Research Article
A systematic review of educators' engagement with AI in problem-based learning
Synthesis: This PRISMA 2020 systematic review (50 peer-reviewed articles, 2015–2024) examines how educators engage with AI-driven Problem-Based Learning (PBL) tools through a Human–Computer Interaction (HCI) lens, aligned with SDG 4 (Quality Education). Three dominant themes emerge: AI-enhanced PBL environments strengthen educator–student collaboration via real-time feedback, intelligent scaffolding, and data-informed decisions; AI supports adaptive and personalized learning that improves engagement and problem-solving; yet persistent ethical concerns (data privacy, algorithmic bias, educator autonomy) and underdeveloped AI-based assessment practices limit adoption. The authors propose an AI-Enhanced Academic Interaction Model (AEAIM) integrating the Technology Acceptance Model, Constructivist Learning Theory, and Krashen's Input Hypothesis, and stress that AI must empower rather than replace educator agency.
Key Findings
- AI-driven PBL environments enhance educator and student collaboration through real-time feedback, intelligent scaffolding, and data-informed instructional decisions.
- AI tools support adaptive and personalized learning experiences that improve learner engagement and problem-solving skills.
- Persistent ethical concerns — data privacy, algorithmic bias, and erosion of educator autonomy — accompany AI integration and require human oversight.
- AI-based assessment within PBL remains underdeveloped: only ~10% of reviewed studies examined AI-driven assessments, indicating weak empirical evidence on automated feedback, formative assessment, and learning analytics.
- Only 45% of educators report confidence using AI-driven PBL platforms without additional training, highlighting a gap in AI usability support and literacy.
- Opaque AI models raise educators' cognitive load and reduce trust and adoption, underscoring the need for Explainable AI (XAI) and user-centered interfaces.
Study Design & Method
- Systematic literature review following PRISMA 2020 (Identification → Screening → Eligibility → Inclusion), starting from 3,500 records and yielding 50 peer-reviewed articles (2015–2024).
- Databases searched: Web of Science, Scopus, IEEE Xplore, SpringerLink, Taylor & Francis Online, Elsevier (ScienceDirect), Consensus, and Google Scholar, using Boolean (AND/OR) and truncation operators.
- Thematic analysis of the 50 studies using NVivo 14 with two independent coders (inter-rater reliability κ = 0.82).
- Results framed around an HCI lens (usability, explainability, cognitive load, decision autonomy) and three thematic codes: educators' perceptions/challenges in AI adoption, AI's effectiveness in supporting PBL, and AI's impact on learning outcomes and SDG 4 equity.
What this means for practice
- Instructors. Keep AI in an assistive role over problem-based learning: the review's model has AI supporting educator–student collaboration through real-time feedback, intelligent scaffolding, and data-informed decisions, with human oversight retained so pedagogical control stays with the educator.
- Instructors. Ask for usability and AI literacy training before the platform arrives, and read low confidence as a systems gap rather than a personal one: the review reports that only 45% of educators feel confident using AI-driven PBL platforms without additional training, and that opaque models raise cognitive load while depressing trust and adoption.
- Instructors. Press for explainable AI and transparent rules when your institution adopts a PBL platform, because data privacy, algorithmic bias, and the erosion of educator autonomy recur across the 50 reviewed studies.
- Instructors. Treat AI-driven assessment as unproven rather than settled: only about 10% of the reviewed studies examined AI-driven assessments, so automated feedback, formative assessment, and learning analytics in PBL have thin empirical support behind them.
- Instructors. Use the review's integrative adoption model (TAM plus constructivism and Krashen's Input Hypothesis) as a lens when evaluating a PBL tool, and raise the structural blockers — funding constraints, inconsistent AI policies, and staff resistance — with institutional leadership, since the review identifies these as what stalls sustainable adoption.
Limitations
- The synthesis rests on 50 articles distilled from 3,500 records through a PRISMA 2020 flow (2,500 screened, 900 full texts, 250 eligible, 50 included) and coded thematically by two independent coders (κ = 0.82); it is a narrative synthesis with no meta-analytic pooling, so no pooled effect sizes are reported.
- Only about 10% of the included studies examined AI-driven assessments, while AI in Education (30%) and Digital Learning Environments (20%) dominated the corpus, so the review's conclusions about automated feedback, formative assessment, and learning analytics rest on a small slice of the evidence base.
- The widely cited 45% educator-confidence figure comes from a single external study cited within the review rather than from the review's own coding of its 50 articles, so that claim is second-hand evidence.
- Studies focused solely on student engagement without educator perspectives were excluded, as were non-peer-reviewed sources, so the review reports nothing on learner-side outcomes and omits gray literature; full texts are said to have been assessed for "methodological rigor," but no validated quality-appraisal instrument is reported.
Citation
Amdan, M.A., Asim, A., Rosman, N.F., Janius, N., Dasuki, F.H., & Johnny, N.K. (2026). A systematic review of educators' engagement with AI in problem-based learning. Quantum Journal of Social Sciences and Humanities, 7(1).