Co-designing Large Language Model Tools for Project-Based Learning with K12 Educators

Research Background and Issues

What problems or challenges did the authors identify?

The authors explored practical challenges in implementing project-based learning (PBL) in K-12 education, including:

• The complexity of designing standards-aligned projects to meet diverse student needs.
• The difficulty of managing planning time and classroom dynamics, particularly balancing student autonomy and teacher guidance.
• Challenges in assessing student learning outcomes, such as teamwork, creativity, and problem-solving skills, which traditional assessments struggle to capture accurately.
• Students' difficulties in adapting to PBL from traditional learning, such as frustration caused by task ambiguity.
• The burden on teachers to provide personalized feedback and manage interdisciplinary projects.

Additionally, teachers implementing PBL must transition from the traditional role of knowledge transmitters to learning facilitators, requiring complex professional development (PD) and new instructional strategies.

Why is this issue important?

PBL is widely regarded as an effective teaching method for fostering deep learning, interdisciplinary learning, and critical thinking in students. However, these challenges have limited its widespread implementation in K-12 schools. Supporting teachers in overcoming these difficulties is crucial for educational innovation.

Research Motivation and Related Work

• Generative Artificial Intelligence (GenAI), particularly large language models (LLMs), is considered a promising tool for addressing the complexities of PBL, offering personalized learning and real-time feedback capabilities to support differentiated instruction and reduce teacher workload.
• Despite the significant potential of LLMs in education, their impact on actual classrooms needs to be validated through teacher-driven design and research.
• This study explores the application of LLMs in PBL teaching within a co-design framework involving teachers, marking a novel contribution to existing literature, especially in the context of K-12 education.

Solutions

What methods or solutions did the authors propose?

• The authors adopted a co-design approach, involving teachers in tool design to ensure LLMs align with actual classroom needs.
• Proposed the development of LLM tools to support PBL, focusing on:
  1. Curriculum Support: Assisting teachers in creating and refining project ideas and lesson plans, providing standards-based templates and references.
  2. Assessment and Grading Support: Automated assessment and personalized feedback features to help teachers create differentiated evaluation criteria and grading mechanisms.
  3. Progress Tracking and Dynamic Management: Visualized student and team progress tracking features to assist teachers in managing contributions and tasks across multiple student groups.

What are the innovative aspects of this solution?

1. By adopting a teacher co-design approach, the tool design is more closely aligned with real teaching contexts, reflecting teachers' needs and values.
2. Focuses on addressing specific challenges in PBL, such as curriculum design, grading, and classroom management, rather than providing generalized educational support.
3. Emphasizes enhancing the teacher's role rather than replacing it, achieving a balance between automation and professional development.

What are the implementation steps? What key technologies were used?

1. Research Design:
   • Two rounds of teacher involvement: in-depth interviews with experienced PBL teachers and multi-stage co-design; followed by testing and feedback from teachers with diverse experience levels.
2. Co-Design Workshops:
   • Phase 1: Brainstorming and scenario analysis to identify teaching pain points and potential applications of LLMs.
   • Phase 2: Developing conceptual design prototypes based on discussions and exploring their impact on stakeholders such as teachers and students.
3. Prototype Design and Evaluation:
   • Prototypes included curriculum planning support tools, adjustable assessment and feedback features, and visualized student progress tracking.
   • Teachers tested these prototypes and provided feedback to optimize tool functionality and interaction design.
4. Technical Support:
   • Using LLMs to generate suggestions, create interactive templates, and provide visualized feedback.
   • Integration with existing learning management systems (LMS) such as Canvas and Google Classroom.

Research Outcomes

What specific outcomes were achieved?

• Needs and Challenges: The study clearly defined teachers' needs and pain points in implementing PBL, such as high workload management, balancing diverse needs, and difficulties in fair grading.
• Tool Prototypes: Developed three core supports:
  1. Project design tools: Helping teachers generate diverse project ideas and classroom activity suggestions.
  2. Differentiated assessment tools: Supporting the creation of personalized evaluation criteria and instant feedback.
  3. Progress management dashboards: Providing real-time monitoring and tracking of student contributions.
• Design Guidelines:
  • Flexible integration of teacher-driven inputs and personalized options.
  • Alignment of technology with curriculum standards to simplify complex lesson planning.
  • Emphasis on privacy and data protection, particularly when using sensitive student data for differentiated teaching.

How does this compare to existing solutions? What are its advantages?

• Compared to general educational AI tools, this solution specifically addresses PBL challenges such as collaborative dynamics management and complex grading assessments.
• Emphasizes teachers' creativity and autonomy while respecting existing teaching models.
• Ensures that the developed LLM tools meet the practical needs of teachers and students through a co-development approach.

What were the experimental or evaluation results?

• Testing results indicated:
  • Teachers highly valued LLMs' potential in project generation and personalized student support, noting significant reductions in lesson planning time and increased student learning independence.
  • Teachers expressed a clear preference for retaining human oversight in grading processes to ensure fairness and flexibility.
• Feedback revealed several directions for tool optimization, such as enhancing support for interdisciplinary collaboration, simplifying teacher training, and improving student data privacy protection.

Limitations and Future Directions

• Limitations:
  • The sample was concentrated within the U.S. K-12 education system, lacking validation in global educational contexts.
  • Participants' preference for LLM tools may underestimate resistance during actual implementation.
  • Prototype tools have not yet been deployed in real teaching environments, requiring further validation of their outputs.
• Future Directions:
  • Develop fully functional tools and conduct multiple rounds of testing on teacher and student interfaces.
  • Expand research to international and resource-constrained schools to evaluate tool scalability.
  • Explore co-designing tools from the student perspective to create a balanced support ecosystem for both teachers and students.

By combining carefully designed LLM tools with teacher-led co-design processes, this study demonstrates the potential of generative AI in PBL implementation. These tools aim to reduce teachers' instructional burdens while maintaining high-quality, personalized, and equitable teaching practices.