Summary

Fulbright Planner is a web application built to address challenges faced by Fulbright grantees in Taiwan's educational settings. This tool uses a multi-phase generation approach to create context-specific lesson plans and workshop outlines. It is meant to help both novice and experienced teachers align their teaching with current educational standards recommended by Taiwan's Ministry of Education (MOE) and the Foundation for Scholarly Exchange (FSE) while incorporating pedagogical best practices. The app generates lessons which include differentiation strategies, multiple assessment approaches, and culturally responsive elements - saving educators hours of planning time as they can start planning from a good foundation and focus more time and energy on students. This article details the tool's development, implementation, and impact on the Fulbright Taiwan community.

Github Project Repository: ETA-lesson-planner
Live Application: Fulbright Lesson Planner

---

Tool Overview

The Challenge

Fulbright Taiwan deploys over 200 grantees annually across diverse educational settings—from urban centers to rural communities, elementary schools to universities. Despite shared titles and general expectations, each grantee faces unique classroom dynamics, co-teaching relationships, and student backgrounds. Traditional one-size-fits-all training struggles to address these differences, leaving many grantees to adapt general principles to specific contexts with limited guidance.

Key Problem: Standard training cannot account for the wide variance in teaching contexts experienced by Fulbright participants across Taiwan.

The Solution

Fulbright Planner used AI workflow prompting to create personalized educational resources through an easily to use web app. The tool generates detailed lesson plans and workshop outlines that:

• Align with both Fulbright program goals and Taiwan's MOE standards
• Accommodate specific grade levels, class durations, and teaching contexts
• Incorporate appropriate co-teaching models for Taiwan's bilingual education initiative

The application was developed iteratively with feedback from approximately 60 Fulbright participants, ensuring its relevance to real teaching scenarios encountered throughout Taiwan.

---

Features

Dual Planning Functionality

Fulbright Planner offers two specialized planning interfaces:

---

Progressive Multi-Phase Generation

Rather than producing plans in a single step, Fulbright Planner combines LLMs with experience using a multi-phase approach that mirrors how many experienced teachers plan:

Phase 1: Objectives Refinement

• Transforms initial objectives into structured, measurable learning objectives with clear and appropriate performance criteria
• Aligns with educational standards and backward design principles

Phase 2: Activities Design

• Creates engaging, culturally and age appropriate learning activities
• Incorporates differentiation strategies for mixed-ability classrooms

Phase 3: Components Preparation

• Generates supporting materials and assessment tools
• Develops scaffolding approaches for language learners

Phase 4: Final Plan Integration

• Assembles all elements into a comprehensive, ready-to-use plan with reflection questions for the teacher
• Formats output in the standard Fulbright lesson planning template

Design Philosophy: The multi-phase approach follows universal backward design principles—the same methodology many teachers use when planning effective lessons.

Taiwan-Specific Educational Elements

The tool incorporates several elements specific to Taiwan's educational landscape:

• Bilingual Education Elements: Supports Taiwan's Bilingual 2030 initiative with appropriate language scaffolding
• Co-Teaching Models: Integrates the six standard co-teaching approaches recognized by FSE
• Cultural Sensitivity: Includes appropriate cultural references and cross-cultural communication strategies
• Local Curriculum Standards: Aligns with Taiwan's curriculum guidelines and educational priorities

---

Technical Implementation

Architecture Overview

Fulbright Planner is built as a responsive single-page web application using:

• Frontend: SvelteKit and TailwindCSS for a clean, accessible interface
• Backend: Node.js and serverless functions deployed via Vercel
• AI: Anthropic's Claude models (primarily Claude-3.7-Sonnet and Claude-3.5-Haiku)
• Development: User-centered design with multiple testing phases and log analysis

Agentic AI Approach

The core of Fulbright Planner is its multi-phase generation system that employs structured prompting techniques to create a coherent planning workflow:

User Input → Phase 1 (Objectives) → Phase 2 (Activities) → Phase 3 (Components) → Phase 4 (Final Plan)

Each phase:

1. Receives the context from user inputs
2. Incorporates outputs from previous phases
3. Performs a specialized transformation task
4. Provides structured output for subsequent phases

This approach creates a chain of reasoning that is meant to follow the same process expert teachers use when planning while maintaining the users specific needs and context throughout the generation sequence.

Phase-Specific Prompt Construction

This snippet demonstrates how you can construct different prompts for each phase of the lesson planning process, showing the core of the agentic AI implementation:

// From /api/generate-emi-lesson/+server.js - Phase-specific prompt construction
switch (phaseName) {
  case 'Refining Objectives':
    prompt = `As an expert English as a Medium of Instruction (EMI) Advisor in Taiwan, you are part of a multi-step workshop planning workflow. Your role in Phase 1 is to refine raw learning objectives and provide supporting context for later phases. Improve on, but do not disregard important details in the INPUT OBJECTIVES. This is what the workshop facilitator wants to work on.

