Part 1: Introduction & System Overview (15 minutes)
0:00 - 0:15
Welcome & Icebreaker (3 min)
- Quick introductions (name, role, institution)
- Poll: "What's your biggest adaptive learning challenge?"
The Adaptive Learning Problem (5 min)
Three critical challenges:
- Philosophical rigidity: Systems designed for single pedagogy
- AI opacity: Black-box decisions without transparency
- ITS fragmentation: Narrow focus (only cognitive OR only assessment)
Real-world impact:
- Students transfer institutions → learning history lost
- Teachers can't validate AI recommendations
- Systems can't adapt to diverse learner needs
UALS Solution (7 min)
Four innovations:
- Multi-Philosophy Framework:
- Curriculum-based (domain → subdomain → concept)
- Competency-based (category → competency → proficiency)
- Identical UI, different content generation
- Three Complementary Modes:
- Knowledge Explorer (exploration)
- Socratic Playground (inquiry)
- Scenario-Based CAT (assessment)
- 100-Agent ITS:
- 30 production-ready, 70 extended
- 17 categories (learner modeling, pedagogy, ethics, neuroscience, etc.)
- 10 coordination levels (4-100 agents, 1-40s)
- Explainable AI:
- "Show AI Thinking" workflow visualization
- Complete transparency for trust-building
- 5-minute cache for immediate replay
View Dashboard
Part 2: Knowledge Explorer (KE) - Theory + Hands-On (25 minutes)
0:15 - 0:40
Theory & Design Principles (7 min)
Pedagogical Foundation:
- Constructivist learning (Piaget, Bruner)
- Cognitive load theory (Sweller)
- Schema construction through exploration
Design Features:
- Interactive concept mapping
- Progressive disclosure (prerequisite-based unlocking)
- Multiple representations (definition, visual, examples, practice)
- Learner-driven navigation (not forced sequence)
Agent Coordination (Level 2, 11 agents):
- Learner Model → Retrieves knowledge state
- Mastery Estimation → Identifies gaps/strengths
- Pedagogical Model → Selects exploration depth
- Adaptive Sequencing → Recommends next concept
- Affective State Monitoring → Assesses confidence
- Knowledge Tracing → Predicts trajectory
- Intervention Timing → Decides if scaffolding needed
- Domain Model → Provides concept relationships
- Misconception Detection → Identifies errors
- Explanation Generation → Adapts complexity
- Feedback Generation → Creates encouragement
Result: Personalized KE content with adaptive scaffolding
Hands-On Activity (15 min)
Instructions:
- Navigate to curriculum mode: Science → Biology → Photosynthesis
- Explore 3-4 concepts in depth
- Complete practice questions (note how hints adapt)
- Try to unlock an advanced topic by mastering prerequisites
- Switch to competency mode: Data Literacy → Statistical Analysis
- Compare navigation and structure
Observation Prompts:
- How does content complexity change as you master concepts?
- What happens when you get practice questions wrong?
- How is prerequisite structure different in competency mode?
Try Knowledge Explorer
Try Competency Mode
Group Discussion (3 min)
- When would you use KE with your learners?
- What subjects are ideal for KE?
- How could KE support diverse learning styles?
Part 3: Socratic Playground (SPL) - Theory + Hands-On (30 minutes)
0:40 - 1:10
Theory & Design Principles (8 min)
Pedagogical Foundation:
- Socratic method (Collins & Stevens, 1982)
- EMT dialogue (Marineau et al., 2019)
- Productive failure (Kapur, 2008)
- Zone of Proximal Development (Vygotsky, 1978)
Design Features:
- Authentic, scenario-based problems
- Five hint agent types:
- Conceptual: "What principle applies?"
- Procedural: "What's the next step?"
- Reflection: "Why didn't that work?"
- Worked Example: Complete solution
- Error Analysis: Systematic mistake identification
- Four interaction modes:
- Socratic (questioning > telling)
- Demonstrative (worked examples)
- Practice-focused (repeated application)
- Study Mate (peer-like conversation)
Agent Coordination (Level 3, 20 agents):
- All 11 from KE PLUS:
- Problem Generation, Worked Example Generation
- Explanation Generation, Metacognitive Support
- Domain Model, Curriculum Alignment
- Misconception Detection, Assessment
- Performance Analytics
EMT Dialogue Framework:
- Elicit expectation ("What do you predict?")
- Detect misconception (learner attempts problem)
- Provide tailored feedback (address specific error)
Hands-On Activity (18 min)
Instructions:
- From photosynthesis KE, click "Practice in SPL"
- Attempt Problem 1 (plant health scenario):
- Try solving WITHOUT hints first
- If stuck, request conceptual hint
- If still stuck, request procedural hint
- As last resort, view worked example
- Attempt Problem 2 (light intensity effects):
- Use "Study Mate" mode for conversational support
- Notice how difficulty adapts to your performance
- Attempt Problem 3 (advanced):
- Try Socratic mode (minimal scaffolding)
- Observe EMT dialogue when you make a mistake
Observation Prompts:
- How do hints guide without giving answers?
- What happens when you make a systematic error?
- How does Study Mate mode differ from regular hints?
- How does difficulty change based on success/failure?
Try Socratic Playground
Group Discussion (4 min)
- How is SPL different from traditional problem sets?
- When would you use demonstrative vs. Socratic mode?
- How could SPL support struggling learners?
