Bridging Educational Gaps in Low-Resource Classrooms: AI-Enhanced Project-Based Learning for STEM Educators in Africa

Abstract

Purpose

This research examines the potential of Artificial Intelligence (AI)-Project-Based Learning (PBL) to enable science, technology, engineering, and mathematics (STEM) educators in resource-limited African classrooms to shift from content-centric teaching to competency-based learning (CBL). It seeks to ascertain how cost-effective AI tools can address deficiencies in teacher capacity, inclusivity, and digital accessibility.

Design/Approach/Method

Using a qualitative research design, data were collected from ten STEM educators across Nigeria, Botswana, Ghana, Namibia, and Sierra Leone through open-ended interviews, document analysis, observations, and analysis of participants’ project artefacts. The intervention involved hands-on engagement with speech-to-text-to-image generation, smartphone-based block coding using the Magnetcode application, circuit simulation, and microcontroller-based prototyping.

Findings

Thematic analysis revealed five outcomes: AI speech-to-text tools enhanced visualization of science concepts; smartphone-based coding increased inclusion; simulations provided cost-effective scaffolding for hardware use; educators gained transition skills in computational thinking; and participants developed concrete strategies for classroom integration.

Originality/Value

This study contributes novel empirical evidence from Sub-Saharan Africa, demonstrating that low-cost, AI-supported PBL can facilitate learner-entered pedagogy and equitable STEM innovation. It highlights AI as a transformative enabler for CBL, offering scalable models for sustainable, inclusive education in resource-constrained environments.

Introduction

The rapid integration of artificial intelligence (AI) into education is reshaping how educators design, deliver, and assess learning experiences worldwide (UNESCO, 2025). AI-driven tools such as intelligent tutoring systems, natural language interfaces, and block-based coding applications have shown significant potential to enhance problem-solving, creativity, and learner autonomy (Taj & Jhanjhi, 2022; Triplett, 2023). However, much of this progress has occurred in high-resource contexts, leaving low-resource classrooms, particularly across Africa, at the margins of digital transformation (African Union Commission, 2025; Atuhura & Nambi, 2024). Addressing this imbalance requires innovative, context-responsive pedagogies that leverage affordable AI technologies to foster inclusive, competency-based science, technology, engineering, and mathematics (STEM) education (Nørgaard et al., 2024).

In response to these challenges, the African Union's Continental Education Strategy for Africa 2026–2035 (CESA 26–35) underscores the urgent need to strengthen STEM education, digital literacy, and AI integration as catalysts for equitable and sustainable development (African Union Commission, 2025). Within this framework, professional development is identified as a key driver for systemic reform (Zickafoose et al., 2024). Yet, educators in many African contexts face persistent infrastructural and capacity limitations that constrain their ability to adopt emerging technologies effectively (Mutsvangwa & Zezekwa, 2021). Consequently, there remains a critical gap between policy aspirations and classroom realities (Gottschalk & Weise, 2023).

Education systems are undergoing a transition from traditional rote-based instruction to competency-based learning (CBL) (Phan, 2024). CBL emphasizes transferable skills such as creativity, adaptability, problem-solving, and CT (Lupeja & Komba, 2021; Phan, 2024). This paradigm shift is particularly urgent in STEM education, where learners are expected not only to acquire conceptual knowledge but also to apply it in addressing pressing challenges such as climate change, sustainable agriculture, and resource management (Öztemur, 2024; UNESCO, 2020). In this regard, CBL frameworks provide an avenue for nurturing twenty-first-century skills that are vital for both local and global development.

Recent advances in AI have introduced new opportunities for transforming teaching and learning. Applications such as speech-to-text, text-to-image generation, adaptive learning platforms, and intelligent tutoring systems are increasingly integrated into classrooms to promote personalized, inclusive, and contextually relevant learning (Kumar et al., 2023; Yim & Su, 2025). Within STEM education, AI not only supports conceptual understanding but also enables hands-on and multimodal learning experiences that mirror real-world problem-solving scenarios (Kong et al., 2024). Parallel to AI adoption, project-based learning (PBL) has emerged as a complementary pedagogy that empowers learners to engage in inquiry, design, and innovation while demonstrating competencies through authentic projects rather than conventional examinations (Dacumos, 2023; Hsieh et al., 2022).

Despite growing international interest in AI-supported PBL, there is limited empirical evidence on how such approaches can be implemented in low-resource African classrooms. The adoption of AI and PBL in African educational contexts is constrained by structural barriers, including unreliable internet access, limited access to digital devices, inadequate laboratory infrastructure, and insufficient teacher preparedness (Muyunda et al., 2023; Oliver et al., 2022; Onaolapo et al., 2020; Phiri et al., 2022). These constraints hinder the scalability of emerging pedagogies and widen existing inequalities (Onaolapo et al., 2020).

