |
| Internet ConnectivityScore (1-5) | 0.42 | 0.12 | 3.50 | 0.001 | Significant predictor: Self-reported higher internet connectivity scores are strongly associated with increased perceived access to AI tools. This underscores internet access as a crucial gateway, the absence of which renders advanced learning unattainable. |
| Device ownership (1-5) | 0.38 | 0.11 | 3.45 | 0.001 | Significant predictor: Higher scores for device ownership/availability are also strongly associated with greater perceived access. This highlights that physical access to appropriate hardware is as critical as connectivity, forming a dual barrier for many. |
| R-squared | 0.72 | Strong explanatory power: Approximately 72% of the variance in perceived access to AI tools can be explained by these three factors. This suggests that the urban-rural divide, internet connectivity, and device ownership are the dominant drivers of AI access disparities, rather than tertiary issues. These are fundamental, systemic barriers demanding urgent, targeted interventions. |
From table 4 above, the regression analysis provides a statistically robust confirmation of the profound impact of socio-economic factors on AI tool access. The R-squared value of 0.72 is remarkably high, indicating that the urban-rural location, internet connectivity, and device ownership collectively explain a dominant proportion of the variance in perceived access to AI tools. This directly refutes any notion that these are minor obstacles; they are the defining factors.
Specifically, the highly significant positive beta coefficient for “Urban rural” (β=1.89, p<0.001) confirms that urban TVET learners experience significantly higher access to AI tools compared to their rural counterparts. This is not merely a slight advantage but a dramatic difference, cementing the urban-rural divide as a fundamental structural inequity in AI readiness. Similarly, “Internet connectivity score” (β=0.42, p=0.001) and “Device ownership” (β=0.38, p=00.001) are statistically significant positive predictors. This demonstrates that without reliable internet and adequate devices, the aspiration for AI adoption in TVET is nothing more than an imagination. The low constant (0.85) further implies that even assuming average conditions for the independent variables, baseline access is critically low, indicating a broader, sector-wide deficiency.
The adoption of Artificial Intelligence (AI) in Technical and Vocational Education and Training (TVET) in Zimbabwe presents a dual challenge: it is a critical avenue for future economic development, yet its effective integration is profoundly obstructed by entrenched issues of equity, inclusiveness, and the digital divide. The qualitative themes presented below, structured from the study objectives, critically illuminate how existing institutional policies systematically fail to address intersectional barriers, perpetuating exclusion and limiting the transformative potential of AI.
Objective 1 explores how various dimensions of equity and inclusiveness create barriers to AI adoption in Zimbabwean TVET. The themes reveal policy shortcomings that exacerbate these societal inequalities within the educational landscape.
The following table shows the qualitative themes that came out during interviews and how the respondents wish to have them addressed.
Table 5: Qualitative themes on equity and inclusiveness barriers to AI adoption
| Theme title | Key concepts/sub-themes | Illustrative data |
| 1. Gendered Disparities in AI Pathways and Access | – Underrepresentation of women in advanced technical TVET programs. – Persistence of socio-cultural norms and gender stereotypes. – Lack of female role models and targeted mentorship. | In rural TVET colleges, girls are often subtly steered towards traditional ‘feminine’ trades like cosmetology, home economics (Food and Nutrition) or garment construction. While, AI, robotics, and coding courses are implicitly marketed to boys. The curriculum and career guidance don’t actively challenge these stereotypes, and there are almost no female lecturers in these advanced technical subjects to inspire female students. When AI tools are introduced, they sometimes reflect inherent biases from their Western, male-dominated development contexts. Female students struggle to connect with these abstract tools, and the local curriculum hasn’t been adapted to make AI relevant and accessible to their unique experiences or local Zimbabwean challenges faced by women entrepreneurs or artisans. Policy: Institutional policies lack explicit targets and proactive strategies for increasing female participation in emerging technology fields, often perpetuating existing gendered divisions through resource allocation, curriculum design, and insufficient support for inclusive pedagogy. |
