Content Generation & Safeguarding

UALS uses AI to generate all learning content dynamically. This page covers how to pre-generate content caches, export content to various formats, and use the content safeguard system to ensure quality.

📦 Content Generation

The Content Generation section provides tools to pre-generate and export AI-powered learning content for your class.

📦

Generate Content Cache

Pre-generate content for all competencies in your class. This speeds up the student experience by having content ready before they request it.

Generates KE (Knowledge Explorer) maps
Generates SPL (Socratic Playground) content
Generates SBCAT assessment items
Creates multiple versions per competency

📄

Export Content

Export generated assessment items to standard formats for use in other systems or for offline review.

PDF format for printing/review
QTI format for LMS import
Blackboard format for Blackboard LMS

🎓

Generate Lecture Content

Create lecture slides and materials based on your class competencies. AI generates structured presentations for each topic.

PowerPoint-compatible slides
Speaker notes included
Aligned to competency levels

Content Cache Wizard

The Content Cache Wizard guides you through pre-generating content:

Step 1: Select Content Types

Choose which content types to generate: Knowledge Explorer maps, Socratic Playground content, and/or SBCAT assessment items.

Step 2: Select Competencies

Choose specific competencies or select all. You can generate content for the entire class or focus on specific topics.

Step 3: Configure Versions

Set how many versions to generate per competency. More versions provide variety but take longer to generate.

Step 4: Generate

Start the generation process. Progress is shown in real-time. Large classes may take several minutes to complete.

💡 Tip: Generate content caches during off-peak hours to minimize wait times and ensure all content is ready when students arrive.

🛡️ Safeguard AI Content

AI can make mistakes. The Content Safeguard system ensures every piece of AI-generated content is reviewed and verified by a human expert before students see it.

🛡️ Human-Verified Content

The Content Editor is Layer 3 of UALS's 6-layer content safeguard system. It provides a dedicated interface for reviewing, editing, and approving all AI-generated content.

Content Editor

The Content Editor allows you to:

View all cached content for your class
Edit any AI-generated text, scenarios, or questions
Mark content as reviewed/approved
Flag problematic content for regeneration
Add teacher notes for other content experts

Quality Review Checklist

When reviewing AI content, verify these key areas:

🎯

Factual Accuracy

Verify dates, names, formulas, and facts. AI can hallucinate incorrect information that sounds plausible.

📚

Pedagogical Fit

Ensure content matches the intended proficiency level. Acquire-level content should be simpler than Create-level.

⚖️

Cultural Sensitivity

Check for bias and ensure inclusive language. Scenarios should be appropriate for your student population.

✅

Assessment Accuracy

Verify that correct answers are actually correct. Check that distractors are plausible but definitively wrong.

The 6 Safeguard Layers

UALS employs a comprehensive 6-layer safeguard system:

Layer	Name	Description
1	Prompt Engineering	Carefully designed prompts that guide AI to produce accurate, appropriate content
2	AI Notes	Teacher-provided context in DSC competencies that guides content generation
3	Content Editor	Human review and editing of all AI-generated content (this page)
4	Version Control	Multiple versions with teacher selection of best content
5	Student Feedback	Students can flag problematic content for review
6	Analytics Review	Performance analytics identify questions that may be problematic

⚠️ Important

While UALS provides multiple safeguards, the ultimate responsibility for content accuracy rests with the teacher. Always review content before making it available to students, especially for high-stakes assessments.

💾 Cache Management

Understanding how content caching works helps you manage your class efficiently.

How Caching Works

Content is cached by a hash of the configuration (competency, level, context)
Up to 4 versions are maintained per configuration
When students request content, a random version is served
Cache is class-specific (isolated between classes)

Cache Storage

Content is stored in Google Cloud Storage with this structure:

uals-storage/
└── cache/
    └── class-{classId}/
        ├── ke/                    # Knowledge Explorer
        │   └── {hash}/
        │       ├── 1/content.json
        │       ├── 2/content.json
        │       ├── 3/content.json
        │       └── 4/content.json
        ├── spl/                   # Socratic Playground
        │   └── {hash}/[1-4]/
        └── sata/                  # SBCAT Assessment
            └── {hash}/[1-4]/

Cache Benefits

Speed: Cached content loads instantly vs. 5-10 second generation
Cost: 94-96% reduction in LLM API costs
Quality: Pre-reviewed content ensures accuracy
Consistency: Students get verified, teacher-approved content

📤 Export Formats

PDF Export

Generates a printable PDF with all assessment items. Useful for:

Paper-based backup assessments
Review by department heads
Accessibility accommodations
Archiving for accreditation

QTI (Question and Test Interoperability)

IMS QTI format for import into compatible Learning Management Systems:

Canvas
Moodle
D2L Brightspace
Other QTI-compatible systems

Blackboard Format

Native Blackboard format for direct import into Blackboard Learn:

Pools and tests
Preserves question metadata
Supports multiple choice and multiple answer

💡 Export Tip: Export content regularly as backup. Even if you primarily use UALS for delivery, having exports ensures you're not locked into a single platform.

Best Practices

✅ Content Generation Best Practices

Pre-generate content caches before the semester starts
Review all content in the Content Editor before students access it
Generate multiple versions for variety
Export content as backup before major system updates
Use AI Notes in your DSC to guide content generation

⚠️ Common Pitfalls

Don't skip content review - AI makes mistakes
Don't generate too many versions (4 is usually sufficient)
Don't edit the same content from multiple browser tabs
Remember to regenerate after significant DSC changes