Previous research has shown that low-cost mobile devices, block-based coding platforms, and simulation tools can serve as effective entry points for educators and learners in resource-constrained environments (Koshiry et al., 2024; Nurul Farahah & Eh Phon, 2020). For instance, block-based coding on smartphones and circuit simulation environments enable students to develop scientific problem-solving skills at minimal cost before transitioning to physical microcontroller-based applications (Sanura et al., 2025; Sanura & Rozniza, 2024). While recent scholarship has advanced the development of AI literacy frameworks (Kong et al., 2024) and systematic reviews of AI integration in K-12 education (Lee & Kwon, 2024), there remains limited empirical evidence on how AI-supported PBL can be effectively contextualized for low-resource African classrooms. Most existing studies on AI integration have been conducted in high-resource or experimental settings, overlooking the pedagogical adaptations required for teachers in Sub-Saharan Africa, where infrastructure and training gaps persist (Angwaomaodoko, 2025). Addressing this gap, this study provides empirical insights from a capacity-building workshop with science educators from Nigeria, Botswana, Ghana, Namibia, and Sierra Leone.

The purpose of this study is to explore how AI-enhanced PBL can bridge educational gaps in low-resource African classrooms by empowering STEM educators to transition from traditional, content-driven instruction toward competency-based and learner-centered practices. Specifically, the study investigates how educators engage with AI-supported PBL approaches to develop teaching conceptual understanding (RQ1) and how they apply these AI-supported strategies to transform instructional practices and promote CBL in their classrooms (RQ2). By addressing these questions, the research aims to provide empirical insights into how low-cost, AI-enabled pedagogies can promote inclusive, equitable, and future-oriented STEM education reform across Africa.

Literature Review

CBL and the Role of AI in STEM Education

In recent years, CBL has gained momentum as a framework to replace traditional exam-driven education with approaches that emphasize mastery of skills and authentic application of knowledge. Lupeja and Komba (2021) stressed that CBL fosters deeper learning by encouraging students to develop adaptability, creativity, and problem-solving skills that are essential for navigating complex, real-world challenges. UNESCO (2020) has further positioned CBL as central to building future-ready learners capable of addressing global sustainability issues, particularly in STEM fields where conceptual understanding must connect with practical innovation. At the same time, AI is increasingly recognized as a transformative driver of educational reform. Angwaomaodoko (2025) argued that AI-supported and digital tools ranging from natural language processing to adaptive tutoring systems offer new opportunities for personalizing learning, providing timely feedback, and creating inclusive learning environments. In resource-limited educational contexts, AI has increasingly been described as a “leapfrogging technology,” a transformative enabler capable of compensating for infrastructural deficits by facilitating interactive, adaptive, and engaging learning experiences (Lupeja & Komba, 2021; Onah, 2025; UNESCO, 2025).

PBL supports inquiry-driven STEM learning by enabling learners to apply knowledge to authentic problems (Fu et al., 2024). In the African context, PBL aligns with the need to prepare learners for sustainability-linked careers, where problem-solving and innovation are critical (Nørgaard et al., 2024; Pérez Torres et al., 2024). Empirical studies affirm the benefits of PBL. Nørgaard et al. (2024) found that students participating in PBL demonstrated greater independence and resilience, while Sanura and Rozniza (2024) highlighted its potential as a catalyst for integrated PBL in STEM education. Fu et al. (2024) further validated the development of reliable scales to measure learners’ PBL experiences, underscoring its academic legitimacy as a pedagogical framework. Thus, implementing PBL in low-resource classrooms presents unique challenges, particularly where laboratories, teaching aids, and ICT infrastructure are lacking. Scholars have emphasized the importance of low-cost technological innovations to make PBL feasible. Block-based coding applications designed for Android smartphones can democratize access to CT by eliminating the need for expensive laptops or continuous Internet (Sanura, et al., 2025). Similarly, Kim et al. (2023) showed that simulation environments allow learners to practice coding and circuit design in a virtual space, building confidence before transitioning to physical hardware. These innovations are particularly relevant in African schools, where teachers often have to improvise due to the scarcity of physical lab equipment (Kim et al., 2023). The integration of microcontroller-based projects further enriches PBL by enabling learners to prototype tangible solutions. For example, simple servo- and sensor-based systems have been used to connect science content with agriculture and sustainability issues, offering students opportunities to apply knowledge in contextually meaningful ways (Berciano et al., 2025; Blackley & Howell, 2019). Therefore, these studies point to PBL as both a feasible and impactful strategy when aligned with low-cost, resource-appropriate tools.