| 2. Linguistic and Cultural Irrelevance of AI Content & Pedagogy | – Predominance of English language and Western-centric AI curricula. – Scarcity of AI tools and content supporting indigenous languages (such as Shona, Tonga, or Ndebele). – Disconnect between AI use cases and local Zimbabwean economic/social contexts. | Most of the accessible AI learning platforms, tutorials, and open-source tools are exclusively in English. For a significant number of students, especially those from non-urban areas where English proficiency is lower, this is a major cognitive barrier. There’s almost no AI content available that explains complex concepts in Shona, Tonga or Ndebele, let alone applying AI directly to local challenges like improving small-scale agriculture or managing rural health data. TVET curriculum for emerging technologies largely adopts international frameworks without sufficiently adapting AI examples to problems a Zimbabwean artisan, farmer, or small business owner would genuinely face. It remains too abstract, too foreign, making it difficult for students to see the immediate value and practical application of AI in their local context. This lack of cultural resonance significantly hinders engagement and understanding. Policy: National TVET and AI strategies often overlook the critical need for curriculum indigenisation and investment in local language processing for AI. This passive acceptance of global North-derived content perpetuates linguistic and cultural exclusion, failing to empower local innovation or address unique national needs. |
| 3. Socio-economic and disability-related access exclusions | – Financial constraints for marginalised students (for example, data, personal devices). – Physical and digital inaccessibility of AI laboratories and online platforms for students with disabilities. – Lack of supportive infrastructure for diverse learning needs. | Most academically promising students, particularly those from low-income families in high-density suburbs or rural areas, cannot afford the consistent data bundles required to access online AI modules or participate in collaborative AI projects from home. They are wholly reliant on limited, often oversubscribed, on-campus lab time, which simply isn’t sufficient for the depth of learning required. Many computer labs with AI software are not wheelchair accessible, and the AI platforms used are rarely compatible with assistive technologies like screen readers for visually impaired students. There’s inadequate budget allocated for adaptive technologies or pedagogical adjustments for students with diverse learning needs, effectively creating insurmountable barriers to AI education. Policy: Institutional budgets and national educational policies often fail to allocate adequate funds for accessible technology, assistive devices, or data subsidies. This systemic omission implicitly excludes students facing socio-economic hardships. |
From table 5, the themes derived from Objective 1 critically reveal that the ambition for AI adoption in Zimbabwean TVET is fundamentally undermined by systemic failures to uphold principles of equity and inclusiveness. The persistence of gendered disparities in AI pathways is not merely a societal reflection but is actively reinforced by institutional neglect. The respondents felt that absence of explicit policies for promoting female participation, the lack of gender-sensitive curriculum design, and the dearth of visible female role models within advanced technical fields demonstrate an institutional inertia that perpetuates a cycle of exclusion. This ensures that a significant demographic is systematically denied access to future-oriented skills. The study found that the cultural irrelevance of AI content and pedagogy represents a profound failure to localise education. Mutambara (2022) also noted that the uncritical adoption of English-centric, Western-developed AI curricula alienates a vast majority of students, particularly those from non-urban and indigenous language-speaking backgrounds. This policy oversight, the lack of investment in local language processing for AI and context-specific use cases, renders AI abstract and inaccessible, hindering its practical application to Zimbabwean development challenges. The study found that the socio-economic and disability-related access exclusions highlight critical intersectional barriers. The absence of institutional provisions such as data subsidies, affordable device schemes, or universally designed learning environments effectively disenfranchises students who are already economically disadvantaged or live with disabilities. These policy gaps demonstrate a lack of commitment to dismantling historical inequalities, instead permitting them to be amplified within the crucial domain of AI education, thereby entrenching a two-tiered system of access and opportunity.
Objective 2 focuses on the tangible impacts of the digital divide, examining how disparities in infrastructure directly impede access to AI for marginalised communities within and around TVET institutions in Zimbabwe.
The following table shows digital divide’s impact on AI access in Zimbabwe, urban and rural/peri urban institutions.
