AIEd Knowledge BaseAIEd KB
On this page

Research Article

Artificial intelligence in interdisciplinary higher education: A systematic review on opportunities, challenges and future directions

added 2026-09-14 · updated 2026-09-19

Synthesis: Xia, Zhang, Xing, Andic, and Chiu (2026) report a PRISMA-guided systematic review of 59 studies examining how generative AI and other AI technologies are reshaping interdisciplinary education in higher education. Across four research questions the review maps six forms of interdisciplinary practice, three dominant research focuses, four functional AI roles — evaluator, agent, monitor and assistant — and stakeholder-level impacts on cognition, behavior and emotion. Its central contribution is an interdisciplinary human-AI interactive learning model organized around boundary mediation and knowledge integration, role and capability distribution, ethical and equity governance, and teacher agency and sustainable design. The review finds AI most often positioned as an assistant (43 of 59 studies) and an evaluator (22), yet warns that much interdisciplinary design lacks depth and contextual grounding, that AI introduces accuracy, bias, Privacy and overreliance risks, and that human-centered curricular models struggle with scalability and generalizability. It concludes that integration must be institutionally guided rather than technology-driven, foregrounding professional development, equitable access and alignment with Sustainability and global citizenship goals.

Key Findings

  • PRISMA screening yielded 59 included studies: From 2,950 records retrieved across Web of Science, Scopus, ERIC, ProQuest and PsycINFO in January 2025, 666 duplicates were removed, 2,284 titles and abstracts were screened, 308 full texts were assessed and 249 excluded, leaving 59 studies; coding reached 0.85 inter-coder reliability after a 20-article pilot.
  • Six forms of interdisciplinary education were identified: STEM dominated (n = 26), followed by cross-disciplinary student backgrounds (n = 14), non-STEM fields (n = 10), cross-disciplinary learning content (n = 10), inter-STEM (n = 8) and STEAM (n = 2), spanning domains from engineering and biology to journalism, design and theater.
  • Three research focuses structured the literature: AI-empowered curriculum and teaching transformation was the largest (n = 25), followed by AI-based learning analysis and educational evaluation (n = 20) and AI tool and system development (n = 14).
  • AI played four functional roles: It acted as an assistant in 43 studies, an evaluator in 22, an agent in 14 and a monitor in 9, across four AI forms — general AI such as ChatGPT, educational AI, technical frameworks and AI-powered robots.
  • Stakeholder impacts spanned cognition, behavior and emotion: Students were the most-studied group (n = 57; cognition 47, behavior 43, affect 14), teachers next (n = 26; behavior 24, cognition 18, affect 7) and staff least (n = 4), revealing a strong learner-centric bias in the field.
  • Three challenges limit the field: Interdisciplinary design often lacks depth and contextual grounding (short interventions, small samples); AI integration brings technological, ethical and learner-related risks (accuracy, bias, Privacy, overreliance and immature assessment instruments); and human-centered curricular models face scalability and generalizability constraints.
  • A four-dimension model was proposed: The interdisciplinary human-AI interactive learning model combines boundary mediation and knowledge integration, role and capability distribution, ethical and equity governance, and teacher agency and sustainable design.
  • Equity and governance gaps persist: Most included studies came from technologically advanced regions with limited representation of under-resourced contexts, prompting calls for institutional policies on Ethics, bias, transparency, student data privacy and equitable access to AI tools.

Review Method & Framework

  • PRISMA 2020 systematic review of 59 studies, guided by four research questions covering the forms of interdisciplinary education, main research focuses, AI's functional roles and effects, and impacts on different stakeholders.
  • Hybrid quality assessment: Because the corpus mixed quantitative, qualitative and mixed methods studies, a hybrid framework applied six criteria scored 0–3 each; total scores ranged from 8 to 15 and no study was excluded on quality grounds.
  • Collaborative coding with verified reliability: The first three authors (two independent coders and one moderator) piloted a coding framework on 20 articles, then coded all studies by region, interdisciplinary form, AI form, AI role, research focus, impacted object and impacted dimension, achieving 0.85 inter-coder reliability.
  • A deliberately broad definition of interdisciplinary: Following Moran (2010), the review treats interdisciplinary as dialogue or interaction among two or more disciplines, acknowledging that multi-, inter-, cross- and transdisciplinary boundaries frequently blur in educational practice.
  • A descriptive framework rather than a meta-analysis: The review maps the field across six interdisciplinary forms, three research focuses and four AI roles (evaluator, agent, monitor, assistant), then synthesizes these into the interdisciplinary human-AI interactive learning model.

What this means for practice

  • Instructors. Design AI-supported activities that mediate disciplinary boundaries rather than substitute for human thinking, keeping AI in the assistant role it played in 43 of the 59 reviewed studies and promoting interdisciplinary Problem Solving.
  • Administrators. Establish policies on AI ethics, bias, transparency and student data privacy before scaling AI across programs, so integration is institutionally guided rather than technology-driven.
  • Administrators. Fund sustained professional development and target access gaps explicitly, since the evidence base comes mostly from technologically advanced regions and educators need to move from passive users to active co-designers.
  • Instructors. Connect AI-supported interdisciplinary work to sustainability and global citizenship goals, embedding ethical reflection alongside technical training.
  • Researchers. Adopt longer interventions, larger samples and validated assessment instruments and report effect sizes, given the field's reliance on short, context-specific designs.

Limitations

  • The search excluded gray literature and was conducted only once, so recent studies in a rapidly evolving field may have been missed.
  • The review used no meta-analytic procedures and reported no effect sizes, limiting the practical applicability of its findings.
  • Generalizability is constrained by geographic concentration: most included studies came from technologically advanced regions with limited representation of under-resourced educational contexts.
  • The corpus is strongly learner-centric — students were the impacted group in 57 studies, teachers in 26 and staff in only 4 — so institutional and staff-level effects are thinly evidenced.

Connected Concepts

Connected Articles

Citation

Xia, Q., Zhang, Z., Xing, T., Andic, B., & Chiu, T. K. F. (2026). Artificial intelligence in interdisciplinary higher education: A systematic review on opportunities, challenges and future directions. Australasian Journal of Educational Technology, 42(3), 1–23.

Metadata

DisciplineStem Education
LevelHigher Ed
Intended audienceResearchers, Instructors, Policymakers
Research methods and evaluationMeta-Analysis and Systematic Review
Equity, ethics, and responsible useEquity, Ethics
Foundations of AI in educationAI in Education, Human AI Collaboration
Page kindFramework, Synthesis
SourceHistoryEditChat

Embed this page

Copy the code below to embed a chromeless version of this page in a learning management system or other website. The embedded view hides the site header, navigation, and footer.