While technological innovations are promising, the role of educators is central to realising sustainable change. Multiple studies emphasize that educators’ beliefs, skills, and professional development determine the extent to which AI and CT are successfully integrated into classrooms (Mnguni, 2024; Onah, 2025; UNESCO, 2025). In Africa, professional development remains a pressing challenge: Educators often lack structured opportunities to engage with AI-supported and digital tools or computational pedagogy, leading to low confidence and a reliance on traditional lecture-based methods (Mnguni, 2024; Mutsvangwa & Zezekwa, 2021). Nonetheless, emerging programs show the potential of targeted training to foster STEM educators’ transformation. Lee et al. (2024) reported that PBL-oriented professional development can shift teachers’ mindsets toward learner-centered instruction (Lee et al., 2024). Similarly, UNESCO (2025) underscored that sustainable adoption of AI in African education requires not just technology access but also culturally relevant professional development that empowers teachers as innovators rather than passive implementers.

Simulation-first strategies and smartphone-based coding platforms are particularly well-suited for teacher training in Africa (Wijnen-Meijer et al., 2022). Sanura and Rozniza (2024) demonstrated that simulations reduce risk and cost while enhancing teachers’ confidence before deploying physical devices, a finding echoed in interventions that introduced microcontrollers in under-resourced schools. Moreover, Mnguni (2024) highlighted that AI must be contextualized for African classrooms, leveraging mobile penetration, local languages, and community-based problem-solving. These insights point to the importance of capacity-building workshops that integrate AI, CT, and PBL in ways that are both technologically feasible and pedagogically transformative. Existing studies discuss AI literacy and teacher readiness but lack empirical evidence from real classroom interventions in Africa. By exposing teachers to speech-to-text-to-image (STTI) tools, mobile block coding, and microcontroller-based projects, such initiatives not only enhance teacher competence but also lay the foundation for systemic curriculum reform aligned with CBL, as illustrated in Figure 1.

Figure 1.

Conceptual model linking artificial intelligence (AI), computational thinking (CT), project-based learning (PBL), and competency-based learning (CBL).

Figure 1 illustrates how AI functions as a technological scaffold that enhances the PBL process, fostering the development of CT skills that ultimately support CBL outcomes. Through iterative, experiential learning cycles, educators engage in reflective practice and adaptive problem-solving, thereby bridging pedagogical innovation and competency development in low-resource STEM classrooms. This relationship is consistent with prior studies highlighting the role of AI-supported and digital tools in mediating inquiry-driven, hands-on learning experiences (Atuhura & Nambi, 2024; Mulenga & Kabombwe, 2019), the effectiveness of PBL in cultivating CT skills (Kong et al., 2024), and the alignment between CT-oriented pedagogy and CBL frameworks that emphasize performance-based mastery and adaptability (Muyunda et al., 2023; Rich et al., 2020). Collectively, these findings support the notion that AI-enhanced PBL environments create transformative, feedback-rich learning ecosystems that promote equity and competence in STEM education across diverse educational contexts.

The Magnetcode Application as a Pedagogical Enabler in Low-Resource STEM Education

Magnetcode is a free, Android-based application designed to introduce coding and electronic circuit concepts through an accessible, block-based programming environment (Kim et al., 2023; Vinueza-Morales et al., 2025). Its simplified interface enables educators and students to create algorithms visually, minimizing the cognitive barriers associated with formal programming syntax. From a pedagogical perspective, this design aligns with constructionist learning theory, which emphasizes knowledge construction through active engagement and artifact creation (Butler & Leahy, 2021). By allowing users to design programs through visual command blocks, Magnetcode transforms abstract algorithmic concepts into tangible, experiential learning processes that foster CT and problem-solving.

Magnetcode's integrated circuit simulation feature extends its pedagogical utility by enabling virtual experimentation prior to physical implementation. This simulation environment allows users to test coding logic and visualize circuit behavior, such as LED activation, buzzer operation, or servo rotation, without relying on actual hardware components (Sanura & Parvinder, 2024; Wan Nurlisa et al., 2023). Analytically, this feature embodies the principles of Kolb experiential learning theory (Villarroel et al., 2025), as it allows learners to progress through cycles of conceptualization, testing, reflection, and refinement (Sutton, 2025). Such iterative engagement strengthens conceptual understanding of cause-and-effect relationships in electronic systems while building confidence in programming accuracy before physical prototyping.