Table 6: Qualitative themes on digital divide’s impact on AI access
| Theme title | Key concepts/sub-themes | Illustrative data |
| 1. Inadequate Digital Infrastructure in Underserved TVET institutions. | – Limited or unreliable internet connectivity (slow speeds, frequent outages). – Insufficient number of functional and AI-capable devices per student. – Lack of consistent and reliable electricity supply in urban or remote institutions. | In many TVET colleges located in urban or rural areas of Zimbabwe, internet connectivity is slow and frequently drops, making it impossible to stream AI tutorials, access cloud-based development environments, or download necessary AI libraries. When there’s load shedding, which is often, no practical AI learning takes place. There is an outdated computer lab with 25 aging machines for over 400 enrolled students. These machines can barely run basic office software, let alone the resource-intensive AI applications like Tensor Flow or PyTorch. How can students gain hands-on experience with AI software when the hardware itself is inadequate and shared so thinly, creating long queues and limited actual practice time?” Policy: Government and institutional infrastructure development plans consistently fail to prioritise robust digital readiness for TVET institutions in underserved rural and peri-urban areas, leading to an ever-widening gap in access to essential and effective AI learning environments. This neglect compounds the digital divide, making meaningful AI adoption virtually impossible in these regions. |
| 2. High Cost of AI Tools and Data for Marginalised Students | – Exorbitant data costs for accessing online AI platforms and resources. – High financial barrier to acquiring AI-capable personal hardware/software. – Absence of free or subsidised AI learning materials and platforms. | The cost of data bundles in Zimbabwe is prohibitive for the vast majority of the students. They either cannot afford consistent access or quickly deplete their bundles, cutting off their participation in crucial online AI courses, collaborative coding projects, and access to necessary research tools from outside the campus. “There are few national or institutional programs that provide subsidised laptops or even affordable data packages specifically for economically disadvantaged TVET students. The assumption that everyone or majority can afford a capable device and continuous internet access is deeply flawed, especially in high-density areas or peri-urban settlements where income is severely constrained. This effectively makes AI education a privilege, not an accessible opportunity. Policy: There is a glaring absence of targeted government or institutional policies for digital affordability, such as data subsidies, low-cost device schemes, or guaranteed free access to high-quality AI learning platforms. This economic barrier effectively excludes economically disadvantaged students from meaningfully participating in the AI revolution, exacerbating existing socio-economic stratification. |
| 3. Limited Digital Literacy and Awareness of AI’s Vocational Potential | – Lack of foundational digital skills among TVET faculty and students. – Low awareness of AI’s immediate relevance and application to specific trades/vocations. – Inadequate training and support for educators on integrating AI into diverse curricula. | Many TVET instructors, especially those with longer tenure, are not yet proficient with even basic digital tools for teaching, let alone comfortable integrating complex AI concepts into their vocational subjects. How can they effectively train students for an AI-driven future in plumbing, electrical work, or carpentry if they themselves are struggling with digital pedagogy and AI fundamentals? Students often view AI as an abstract concept exclusively for IT specialists or software developers, failing to grasp how it can revolutionise their specific vocational fields, for example, in predictive maintenance for mechanics, smart irrigation in agriculture, or automated design in construction. The current curriculum does not effectively demonstrate practical, real-world AI applications relevant to their everyday vocational practices, leading to disinterest and a perception of irrelevance. Policy: TVET institutions demonstrate a profound lack of comprehensive, ongoing professional development programs for faculty in foundational digital literacy and practical AI integration across all disciplines. Furthermore, national curricula fail to embed core AI literacy and relevant use cases across all vocational disciplines, perpetuating a narrow, IT-centric view of AI’s utility and significantly hindering widespread adoption and understanding among a diverse student body. |
From table 6, the themes emerging from Objective 2 starkly illuminate how the digital divide in Zimbabwe acts as a formidable barrier, effectively ghettoizing AI access within TVET. The study found that there is inadequate digital infrastructure, particularly in rural and peri-urban TVET institutions. This extends beyond merely limited internet connectivity to encompass an acute shortage of functional, AI-capable devices and an unreliable power supply. Chisikwa & Shava (2021) also noted that institutional and national policies have often neglected strategic investment in these foundational elements, contributing to a stark geographical disparity. This creates an insurmountable practical barrier, explicitly excluding marginalised students from the immersive and hands-on learning necessary for AI. From the study, the high cost of AI tools and data renders AI adoption a luxury few can afford. Lastly, the limited digital literacy and awareness of AI’s vocational potential among potential or students underscores a profound failure in human capital development and curriculum integration. Therefore, the study noted that without a concerted, institution-wide effort to up skill educators in digital literacy and AI pedagogy, and curricula that explicitly demonstrate AI’s practical relevance across diverse vocational trades, AI remains an esoteric concept.