        INPUT OBJECTIVES:
        ${objectives}

        WORKSHOP CONTEXT:
        Topic: ${topic}
        Target Audience: ${audience}
        Workshop Duration: ${duration}
        Number of Participants: ${participantCount}
        ${workshopContext ? `Additional Context: ${workshopContext}` : ''}

        Your response should have two parts:

        PART 1: REFINED OBJECTIVES (2-3 total)
        Create 2-3 focused, achievable objectives based on the INPUT OBJECTIVES that are:
        - Specific and measurable using clear action verbs
        - Appropriate for the target audience and achievable in the given time
        - Written to allow for different participation styles
        - Structured as: "Participants will be able to [action verb] [specific skill/knowledge] [conditions] [criteria]"

        PART 2: PLANNING CONSIDERATIONS
        Provide key information for subsequent planning phases about:
        a) Differentiation Opportunities
        - How these objectives can be modified for different proficiency levels
        - Suggested scaffolding approaches
        
        // Additional prompt instructions...`;
    
    model = 'claude-3-5-haiku-20241022';
    max_tokens = 1000;
    break;
    
  // Additional cases for other phases...

Technical Note: Phase-specific prompts range from 500-1000 tokens, while outputs range from 1000-4000 tokens depending on the complexity of the planning task.

Error Handling and Edge Cases

The multi-phase architecture provides built-in error management:

• Each phase validates and refines outputs from previous phases
• The progressive generation approach prevents catastrophic failure by compartmentalizing generation tasks
• User feedback is displayed during generation through a visual progress indicator
• Inappropriate requests are redirected toward educational purposes through the structured prompt design

---

Installation & Deployment

Current Deployment

Fulbright Planner is deployed as a web application accessible to all Fulbright Taiwan participants through:

• Hosted on Vercel
• Shared through the Fulbright Taiwan community forum
• No login required to maximize accessibility

Technical Requirements

For those wishing to deploy or modify their own version:

---

Usage Examples

User Feedback: "I loved your lesson planning AI and shared it with my LET from last year! [...] Thanks again for your model, it's a really cool tool! I used it to get out of a workshop planning rut earlier last week." — EMI in Kaohsiung

Use Case: Taipei ETA Elementary Middle School Ramadan Lesson
A Fulbright ETA working on a culturally responsive lesson about Ramadan for 7th graders with varying levels of English proficiency:

1. Input: The ETA provided essential parameters including the topic (Ramadan), grade level (7th), lesson duration (45 minutes), and selected the "one teach, one assist" co-teaching model for a class of 10-15 students
2. Process: The planning system generated refined objectives focusing on cultural understanding, vocabulary development, and critical thinking about discipline and reflection, then created differentiated activities with appropriate scaffolding for middle school learners with diverse language abilities
3. Output Summary:

• Warm-up using a world map and sticky notes to activate prior knowledge
• Introduction with bilingual vocabulary support and visual aids
• Main activities including a comparative graphic organizer and reflective pair-share discussions
• Assessment components with exit tickets and peer review
• Clearly defined roles for both the lead teacher (cultural context delivery) and the ETA (language support and facilitation)
• Differentiation strategies for varying language proficiency levels
• Cultural sensitivity considerations to ensure respectful engagement

Lesson Output Example:
Bilingual Lesson on home

---

Impact & Future Directions

Observed Impact

Since its deployment, Fulbright Planner has demonstrated several positive effects:

Future Development Plans

Based on user feedback and observed usage patterns, several enhancements are planned:

1. Chinese Language Interface: Creating a bilingual version to better support Taiwanese co-teachers
2. School-Specific Templates: Developing customized versions for specific educational institutions
3. Expanded Content Areas: Adding specialized support for STEM, arts, and physical education contexts

---

Conclusion

Fulbright Planner demonstrates how targeted agentic AI applications can address specific challenges in specialized domains. By combining educational expertise, cultural context, and advanced AI capabilities, the tool provides personalized support that traditional training approaches cannot deliver at scale.

Key Insight: This project exemplifies how thoughtfully designed AI systems can be informed by human experience and enhance human capabilities rather than replace them, allowing educators to focus more on classroom relationships and less on administrative planning.

Acknowledgments

Special thanks to the Fulbright Taiwan community for their invaluable feedback and testing support, and to the Foundation for Scholarly Exchange for fostering the collaborative environment that made this project possible.