In the context of low-resource African classrooms, this simulation-first strategy represents a transformative approach to STEM pedagogy. It mitigates cost and equipment constraints while offering a safe environment for exploration and error correction. In doing so, Magnetcode contributes to what UNESCO (2025) described as “technological inclusivity” using accessible digital tools to promote equitable participation in emerging technologies. Moreover, by supporting both teacher-led demonstrations and student-driven exploration, the platform reinforces CBL's focus on competency demonstration through authentic, hands-on problem-solving.

The educational implications of Magnetcode thus extend beyond its functional simplicity. It exemplifies how AI-supported PBL can be localized for contexts characterized by infrastructural limitations. Through its smartphone-based accessibility and simulation capability, the application facilitates a continuum between conceptual learning and practical implementation, bridging the gap between digital literacy and applied STEM competence. As educators integrate Magnetcode into their practice, they are not merely teaching coding; they are cultivating adaptive thinking, a design orientation, and reflective problem-solving as core competencies for twenty-first-century STEM education in Africa.

Within low-resource educational contexts, this simulation-first strategy exemplifies an equity-oriented approach to STEM pedagogy. By enabling coding and circuitry experimentation via smartphones without reliance on laboratory infrastructure, Magnetcode transforms resource limitations into opportunities for inclusive participation. Thus, the platform extends the reach of project-based and computational learning, demonstrating how accessible digital tools can foster inquiry, creativity, and technological empowerment in under-resourced African classrooms, as illustrated in Figure 2.

Figure 2.

An example of a circuit simulation created using the Magnetcode application, demonstrating how educators visualize and test coding logic before transitioning to physical microcontroller-based projects.

Applying Kolb's Experiential Learning Theory in AI-Enhanced PBL

In this study, Kolb's experiential learning theory is applied as the core learning process model guiding the design, implementation, and analysis of AI-enhanced PBL. The theory provides a structured lens for understanding how educators and learners transform experience into knowledge through iterative engagement, reflection, conceptualization, and experimentation (Hung et al., 2023). In the context of AI-enhanced PBL, such scaffolding is provided not only by teachers and peers but also by digital technologies, including block-based coding platforms, STTI applications, and AI-supported and digital tools simulators that guide and extend learners’ capabilities (Nkosi & Ramaligela, 2025). These tools act as cognitive and social mediators, allowing educators and students to engage with complex STEM concepts that might otherwise be inaccessible in low-resource classrooms (Atteh, 2023; Mafi, 2022). The integration of AI thus transforms traditional scaffolding into a techno-pedagogical form, facilitating both individualized and collaborative learning. Kolb's (1984) experiential learning theory complements these constructivist underpinnings by articulating how experience is transformed into knowledge through an iterative cycle of concrete experience, reflective observation, abstract conceptualization, and active experimentation (Hung et al., 2023). Within this study, AI-enhanced PBL embodies this cycle as illustrated in Figure 3.

Figure 3.

Integration of Kolb's experiential learning in AI-enhanced PBL.

Methodology

This study employed a qualitative research design to investigate how science educators from selected African countries experienced and applied AI-supported PBL in a professional development context. An interpretivist qualitative perspective was conducted, which assumes that researchers really come alive at the stage of being able to enter the study setting and interact with their sources of data (Thorne, 2025). To increase and provide evidence of the rigor of participants’ experiences and social interactions (Maher, 2025), the researcher adopted a dual role as both facilitator of the AI-supported and digital tools enhancing the PBL workshop and qualitative researcher. This insider position enabled close observation of participants’ learning processes, reflections, and pedagogical shifts, while also requiring ongoing reflexivity to minimize potential bias (Maher, 2025; Oluka, 2025; Thorne, 2025).

To address this, the researcher maintained analytic memos throughout data collection and analysis, documenting assumptions, emerging interpretations, and moments of pedagogical influence (Creswell & Guetterman, 2021; Thorne, 2025). The design combined multiple strategies, including document analysis, project-based activities, and the evaluation of participants’ project outcomes, to capture a holistic view of the training process (Merriam & Elizabeth, 2016; Nicmanis, 2024; Yazan & De Vasconcelos, 2016). According to Thorne (2025), this approach supported transparency in how interpretations were shaped to ensure the analytic claims remained grounded in participants’ voices rather than the researcher's instructional intentions.

