Prompt Engineering for Assessment

Prompt Engineering for Educational Assessment

Guide the analysis of prompts and identification of required materials for generating xAPI statements in educational assessment contexts.

Core Concept: Context Material Mapping

To generate valid xAPI statements, identify and collect these material categories:

xAPI Component	Required Materials	Source
Actor	User ID, name, account info	Authentication system
Verb	Action type, interaction mode	Assessment design
Object	Question/quiz metadata, content	Content management
Result	Answer, score, duration	Runtime collection
Context	Course hierarchy, session info	LMS context

Prompt Analysis Framework

Step 1: Identify the Learning Activity Type

Determine which assessment scenario the prompt describes:

Scenario	Characteristics	Primary Verb
Quiz attempt	Multiple questions, scored	attempted → completed
Single question	One interaction	answered
Course progress	Module completion	progressed
Certification	Pass/fail criteria	passed/failed
Practice	No scoring	experienced

Step 2: Extract Actor Requirements

Identify learner identification needs:

Required materials:

• User identifier (account ID, email, or unique name)
• Account home page (LMS URL)
• Display name (optional but recommended)

Analysis questions:

• How are users authenticated?
• What unique identifier is available?
• Is the actor an individual or group?

Step 3: Map Verb Selection

Match the described action to appropriate verb:

Decision tree:

Is it a question response? → answered
Is it starting an activity? → attempted / launched
Is it finishing an activity? → completed
Is there a pass/fail criterion? → passed / failed
Is it passive consumption? → experienced
Is it progress without completion? → progressed

Step 4: Define Object Structure

Extract activity/question metadata:

For assessments:

• Quiz/test unique identifier (URI)
• Title and description
• Activity type (assessment, cmi.interaction)
• For questions: interaction type, choices, correct answers

For interactions:

• Question text
• Answer options (for choice types)
• Correct response pattern
• Scoring rules

Step 5: Determine Result Requirements

Identify what outcome data to collect:

Result Field	When Required	Data Source
score.scaled	Scored assessments	Calculation
score.raw/min/max	Detailed scoring	Assessment config
success	Pass/fail activities	Threshold comparison
completion	Trackable activities	Completion logic
response	Question interactions	User input
duration	Timed activities	Timer

Step 6: Build Context Requirements

Identify contextual relationships:

Hierarchical context:

• Parent: Direct container (quiz for question)
• Grouping: Broader context (course, program)
• Category: Profile compliance (cmi5)

Session context:

• Registration UUID (attempt identifier)
• Platform identifier
• Language

Material Collection Checklist

For Quiz/Assessment Scenarios

Before quiz starts (static materials):

• Quiz ID and metadata (title, description)
• Question bank with IDs
• For each question:
  • Question text
  • Interaction type
  • Answer options (if applicable)
  • Correct answer(s)
  • Scoring weight
• Pass/fail threshold (if applicable)
• Time limit (if applicable)
• Course/module hierarchy

At runtime (dynamic materials):

• Learner account information
• Session/registration ID
• Timestamp for each action
• Learner responses
• Time spent per question
• Calculated scores

For Single Question Scenarios

Static materials:

• Question ID (URI)
• Question text
• Interaction type
• Options/choices (if applicable)
• Correct response pattern
• Parent activity ID

Dynamic materials:

• Learner identifier
• Response value
• Response timestamp
• Duration

Prompt-to-xAPI Mapping Examples

Example 1: Multiple Choice Quiz

Prompt: "Track when a student answers a multiple choice question about xAPI verbs"

Analysis:

• Actor: Student (need user ID)
• Verb: answered
• Object: Question with interaction type "choice"
• Result: response, success, score

Required materials:

Static:
- Question ID: "https://lms.example.com/questions/xapi-verbs-q1"
- Question text: "Which verb indicates completion?"
- Choices: ["completed", "attempted", "launched", "terminated"]
- Correct answer: "completed"

Dynamic:
- Student account: { homePage, name }
- Response: student's selection
- Timestamp: when answered
- Duration: time to answer

Example 2: Course Completion

Prompt: "Generate xAPI when learner finishes the xAPI fundamentals course"

Analysis:

• Actor: Learner
• Verb: completed
• Object: Course activity
• Result: completion=true, duration

Required materials:

Static:
- Course ID: "https://lms.example.com/courses/xapi-fundamentals"
- Course title
- Course description

Dynamic:
- Learner account
- Total duration
- Completion timestamp

Example 3: Scored Assessment with Threshold

Prompt: "Track quiz results where 80% is passing"

Analysis:

• Actor: Learner
• Verb: passed or failed (based on score)
• Object: Assessment activity
• Result: score, success, completion

Required materials:

Static:
- Assessment ID
- Assessment metadata
- Mastery score: 0.8
- Maximum possible score

Dynamic:
- Learner account
- Raw score achieved
- Calculated scaled score
- Pass/fail determination
- Total duration

Common Patterns

Pattern: Question within Quiz

Context hierarchy for nested tracking:

Quiz (parent)
└── Question (object)

Statement structure:

{
  "object": { "id": "question-uri" },
  "context": {
    "contextActivities": {
      "parent": [{ "id": "quiz-uri" }]
    }
  }
}

Pattern: Attempt Tracking

Track individual attempts with registration:

{
  "context": {
    "registration": "attempt-uuid"
  }
}

Use same registration for all statements in one attempt.

Pattern: Multi-Language Content

Support internationalization:

{
  "object": {
    "definition": {
      "name": {
        "en-US": "Quiz Title",
        "zh-TW": "測驗標題"
      }
    }
  }
}

Validation Questions

Before generating statements, verify:

1. Actor clarity: Is the learner identifier unique and consistent?
2. Verb appropriateness: Does the verb match the actual action?
3. Object completeness: Are all required definition fields present?
4. Result accuracy: Are scores calculated correctly?
5. Context relationships: Are parent/grouping activities correct?
6. Timestamp precision: Are timestamps in ISO 8601 format?

Additional Resources

Reference Files

• references/material-templates.md - Templates for common scenarios
• references/analysis-examples.md - Detailed prompt analysis examples

Related Skills

• xAPI Specification - Detailed xAPI structure and vocabulary