AI Co-designer – HKUST’s Hybrid-Intelligence Course Designer for Evidence-Based Course Quality

HKUST’s AI Co-designer is a hybrid-intelligence course design platform that supports faculty in making thoughtful, pedagogically sound decisions when designing new courses or redesigning existing ones. Powered by Retrieval-Augmented Generation and informed by learning science and institutional policy data, the platform enables faculty to brainstorm, evaluate pedagogical options, and align course components with university requirements.

Designed for both individual and institutional use, the AI Co-designer functions as a professional development tool for teaching and learning centres, supports junior faculty through structured instructional design guidance, and facilitates quality assurance by generating intended learning outcomes and syllabus templates aligned with institutional standards.

The AI Co-designer was developed using Retrieval-Augmented Generation, allowing learning science, instructional design theory, and HKUST institutional policy data to directly inform its outputs. Grounded in Outcome-Based Education and the principle of Backward Design, the platform ensures that course design begins with clearly articulated learning outcomes and proceeds through aligned assessment and learning activities.

Faculty engage with the platform through a structured workflow aligned with established instructional design practices. The process begins with submission of basic course information, such as course descriptions or supporting files. Faculty are then guided to formulate Intended Learning Outcomes aligned with Bloom’s Taxonomy, with prompts encouraging clarity, appropriate cognitive levels, and measurability.

Based on these outcomes, the platform proposes aligned assessments and learning activities drawing on instructional models such as those of Gagné and Merrill. Faculty can select discipline-specific and pedagogical knowledge bases to tailor outputs to their contexts, supporting exploration of pedagogical options while remaining aligned with institutional requirements. Curated, editable sample prompts are provided at each stage to support efficient and meaningful use.

Throughout the workflow, built-in constructive alignment checks help faculty visualise coherence across outcomes, assessments, and activities, enabling early identification and correction of misalignment. The process concludes with the generation of detailed lesson plans grounded in instructional theory and a standardized syllabus consolidating all course components. Each AI-generated output is accompanied by reflection questions that prompt critical evaluation and refinement, reinforcing effective human–AI collaboration and maintaining faculty control over design decisions.

The platform also supports collaborative course design, allowing co-instructors and teaching assistants to work within a shared workspace. An analytics dashboard enables administrators to review usage patterns, ratings, and qualitative feedback, with AI-supported analysis providing insights for continuous improvement. The platform has been piloted and refined through three cycles of User Acceptance Testing, with iterative enhancements informed by extensive faculty feedback.

The preliminary impact of the HKUST AI Co-Designer was assessed through three cycles of User Acceptance Testing (UAT) involving faculty from all schools, the final cycle having recently concluded. The UAT survey utilized targeted questions as a measurement strategy, asking faculty to quantitatively rate the platform on four critical dimensions: 

1. its ability to produce pedagogical and tailored responses to their course contexts, 

2. the coherence of outputs based on the principle of constructive alignment throughout the course design process, 

3. perceived time savings as compared to a course design task without it, 

4. the overall effectiveness of the platform. 

These quantitative scores were meticulously analyzed together with open-ended comments from faculty to provide essential context regarding their experiences.       

Overall, favourable feedback from faculty demonstrated the clear positive impact of the platform. Users consistently found the platform’s outputs to be well-tailored to their specific course contexts and praised its capacity for ensuring clear alignment of course components, which significantly simplified the creation of a coherent course design. The overall effectiveness ratings of the platform were similarly positive, confirming that the structured design process fundamentally streamlined their workflow. The most compelling metric of the platform's success was a reduction in course planning time, with some faculty reporting savings of up to 50%. The collected evidence confirms the AI Co-Designer’s powerful role in supporting faculty efficiency and course quality.

• Learning-science-informed instructional design principles
• Retrieval-Augmented Generation architecture
• Outcome-Based Education and Backward Design frameworks
• Curated pedagogical and disciplinary knowledge bases
• Structured workflows with built-in constructive alignment checks
• Reflection prompts supporting human-AI collaboration
• Faculty and administrator feedback through iterative pilot testing
• Institutional policy and quality assurance alignment