Data Analysis Procedures

The analysis process followed an iterative and comparative approach to integrate findings from the different data sources (Papavasileiou & Dimou, 2024). Data analysis followed a reflexive thematic analysis approach (Nicmanis, 2024), combining inductive and deductive strategies. Interview transcripts, observation notes, participant artefacts, and digital logs were analyzed through multiple iterative coding cycles (Thorne, 2025).

In the first cycle, open inductive coding was conducted to identify recurring patterns in participants’ descriptions of their learning experiences, challenges, and pedagogical shifts. In the second cycle, deductive coding was applied using sensitizing concepts derived from Kolb's experiential learning theory (i.e., concrete experience, reflective observation, abstract conceptualization, and active experimentation) and the study's conceptual framework linking AI, PBL, CT, and CBL.

Interview transcripts and observation notes were thematically coded using both inductive and deductive strategies derived from the study's research questions and theoretical framework. Digital artefacts from the Magnetcode platform, such as coding logs, simulation results, and project prototypes, were analyzed to corroborate participants’ reported experiences and teaching practices. The researcher identified patterns across these datasets and cross-validated them to ensure consistency and to identify convergence or divergence in perspectives. This comparative synthesis enhanced the analytical rigor of the study and reflected the triangulation logic described by Papavasileiou and Dimou (2024), where multiple evidence streams are integrated to reinforce validity and interpretive depth. Thematic saturation was considered achieved when no substantively new codes or conceptual insights emerged from successive interviews and artefact analyses (Creswell & Guetterman, 2021; Göçen, 2021; Maher, 2025; Naeem et al., 2023; Nicmanis, 2024; Thorne, 2025).

Participants and Sampling

The selection of participants reflected a deliberate effort to capture diversity across educational environments in the region. The cohort comprised ten STEM educators (five male and five female) actively engaged in teaching science at the secondary and post-secondary levels. All participants joined the workshop as part of their ongoing professional development initiatives. They were drawn from five African countries, as shown in Figure 4.

Figure 4.

Participants’ demographic.

Although the sample size was modest, it was considered appropriate for a qualitative, exploratory design (Schoch, 2020), where the emphasis lies on obtaining rich, contextually grounded insights rather than generalizable statistical outcomes (Nicmanis, 2024). By intentionally including educators from multiple African countries, the study ensured that the perspectives represented a broad spectrum of challenges and opportunities characteristic of low-resource STEM classrooms across the continent.

Trustworthiness and Rigor

Before conducting the workshop, the researcher rigorously followed in accordance with the research ethics guidelines, where ethics accountability was crucial for different types of research-related activities (Knight et al., 2024; Kolstoe et al., 2025). For high-quality qualitative research, that is planned, designed, and conducted, it must be rigorous (Maher, 2025). To strengthen the research, the researcher must follow the plans by carefully examining the reasoning and ensuring precision in conducting the study (Kumar et al., 2025). According to Kumar et al. (2025), due to its in-depth presentation, Lincoln and Guba's model has outlined the four criteria for the evaluation of trustworthiness.

In this research, the credibility was enhanced through triangulation of interviews, observations, document analysis, and participants’ digital artefacts. This prolonged engagement during the AI-enhanced PBL workshop enabled the researcher to observe participants’ learning processes across multiple experiential cycles. It started with initial exposure to AI tools through reflection and application. Verbatim quotations were used to ensure that findings accurately represented participants’ perspectives (Noble & Smith, 2025).

Trustworthiness was ensured through credibility, dependability, confirmability, and reflexivity (Kumar et al., 2025). To ensure the dependability of the research, the researcher applied a systematic and well-documented analytic procedure by combining inductive and deductive strategies. To address the confirmability of the research, the researcher maintained reflexive journals to acknowledge positionality as a workshop facilitator and to monitor potential bias. Thus, the researcher grounded data interpretations in participants’ narratives and corroborated them with observational evidence and digital artefacts, ensuring that the findings reflected participants’ experiences rather than researcher expectations.

Prior to participation, all educators were provided with comprehensive information about the study. This included detailed descriptions of the research context, participant characteristics, workshop design, learning activities, data collection methods, and the intended use of research outputs. These details were communicated together with the informed consent process, ensuring that participants clearly understood the scope, purpose, and ethical considerations of the study before agreeing to participate (Adley et al., 2024; Knight et al., 2024). By signing the consent form before data collection commenced, participants agreed to the terms outlined in the consent letter, which specified the study's purpose, procedures, confidentiality provisions, their right to withdraw at any stage without penalty, and the assurance that all responses would be used solely for academic purposes (Noble & Smith, 2025; Thorne, 2025).