☕ BREAK (10 minutes) - 1:10 to 1:20
Part 4: Scenario-Based CAT (SBCAT) - Theory + Hands-On (20 minutes)
1:20 - 1:40
Theory & Design Principles (6 min)
Pedagogical Foundation:
- Item Response Theory (Lord, 1980)
- Computerized Adaptive Testing (Wainer, 2000)
- ITS-inspired assessment (Hu et al., 2023)
Design Features:
- IRT 2PL model: P(correct) = 1 / (1 + e^(-a(θ - b)))
- θ = learner ability (estimated in real-time)
- b = item difficulty
- a = item discrimination
- Adaptive algorithm:
- Start at θ = 0 (or prior from learner model)
- Select item with b ≈ θ ± 0.3 (maximum information)
- Update θ after each response
- Terminate when SE(θ) < 0.3
- Scenario-based items (not multiple choice trivia)
- Diagnostic feedback (not just correctness)
- Formative guidance (hints available during test)
Key Innovation: Unlike traditional CAT (no feedback), SBCAT makes assessment a learning opportunity
Hands-On Activity (12 min)
Instructions:
- From SPL, click "Assess Mastery with SBCAT"
- Complete adaptive assessment (8-12 items)
- Observe:
- How difficulty adjusts after each response
- Feedback provided (not just "incorrect")
- Option to request hints (with small θ penalty)
- Review final diagnostic report:
- Final ability estimate (θ with confidence interval)
- Mastery level (novice/intermediate/advanced)
- Strengths identified
- Gaps for remediation
- Recommended next steps
Observation Prompts:
- How does question difficulty change?
- What feedback do you receive after wrong answers?
- What insights does the diagnostic report provide?
- How could you use this report to guide instruction?
Try Adaptive Assessment
Group Discussion (2 min)
- When would you use SBCAT: pre-assessment, formative, summative?
- How is SBCAT different from traditional quizzes?
- What are ethical considerations of adaptive assessment?
Part 5: The 100-Agent Architecture + "Show AI Thinking" (20 minutes)
1:40 - 2:00
Agent Overview (5 min)
17 Agent Categories (100 total):
- Core Learner Modeling (4)
- Tutoring Interaction (11)
- Content Generation (17)
- Pedagogical Strategies (15)
- Assessment (11)
- Cognitive Support (10)
- Social-Emotional Learning (9)
- Social & Collaborative (6)
- Domain-Specific (6)
- Accessibility & Inclusion (3)
- Teacher Support (4)
- Analytics & Research (7)
- AI Ethics & Fairness (4)
- Longitudinal Learning (4)
- Real-World Learning (4)
- Parent & Community (3)
- System Intelligence (2)
10 Coordination Levels:
| Level |
Name |
Agents |
Time |
Use Case |
| 1 |
Lightning |
4 |
1-2s |
Quick checks |
| 2 |
Quick |
11 |
2-4s |
Recommended default |
| 3 |
Standard |
20 |
4-8s |
Comprehensive tutoring |
| 5 |
Professional |
50 |
12-18s |
Advanced analysis |
| 10 |
Complete |
100 |
25-40s |
Full ecosystem |
"Show AI Thinking" Demo (12 min)
Activity Instructions:
- Click "Get AI Recommendation" from dashboard
- Enter goal: "Understand photosynthesis"
- Try Level 2 (Quick):
- 11 agents, 2-4 seconds
- Recommends SPL with rationale
- Click "🧠Show AI Thinking"
- Explore workflow timeline
- Play/pause, adjust speed
- Click agents to see reasoning
- Try Level 5 (Professional):
- 50 agents, 12-18 seconds
- More comprehensive analysis
- Includes multimodal, social intelligence
- Compare workflow to Level 2
- Try Level 10 (Complete) (if time permits):
- 100 agents, 25-40 seconds
- Complete ecosystem consultation
- Includes neuroscience, ethics, longitudinal
Key Insights:
- Transparency builds trust in AI
- Different levels for different use cases
- Educators can validate recommendations
Try AI Recommendation
Group Discussion (3 min)
- How does explainability help educators?
- Which analysis level would you use most?
- What research questions could workflow data answer?
Part 6: Wrap-Up, Q&A, Next Steps (15 minutes)
2:00 - 2:15
Q&A Session (10 min)
Common Questions (be prepared):
- "What if students game the system?"
→ Agent monitoring detects patterns; system adapts
- "How much do LLMs cost?"
→ <$0.01/student/hour (94% cache hit rate)
- "What about incorrect content?"
→ Validation layers + educator review dashboard
- "Can I customize content?"
→ Yes, custom prompts + domain model import
- "How long to implement?"
→ Pilot: 2-4 weeks; Full deployment: 2-3 months
- "LMS integration?"
→ LTI 1.3 (Canvas, Moodle, Blackboard, D2L)
Next Steps (5 min)
Takeaways:
- Workshop materials: Slides, recording (if virtual)
- Demo access: 30-day trial credentials
- Documentation: Architecture, API, deployment guides
- Implementation checklist: Steps for pilot program
Follow-Up Options:
- 3-hour workshop: Deep dive into agents, analytics, customization
- Office hours: Monthly Q&A sessions
- Pilot program: 3-month institutional trial
- Community: Join UALS educator Discord/Slack
Call to Action:
"Ready to transform adaptive learning at your institution? Schedule a 30-minute consultation to design your pilot program."
Contact:
- Email: pilot@uals.edu
- Office hours: First Friday/month, 10 AM PT
- Community: discord.gg/uals-educators