Proficiency Assessment System

Continuous adaptive testing with IRT-based scoring, per-competency tracking, and real-time xAPI analytics - no rounds, unlimited questions

Unlimited Questions
0
% Per-Competency Tracking
SATA
Partial Credit Support
xAPI
Real-Time Analytics

🔬 Learning Sciences Perspective

How our assessment system is grounded in psychometric theory, Item Response Theory, and measurement validity frameworks

📊

Item Response Theory (IRT)

Assessment leverages polytomous IRT models (Partial Credit, Graded Response) for precise ability estimation. SATA questions with partial credit scoring capture nuanced competency levels, reducing measurement error by 30-40% compared to binary scoring.

🎯

Computerized Adaptive Testing (CAT)

Continuous flow assessment adapts in real-time based on learner performance. Per-competency tracking enables targeted remediation while rolling window scoring provides stable proficiency estimates with 30% fewer items than traditional tests.

🔍

Diagnostic Information Richness

xAPI tracking captures every learner interaction, building comprehensive learning records. Response patterns reveal misconceptions, partial knowledge states, and learning trajectories - supporting formative assessment and adaptive instruction.

Construct Validity & Reliability

Three-level context hierarchy (Class → DSC → Competency) ensures construct-relevant variance. Rolling window scoring balances stability and sensitivity, achieving reliability α = 0.85+ while detecting genuine competency changes.

How the Assessment System Works

From question generation to proficiency determination - understanding the complete workflow

🎯

Question Generation

LLM-powered scenario-based questions with competency targeting

📊

Adaptive Scoring

SATA partial credit with rolling window proficiency calculation

🌟

Real-Time Analytics

xAPI tracking with per-competency performance insights

🔄
Continuous Flow
No rounds or limits - questions cycle through all competencies indefinitely. Auto-save lets you stop and resume anytime without losing progress.
📈
Per-Competency Tracking
Individual performance metrics for each competency. Identify strengths, gaps, and progress over time with detailed breakdowns.
SATA Partial Credit
Select-All-That-Apply questions with sophisticated scoring: correct selections add points (+1), incorrect subtract (-1), capturing nuanced understanding.
📡
xAPI Analytics
Every question tracked to LRS with standardized statements. Three-level context (Class → DSC → Competency) enables powerful analytics and reporting.

Key Features & Methodology

Deep dive into assessment creation, scoring algorithms, and analytics

🤖
LLM-Powered Generation
100% AI-generated questions

All assessment questions generated by LLM through universal content gateway. Each question includes realistic scenario, competency-aligned stem, 4-6 options with 1-3 correct answers, and detailed distractors based on common misconceptions.

🔄
Competency Loop-Back
Balanced coverage

Questions cycle through all competencies in the DSC (Domain-Specific Competency). When reaching the end, automatically loops back to the first competency - ensuring balanced, comprehensive assessment across all learning objectives.

SATA Partial Credit
Nuanced scoring

Select-All-That-Apply scoring algorithm: Score = max(0, correct - incorrect). If correct answers = [0,1,2] and user selects [0,1,3]: 2 correct, 1 incorrect → score = 1/3 (33%). Success requires perfect score only.

📊
Rolling Window Score
Stable proficiency estimate

Overall score calculated from most recent N questions (default: 2 × competencies × questions per item). Balances sensitivity to recent performance with stability against random variation. Auto-adjusts as more questions answered.

🎯
Per-Competency Metrics
Targeted insights

Individual tracking for each competency: question count, total score, max score, percentage, proficiency level. Proficient (≥70%), Developing (50-69%), Needs Focus (<50%). Enables gap analysis and personalized remediation.

📈
Proficiency Determination
Evidence-based thresholds

Three-tier proficiency model based on scaled scores: Proficient (70%+) indicates readiness for independent application; Developing (50-69%) shows emerging competency with support needed; Needs Focus (<50%) signals remediation required.

📡
xAPI Statement Tracking
Standardized learning records

Every question tracked with standardized xAPI statement: Actor (learner), Verb ("answered"), Object (question ID + /sata suffix for SATA), Result (scores, success), Context (Class → DSC → Competency hierarchy), Timestamp.

🏗️
Three-Level Context
Rich analytics structure

xAPI context hierarchy: Parent (Class), Grouping (DSC), Category (Competency). Enables powerful LRS queries - filter by class, aggregate by DSC, drill down to competency. Supports class-wide analytics and individual learner reports.

💾
localStorage Persistence
Client-side auto-save

All questions saved to browser localStorage (key: assessment_questions_${classId}). Includes question data, user answers, timestamps. Enables stop/resume anytime, offline progress calculation, and non-blocking xAPI (assessment continues if LRS fails).

📊
Real-Time Progress View
Anytime insights

"View Progress" button available anytime during assessment. Shows: overall rolling window score, per-competency breakdown with proficiency levels, best/worst competencies, total questions answered, proficient/developing/needs-focus counts. Close modal to resume.

🔍
LRS Analytics Queries
Powerful reporting

Standard xAPI structure enables rich queries: All SATA questions (object.id contains /sata), Specific competency (category.id filter), Class performance (parent.id filter), Student history (actor.mbox filter). Supports report cards, gap analysis, longitudinal tracking.

💾
Auto-Save on Next
Zero data loss

When user clicks "Next", system immediately: (1) Saves question to localStorage, (2) Sends xAPI statement to LRS (non-blocking), (3) Generates next question. Even if browser crashes or LRS fails, progress preserved locally.

💡
AI-Generated Pedagogical Feedback
Educational explanations

Every option includes detailed AI-generated feedback (2-3 sentences, 40+ words minimum). Correct options explain WHAT concept it demonstrates, HOW it applies, and WHY it's valid. Incorrect options explain misconceptions, why students might choose it, and correct understanding. Feedback stored in LRS options-analytics extension and displayed in review modal.

📖
Learn More Integration
Deep learning pathways

Review modal includes "Learn More" buttons linking to Knowledge Explorer (KE) and Socratic Playground (SPL) for each competency. URL format: /learning-session.html?classId=X&DSC=[competency]&LEVEL=Y&Mode=[ke/spl]. Enables seamless transition from assessment to deeper learning.

True SATA Enforcement
Multiple correct answers

LLM prompts enforce MANDATORY SATA format with MINIMUM 2 correct options (typically 2-3 out of 4). Distribution: either 2 correct + 2 incorrect OR 3 correct + 1 incorrect. Each correct option represents different valid approach or aspect of competency mastery. NEVER generates single-answer multiple choice questions.

Getting Started

Three simple steps to begin your proficiency assessment

1 Choose Your DSC

Navigate to /start-assessment.html or click "Start Assessment" from your dashboard. Select a Domain-Specific Competency (DSC) to assess - each DSC contains multiple competencies you'll be tested on.

2 Answer Questions

Read each scenario-based question carefully. For SATA questions, select ALL correct answers - partial credit awarded. Click "Next" to save and continue. Stop anytime - progress auto-saves to browser.

3 Track Progress

Click "View Progress" anytime to see overall score, per-competency breakdown, and proficiency levels. Resume assessment anytime by returning to same DSC - picks up where you left off.

Quick Access URLs

  • Start New Assessment: /start-assessment.html
  • Direct Assessment (with DSC ID): /assessment.html?dscId=<dsc-id>
  • Full Documentation: /docs/ASSESSMENT_SYSTEM.md

Ready to Assess Your Proficiency?

Continuous adaptive testing with IRT-based scoring and real-time analytics