Results and Discussion

The purpose of this study was to explore how science educators from five African countries— Nigeria, Botswana, Ghana, Namibia, and Sierra Leone—experienced and reflected on their participation in an AI-supported and digital tools project-based training workshop. Data were collected primarily through open-ended interviews, complemented by observation, document analysis, and participants’ project outcomes. Using thematic analysis, five major themes emerged, illustrating how educators engaged with the training content and how it influenced their transition toward competency-based teaching practices in low-resource contexts. The themes are illustrated in Figure 5.

Figure 5.

The five major themes.

Across the dataset, educators reported that (1) AI-enabled multimodal tools enhanced conceptual understanding and inclusivity, (2) smartphone-based coding expanded access to CT, (3) simulation tools functioned as effective scaffolds for physical prototyping, (4) educators developed transition skills aligned with CT and facilitative pedagogy, and (5) participants articulated concrete plans for classroom integration despite contextual constraints. In relation to the research questions, themes 1, 2, and 3 align with RQ1 by capturing educators’ conceptual understanding and pedagogical engagement with AI-supported and digital tools, whereas themes 4 and 5 align with RQ2 by evidencing instructional transformation and the planned application of AI and CT strategies in classroom practice.

Theme 1: Exploring AI STTI for Science Education

Educators consistently reported that AI-supported and digital tools, including STTI applications, enhanced learners’ understanding of abstract science concepts by providing immediate visual representations linked to verbal input. Participants emphasized that this multimodal support was particularly valuable in classrooms where learners faced language barriers or limited prior exposure to scientific imagery.

“It helps link three concepts together, and even supports learners in pronunciation practice.” (P7)

“With AI-generated images, learners can see practical applications of what they study.” (P6)

[26 September 2025]

These responses highlight that AI-generated visualization acted as a cognitive scaffold within project-based tasks, enabling learners to connect spoken explanations with concrete representations rather than functioning as a standalone instructional approach. The findings are consistent with multimedia learning principles (Alam, 2024). Thus, educators framed AI as an assistive pedagogical aid rather than a replacement for instructional design or teacher guidance. Prior studies also suggest that AI-generated visualizations can strengthen student engagement by bridging language and conceptual gaps (Bers et al., 2023). This activity demonstrated the accessibility of STTI tools for science visualization, as illustrated in Figures 6 and 7.

Figure 6.

Participants coding the commands using the Magnetcode application for AI speech-to-text-to-image contents.

Figure 7.

Participants using an AI speech-to-text-to-image tool on a smartphone to generate a cow image.

Below are the commands that the participants used for the AI STTI coding.

Text_Speech_Enable = On

Speech_Text_Enable = On

LCD_Show = On

LCD_Text_Size = 50

LCD_Back = Transparent

LCD_Front = Blue

Speech_Text

∼If speech is a cow

If P_Speech_Text = a cow

LCD_Show_Text = a cow

LCD_Update

Background_Image = cow.jpg

Text_Speech_Text = This is a cow. They live in groups

End If

∼If speech is a chicken

If P_Speech_Text = a chicken

LCD_Show_Text = a chicken

LCD_Update

Background_Image = chicken.jpg

Text_Speech_Text = This is a chicken. They live in the flock

End If

Timer=2 sec

Goto=1

Figure 7 shows a participant holding a smartphone with an AI-generated visualization of a cow. This activity illustrates how STTI tools can be applied in science education to support concept identification and visualization of biological content. By converting verbal prompts into immediate visual feedback, participants were able to connect abstract ideas with concrete representations, making learning more engaging and accessible. This aligns with the participants’ reflections, such as

“It transformed abstract ideas into vivid, accessible visuals.” (P3)

“It helps link three concepts together.” (P7)

[26 September 2025]

Theme 2: Block-Based Coding on Smartphones in Low-Resource Contexts

Participants identified smartphone-based block coding as a practical and inclusive entry point to CT in low-resource contexts. While some educators expressed an initial preference for laptops, most acknowledged that smartphones were more readily available and realistic for classroom use.

“Using the smartphone for block-based coding was awesome and rewarding.” (P5)

“With phones, learners in rural schools won’t be left out.” (P1)

[26 September 2025]

The findings indicate that mobile devices enabled broader participation in coding activities without reliance on expensive infrastructure. Educators noted that block-based interfaces reduced syntactic complexity, allowing learners to focus on problem-solving and logic rather than programming language rules (Tawfik et al., 2024). Challenges such as small screen sizes and school policies restricting phone use were acknowledged, but these were framed as implementation constraints rather than pedagogical barriers. These insights underscore the role of mobile devices in democratizing coding education in contexts where laptops and Internet access are scarce. This supports findings from Villarroel et al. (2025), who argued that mobile technologies extend learning opportunities in under-resourced settings. However, the challenges of small screen sizes, institutional restrictions on phone use, and initial learning curves highlight the need for adaptive pedagogies and supportive school policies (Hadi et al., 2020). The participants demonstrated how to use the smartphone to practice coding in low resource, as shown in Figure 8.

