A Systematic Literature Review of Artificial Intelligence-Enabled Coding Education: Humanizing Pedagogy, Learner Agency, and Equity in Global and Low-Resource Contexts

Abstract

Abstract Artificial intelligence (AI) is increasingly being incorporated into coding and programming education through conversational tutors, automated feedback systems, adaptive scaffolding, multimodal learning environments, and AI-assisted programming tools. While these technologies can provide immediate assistance and personalized support, their educational value depends on how they are designed and how learners interact with them. This systematic literature review examines AI-enabled coding education through three interconnected dimensions: humanizing pedagogy, learner agency, and equity in resource-constrained contexts. The empirical analysis (56 research papers) examines AI modes, Reflexive Thematic Analysis (RTA) themes, contextualization, Critical Discourse Analysis (CDA) narratives, pedagogical moves, educational level, prevalence, and causal mechanisms. The analysis extends to examining the potential for digital bricolage in Ghana and Nigeria, data and bandwidth constraints, linguistic and local marginalization, and associated CDA narratives. The results reveal a recurring distinction between two pedagogical orientations. The Socratic Interlocutor emphasizes questioning, formative intervention, logic critique, scaffolding, authentic probing, verification, and learner-directed activity. The Oracle, by contrast, emphasizes rapid summarization, adaptive assistance, efficiency, and automated support. These orientations are associated with different pedagogical and discursive narratives (IEEE 3301, 2025; Blasquez, 2025; Tan, 2024). The analysis demonstrates that the effectiveness of AI-enabled coding education is also shaped by infrastructure and local context. Several entries identify high or critical bandwidth requirements for real-time AI interaction, while others identify offline, low-bandwidth, localized, or small-language-model approaches as possible alternatives. The analysis also identifies English-language dominance, Western-centric programming conventions, and inadequate representation of local linguistic and cultural contexts as equity concerns. The paper argues that AI-enabled coding education is most educationally defensible when AI operates as a scaffold for human reasoning rather than a substitute for it. In addition, equitable implementation requires attention to bandwidth, localization, language, cultural relevance, and the ability of learners to adapt technologies to their own circumstances. Keywords: artificial intelligence, coding education, learner agency, Socratic AI, scaffolding, digital bricolage, educational equity, Critical Digital Pedagogy, generative AI, systematic literature review