Intelligent Tutoring System (ITS) Agents

UALS includes a comprehensive Intelligent Tutoring System with 100 specialized AI agents that adapt to each student's learning style and progress. This page explains how to configure ITS agents for your class.

🧠 What Are ITS Agents?

ITS Agents are AI-powered tutoring assistants that work behind the scenes to personalize the learning experience for each student. They analyze student behavior, performance, and preferences to make intelligent recommendations.

💡 How It Works

When a student uses UALS, the ITS agents analyze their interactions and generate personalized recommendations. These recommendations influence content generation, hint timing, difficulty adjustment, and more.

Agent Categories

UALS includes 100 agents organized into categories:

Category Agents Purpose
Knowledge Representation 3 Domain model, curriculum alignment, misconception detection
Learner Modeling 4 Student profiles, knowledge tracing, mastery estimation
Pedagogical Decision-Making 9 Teaching strategies, content sequencing, intervention timing
Tutoring Interaction 5 Dialogue, hints, scaffolding, feedback
Content Generation 4 Problems, examples, explanations, metacognitive support
Assessment & Analytics 2 Performance evaluation, learning data analysis
Extended Capabilities 27 Advanced learner support, social learning, multimodal
Micro-Behavior Monitoring 43 Detailed activity monitoring, pattern detection
Coordinator 1 Orchestrates all other agents

📊 Analysis Levels

The ITS Analysis Level setting controls how many agents are involved in analyzing each student. Higher levels provide deeper personalization but require more processing time.

Level Name Agents Time Use Case
1 Lightning 4 1-2s Large classes, basic personalization
2 Quick 11 2-4s Recommended for most classes
3 Standard 20 4-8s Content generation focus
4 Enhanced 35 8-15s Advanced pedagogical strategies
5 Professional 50 12-18s Multimodal & social intelligence
6 Research 65 15-22s Learning sciences research
7 Advanced 80 18-28s Domain-specific expertise
8 Enterprise 90 22-32s Longitudinal learning & AI ethics
9 Stakeholder 95 24-35s Multi-stakeholder intelligence
10 Complete 100 25-40s Full adaptive learning ecosystem
✅ Recommendation

Level 2 (Quick) is recommended for most classes. It provides good personalization with minimal wait time. Use higher levels for smaller classes or research contexts.

🎛️ ITS Agent Customizer

The ITS Agent Customizer allows you to fine-tune which agents are active and their behavior settings for your class.

Accessing the Customizer

  1. Open Class Management for your class
  2. Navigate to "ITS Agents" in the sidebar
  3. Click "Customize ITS Agents"

Available Settings

📊

Analysis Level

Set the default analysis depth (1-10). Higher levels engage more agents for deeper personalization.

🔄

Agent Activation

Enable or disable specific agent categories based on your teaching needs and class size.

⚙️

Behavior Tuning

Adjust individual agent parameters like hint frequency, intervention thresholds, and feedback styles.

🧠

AI Thinking Display

Control whether students can see the "Show AI Thinking" button that displays agent reasoning.

🔑 Key Agents Explained

Learner Model Agent

Builds and maintains a comprehensive profile of each student including:

  • Prior knowledge and skill levels
  • Learning pace and patterns
  • Preferred content types
  • Emotional and motivational states

Pedagogical Model Agent

Makes high-level teaching strategy decisions:

  • When to introduce new concepts vs. reinforce existing ones
  • Which teaching method suits the current learner state
  • Optimal sequence of topics
  • When to provide breaks or changes of pace

Socratic Tutoring Agent

Implements inquiry-based learning in the Socratic Playground:

  • Generates probing questions
  • Guides without giving direct answers
  • Challenges assumptions
  • Encourages self-discovery

Hint Generation Agent

Provides progressive hints during assessments:

  • Determines when a hint is needed
  • Chooses appropriate hint level (subtle to explicit)
  • Tracks hint usage for analytics
  • Adjusts based on student frustration level

Misconception Detection Agent

Identifies common student errors:

  • Recognizes patterns in wrong answers
  • Maps errors to known misconceptions
  • Triggers targeted remediation
  • Updates learner model with gap information

🔍 Explainable AI Feature

UALS includes a "Show AI Thinking" feature that provides transparency into how the ITS makes decisions.

How It Works

  1. Student completes the learner onboarding
  2. ITS agents analyze the student and generate recommendations
  3. "🧠 Show AI Thinking" button appears on results screen
  4. Clicking reveals a timeline of agent invocations and reasoning

Benefits

  • Transparency: Students understand why content is personalized
  • Trust: Seeing the reasoning builds confidence in the system
  • Learning: Students gain metacognitive awareness
  • Debugging: Teachers can verify recommendations make sense
💡 Note: The workflow recording expires after 5 minutes for privacy reasons (GDPR compliance). Students can regenerate recommendations to see fresh AI thinking.

📈 Impact on Learning Systems

The ITS recommendations directly influence all three learning systems:

Knowledge Explorer (KE)

  • Novice learners get simpler concept maps
  • Advanced learners get dense, interconnected networks
  • Gaps trigger targeted expansion nodes

Socratic Playground (SPL)

  • Socratic mode uses questioning techniques
  • Novices get more detailed hints
  • Frustration triggers encouragement and easier questions

SBCAT Assessments

  • Diagnostic focus on knowledge gaps
  • Difficulty calibrated to mastery level
  • Sophisticated distractors for advanced students

⚡ Performance Considerations

Operation Time
Single agent invocation 50-150ms
Parallel agent group (3 agents) 150-250ms
Sequential workflow (5 steps) 500-800ms
Complex workflow (10 steps) 1000-1800ms
⚠️ Class Size Considerations

For large classes (100+ students), consider using lower analysis levels (1-3) to minimize processing time. The system can handle 100+ concurrent learners per server with level 2 analysis.

Best Practices

✅ Recommended Practices
  • Start with Level 2 and adjust based on class feedback
  • Enable the "Show AI Thinking" feature for transparency
  • Monitor analytics to see if personalization improves outcomes
  • Consider higher levels for struggling students who need more support
⚠️ Common Pitfalls
  • Don't use Level 10 for large classes (processing time issues)
  • Don't disable all agents (some personalization is always beneficial)
  • Remember that agent recommendations are suggestions, not guarantees