Figure 8.

Participants practicing block-based coding on a smartphone using the Magnetcode application.

Theme 3: Simulation as a Bridge to Physical Microcontroller Projects

Simulation tools were widely valued as a preparatory stage before engaging with physical microcontroller projects. Participants described simulation as a low-risk environment that allowed experimentation, debugging, and conceptual testing without fear of damaging hardware. Participants strongly valued simulation as an intermediate stage that prepares learners for real-world hardware engagement. They described simulations as safe, cost-effective, and confidence-building:

“It gave me confidence to build a real circuit after testing it in the simulator.” (P4)

“The immediate feedback from the simulator was very helpful.” (P7)

[26 September 2025]

Simulation enabled educators to iteratively refine logic and design decisions, supporting a gradual transition from abstract coding to tangible artefact construction. Participants emphasized that this process was especially important in contexts where hardware resources were limited, and mistakes could be costly. As such, simulation was perceived not as a substitute for hands-on learning but as a bridge that strengthened readiness and confidence (Fiandra et al., 2022; Kuehne, 2020). These findings align with constructivist perspectives that emphasize iterative problem-solving and experiential learning (Hung et al., 2023). By offering immediate feedback and reducing the risk of costly mistakes, the educators emphasize that simulations act as cognitive scaffolds, smoothing the transition from abstract coding to hands-on prototyping. Similar results have been documented in engineering education, where simulation increases self-efficacy and shortens the learning curve (Fiandra et al., 2022).

Figure 9 illustrates the participants’ successful circuit simulation linking a microcontroller (CARROT board) to an LED, illustrating how simulation builds confidence and reduces risk before transitioning to physical projects.

Figure 9.

Participants’ successful creation of a simple circuit simulation using the Magnetcode application.

Theme 4: Transition Skills into AI and CT

The workshop not only improved technical competence but also facilitated a pedagogical shift. Educators reported a shift from teacher-centered delivery toward facilitation and learner-centered design:

“This training has built my confidence to integrate computational thinking into STEM projects.” (P10)

“I learned to focus more on problem-solving and creativity rather than just content delivery.” (P3)

[26 September 2025]

These skills enabled participants to design learning experiences where students actively constructed knowledge through projects rather than passively receiving information. This shift reflects alignment with CBL principles, where the demonstration of skills and processes takes precedence over rote content acquisition. These responses highlight the development of transition skills such as adaptability, problem-solving, and a growth mindset, which are crucial for embedding AI and CT in competency-based education. This aligns with Mnguni's (2024) call for integrating AI literacy as a critical twenty-first-century competency. By reframing their roles as facilitators, participants demonstrated readiness to create inquiry-based and student-driven learning environments.

Figure 10 illustrates participants actively engaged in coding tasks and collaborative hands-on activities. They successfully integrated AI applications with Magnetcode programming, demonstrating their ability to apply CT skills, such as problem decomposition, abstraction, and debugging, in real time. Participants reflected on their shifting pedagogical roles as follows:

“I now see myself more as a facilitator, giving space for learners to problem-solve with AI tools.” (P8)

“This training has built my confidence to integrate computational thinking into STEM projects.” (P10)

[26 September 2025]

Figure 10.

Participants actively engaged in coding tasks and collaborative hands-on activities.

This finding is consistent with Kolb's (1984) experiential learning theory, which underscores the importance of reflective engagement through hands-on experimentation. The activities demonstrated that participants were not passive recipients of knowledge but active designers of learning experiences, collaboratively testing solutions, sharing strategies, and iteratively refining their projects.

Theme 5: Plans for Classroom Integration of AI and CT Strategies

Educators articulated concrete strategies for classroom implementation, ranging from co-curricular clubs to PBL initiatives:

“I will create an AI club in my school so learners can continue experimenting.” (P9)

“I will use speech-to-text-to-image for concept art and pronunciation exercises in science.” (P6)

[26 September 2025]

Participants demonstrated strong agency in adapting workshop practices to their local contexts, indicating that professional development had fostered practical, context-responsive implementation planning rather than abstract aspiration. These statements reflect the proactive adoption of AI and CT pedagogies, but also acknowledge systemic barriers such as restrictive device policies and resource constraints. Similar constraints have been reported in technology integration research, where institutional support and the availability of resources are decisive factors for sustainability (Maambo, 2023; Mulenga & Kabombwe, 2019).