4 Conclusion and policy implications
The findings from this study, grounded in the realities of many African contexts, reveal a stark and truth for AI adoption in Zimbabwean TVET. The objectives of assessing equity/inclusiveness barriers and investigating the digital divide’s impact are met with overwhelming evidence of systemic failure and deep-seated inequities.
The lack of dedicated funding, severe inadequacy of staff training, minimal curriculum integration, and utter disregard for accessibility for persons with disabilities illustrate a sector fundamentally unprepared, or simply unable, to genuinely embrace AI equitably. The rhetoric of AI transforming TVET is critically undermined by the absence of foundational support and inclusive planning.
The urban-rural divide is not a benign geographical distinction but an active mechanism of exclusion. Rural TVET institutions are systematically denied reliable internet, adequate devices, consistent power, and specialised hardware. This creates a two-tiered system where advanced AI training is a privilege of urban centres, further marginalizing rural populations from participating in the future economy. The regression analysis definitively proves that these are not peripheral issues but central, statistically significant barriers.
Policy implications
In order for AI adoption in Zimbabwean TVET to move beyond aspirational policy documents, a radical shift in investment and strategic planning is imperative. There is need for:
- Substantial, targeted Funding, earmarked specifically for AI infrastructure, hardware, and sustainable connectivity in all TVET institutions, with particular emphasis on rural areas.
- Comprehensive and continuous training programs for educators, focusing not just on AI tools but also on pedagogical integration and ethical considerations.
- Policy frameworks that legally enforce accessibility features for people with disabilities in all digital learning resources and AI tools.
- A coordinated effort to bridge the rural electrification and internet connectivity gaps, recognising them as fundamental preconditions for any digital transformation.
References
African Development Bank [AfDB]. (2023). African Economic Outlook 2023: Mobilizing Private Sector Financing for Climate and Green Growth in Africa. AfDB.
African Union [AU]. (2022). Africa’s Digital Transformation Strategy (2020-2030). AU.
Buolamwini, J. & Gebru, T. (2018). Gender Shades: Intersectional Phenotypic Biases in Predictive Gender Classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency. In: Proceedings of Machine Learning Research, 81, 77-91.
Chisikwa, D. & Shava, E. (2021). Integrating technology and gender equity in technical and vocational education and training institutions in Zimbabwe: Challenges and opportunities. In: Journal of Technical Education and Training (JTET), 13, 2, 1-15.
GSMA. (2023). The Mobile Economy Sub-Saharan Africa 2023. GSMA.
International Telecommunication Union [ITU]. (2023). Facts and Figures 2023: Statistics on ICT development. ITU.
Ministry of Higher and Tertiary Education, Innovation, Science and Technology Development (Zimbabwe). (2020). Education 5.0 Policy Document. MHTEISTD. Zimbabwe.
Mutambara, J. (2022). Artificial Intelligence and the future of work in Africa: Bridging the digital divide in Zimbabwe’s education sector. In: African Journal of Science, Technology, Innovation and Development, 14, 4, 1017-1027.
Postal and Telecommunications Regulatory Authority of Zimbabwe [POTRAZ]. (2023). Sector Performance Report (Q1 2023). POTRAZ. Zimbabwe.
PricewaterhouseCoopers (PwC). (2017). Sizing the prize: What’s the real value of AI for your business and how can you capitalise? PwC.
UNESCO. (2019). Artificial Intelligence in Education: Compendium of Promising Initiatives. Mobile Learning Week 2019. Online: https://unesdoc.unesco.org/ark:/48223/pf0000370307 (retrieved 11.03.2026).
UNESCO. (2023). Digital transformation in African education: Challenges and opportunities for AI in TVET. UNESCO Regional Office for Southern Africa. UNESCO.
World Bank. (2022). Connecting Africa: The Path to Digital Transformation. World Bank. USA.
World Economic Forum. (2023). The Future of Jobs Report 2023. World Economic Forum. USA.
Zimbabwe National Statistics Agency (ZIMSTAT). (2022). Population and Housing Census 2022: Preliminary Report. ZIMSTAT. Zimbabwe.