Figure 11 shows participants constructing a prototype with a Magnetcode microcontroller, coding its functions using Magnetcode, and linking them to AI-enhanced tasks. For instance, one group created a simple IoT-based scarecrow system using a servo motor, sensors, and microcontrollers—a system that could be easily replicated by students in resource-constrained classrooms. Participants articulated their intentions to transfer these practices into their schools, as shown in the responses below.

“I will create an AI club in my school so learners can continue experimenting.” (P9)

“We plan to design models using recycled bottles and code them with Magnetcode.” (P2)

[26 September 2025]

Figure 11.

Participants constructing a prototype.

The evidence shows that participants not only gained technical skills but also designed context-relevant strategies for classroom adoption. Their hands-on activities illustrate the feasibility of embedding AI and CT practices into STEM projects. These tools complement AI-enabled features by providing structured environments for experimentation and application. This distinction underscores that learning outcomes were driven by pedagogical design and project-based implementation, with AI serving as an enabling component rather than the sole driver of change. In this study, the role of AI is the key composite element of the learning process. However, AI should be considered not only from the technological point of view; it also requires the human “perspective” (Mainardi, 2025) to enhance PBL for STEM Educators in Africa.

Conclusion and Recommendations

This study explored educators’ experiences in integrating AI and CT through hands-on workshops that combined STTI tools, smartphone-based block coding, simulations, and physical microcontroller projects. The findings revealed that participants perceived AI tools as valuable in making abstract science concepts more concrete and accessible, particularly through multimodal representations that enhanced learner engagement and understanding (Baboolal & Singaram, 2023; UNESCO, 2025). The use of smartphones for block-based coding was highlighted as a practical and inclusive strategy, enabling participation even in low-resource contexts where laptops are scarce, thereby affirming earlier research on mobile learning as a driver of digital equity (Sanura, et al., 2025; Taha & Dahabiyeh, 2021). Simulation was particularly effective as a scaffold, providing immediate feedback and reducing risk before transitioning to physical projects, thereby increasing confidence and competence among educators (Sanura et al., 2025). Importantly, the workshops also fostered transition skills, with participants shifting their pedagogical orientation from content delivery to facilitation, aligning with the call for AI literacy and CT as essential competencies for twenty-first-century education.

Based on these insights, several recommendations can be drawn. First, AI and CT should be systematically embedded into STEM curricula through project-based and inquiry-driven modules that emphasize both simulation and physical prototyping, ensuring learners develop authentic problem-solving skills (Grover & Pea, 2017). Second, resource accessibility remains critical, and schools should adopt smartphone-first and simulation-based approaches as viable alternatives in low-resource environments. Third, sustained professional development is needed to equip educators with transition skills, particularly facilitation, formative assessment, and growth mindset practices, so that the use of AI and CT moves beyond technical novelty into pedagogical transformation (UNESCO, 2025). Finally, schools are encouraged to establish AI and CT clubs or extracurricular initiatives that allow learners to continue experimenting with coding and recycled-material prototypes, thereby fostering sustainable innovation cultures. Collectively, these recommendations emphasize the importance of aligning policy, pedagogy, and practice to create equitable and future-ready STEM learning environments.

While this study contributes valuable insights into the implementation of AI-supported PBL in low-resource African classrooms, several limitations must be acknowledged. First, the sample size (n = 10) and the geographic scope, limited to five countries, restrict the generalizability of the findings. Second, the study was conducted within the short-term context of a workshop, leaving the long-term effects on teaching practices and educators’ learning outcomes unexplored. Third, although qualitative data provided rich contextual detail, future research would benefit from incorporating quantitative measures to assess educator competence, confidence, and student achievement more systematically. To advance the field, future investigations should adopt longitudinal designs that examine the sustained integration of AI and CT practices across diverse educational settings. Expanding the sample to encompass a broader range of countries and educational levels would enhance external validity and provide a more comprehensive understanding of contextual variability. Moreover, adopting mixed-methods approaches in future research could generate deeper insights into the relationship between AI-supported PBL and measurable learning outcomes, thereby bridging the persistent gap between pedagogical innovation and empirical validation. The present findings contribute to extending the discourse on AI-supported PBL beyond high-resource educational settings by demonstrating its relevance and adaptability within low-resource contexts. In doing so, the study offers scalable and context-responsive models for advancing inclusive, technology-enhanced STEM education globally.