# 2. Question Bank Management

User Type: **Super Administrator**
Source: *Mi Digital Academy - Education CRM Features Document*

---

## 2. Question Bank Management

### 2.1 Create and Curate Questions Manually
**What it does:** Manages the manual creation and curation of questions in the question bank: the Super Administrator (or the content owners they coordinate) authors questions by hand, with the full detail (the question, the options, the correct answer, the explanation), and curates the bank so it is a high-quality, organized source of questions for the assessments. Manual curation ensures the questions are accurate, well-phrased, and fit for purpose.

**Sub-features:**
- Manual question authoring: the question, options, correct answer, and explanation written by hand
- Question detail: the full content of the question (stem, options, answer, explanation)
- Curation: the bank reviewed for quality (accuracy, phrasing, clarity)
- Organization: the questions organized by topic, subject, and type
- Duplicate detection: similar questions flagged to avoid redundancy
- Quality standard: the questions meet the bank's quality bar
- Audit logging of question create/edit

**Super Administrator User Journey:**
1. A new question is needed for a topic. Super Admin (or the content owner) authors it manually: the stem, the options, the correct answer, and the explanation.
2. Sets the question's attributes: the type, the difficulty, and the topic/subject it belongs to.
3. Saves; the question is added to the bank and is available for assessment building.
4. Cures the bank: reviews the questions for quality (accuracy, phrasing, clarity); a poorly phrased or inaccurate question is corrected.
5. For a duplicate (a similar question already in the bank), the duplicate is flagged; Super Admin consolidates (keeps the better one, removes the redundant).
6. Organizes the bank: the questions by topic, subject, and type, so the right question is found quickly when building an assessment.
7. Confirms the quality standard: the bank's questions meet the bar (accurate, clear, well-phrased), so the assessments built from it are sound.
8. The question create and edit events are audit-logged with the question, the actor, and the timestamp.

**Rules & Edge Cases:**
- A manually authored question has the full detail (stem, options, answer, explanation); it is complete, not a stub.
- The bank is curated for quality; an inaccurate or poorly phrased question is corrected, not left.
- Duplicates are flagged and consolidated; the bank is not redundant.
- The questions are organized by topic, subject, and type; the right question is findable.
- The quality standard is the bank's bar; a question below it is corrected before it is used.
- Question create/edit is audit-logged with the question, actor, and timestamp.

### 2.2 AI-Generated Questions from Video Content
**What it does:** Generates questions automatically from video content using the platform's in-house AI: the system produces draft questions based on a video's content, which the Super Administrator (or the content owners) then reviews and approves before they enter the bank. AI generation accelerates question creation; human review ensures the generated questions are accurate and fit for use.

**Sub-features:**
- AI generation: questions generated from a video's content
- Draft status: the generated questions are drafts, not final
- Human review: the generated questions reviewed for accuracy before approval
- Approval: the reviewed questions approved into the bank (or corrected/rejected)
- Generation parameters: the question count, types, and difficulty requested
- Traceability: the AI-generated questions marked as such (the source video)
- Audit logging of generation and approval

**Super Administrator User Journey:**
1. For a video, Super Admin requests AI-generated questions: the count, the types, and the difficulty, from the video's content.
2. The platform's in-house AI generates the draft questions from the video content; they appear as drafts (not yet in the bank).
3. Super Admin reviews the generated questions for accuracy: the questions match the video's content and the answers are correct.
4. For a correct question, approves it into the bank (it is now available for assessment building).
5. For an inaccurate or off-topic question, corrects it (and approves the correction) or rejects it (it does not enter the bank).
6. Confirms the traceability: the AI-generated questions are marked as such, with the source video, so their origin is known.
7. Reviews the generation parameters for the next batch (adjusting the count/types/difficulty based on the review outcome).
8. The generation and approval events are audit-logged with the video, the questions, the actor, and the timestamp.

**Rules & Edge Cases:**
- AI-generated questions are drafts; they enter the bank only after human review and approval.
- The review checks accuracy (the question matches the content, the answer is correct); an inaccurate question is corrected or rejected.
- The generation parameters (count, types, difficulty) are set per request; they are explicit.
- The AI-generated questions are marked as such, with the source video; the origin is traceable.
- A rejected question does not enter the bank; the rejection is recorded.
- Generation and approval are audit-logged with the video, questions, actor, and timestamp.

### 2.3 Question Types, Difficulty Tagging, and Bloom's Taxonomy Alignment
**What it does:** Defines the question's classification attributes: the question type (multiple choice, fill-in-the-blank, true/false, short answer, essay), the difficulty tag, and the Bloom's Taxonomy level (the cognitive level the question targets). The Super Administrator sets these attributes so the questions are classified consistently, the assessments can be built to a target mix (types, difficulties, cognitive levels), and the question bank is analyzable by these dimensions.

**Sub-features:**
- Question types: multiple choice, fill-in-the-blank, true/false, short answer, essay
- Difficulty tagging: the question tagged by difficulty (e.g., easy, medium, hard)
- Bloom's Taxonomy alignment: the question mapped to a cognitive level (remember, understand, apply, analyze, evaluate, create)
- Consistent classification: the attributes set for every question
- Mix building: assessments built to a target mix of types, difficulties, and levels
- Bank analysis: the bank analyzed by the classification dimensions
- Audit logging of classification changes

**Super Administrator User Journey:**
1. For a question, Super Admin sets its type (multiple choice, fill-in-the-blank, true/false, short answer, or essay), matching the question's actual format.
2. Tags the difficulty (easy, medium, hard), reflecting the question's actual challenge level.
3. Maps the Bloom's Taxonomy level (the cognitive level the question targets: remember, understand, apply, analyze, evaluate, create).
4. Saves; the question is now classified, and it is available for mix-based assessment building.
5. For an assessment, builds it to a target mix: the types, the difficulties, and the cognitive levels, so the assessment is balanced (not all easy, not all one type).
6. Analyzes the bank by the classification dimensions: the coverage per type, difficulty, and level, so a gap (no hard questions on a topic) is visible.
7. For a question mis-classified (a wrong difficulty tag), corrects the attribute; the change is recorded.
8. The classification changes are audit-logged with the question, the attributes, and the timestamp.

**Rules & Edge Cases:**
- Every question has a type, a difficulty tag, and a Bloom's level; the classification is complete.
- The attributes reflect the question's actual format, challenge, and cognitive level; a mis-classification is corrected.
- The classification enables mix-based assessment building (a target mix of types, difficulties, levels).
- The bank is analyzable by the dimensions; a coverage gap is visible, not silent.
- The classification is consistent across the bank; the same question type/level is classified the same way.
- Classification changes are audit-logged with the question, attributes, and timestamp.

### 2.4 Regular Question Refreshing
**What it does:** Keeps the question bank fresh: the Super Administrator manages the regular refreshing of the bank — replacing outdated questions, adding new ones, and rotating the question set so the assessments do not become stale or memorizable. Refreshing maintains the bank's quality and the assessments' integrity over time.

**Sub-features:**
- Refresh cycle: the regular review and update of the bank
- Outdated question replacement: stale questions replaced with current ones
- New question addition: fresh questions added to keep the bank current
- Rotation: the question set rotated so assessments are not repetitive
- Staleness detection: questions flagged as outdated (content changed, over-used)
- Refresh review: the refreshed bank reviewed for quality and coverage
- Audit logging of refresh actions

**Super Administrator User Journey:**
1. On the refresh cycle, Super Admin reviews the bank for staleness: the questions flagged as outdated (the content they cover changed, or they are over-used).
2. Replaces the outdated questions: the stale ones are retired, and current replacements are added (authored or AI-generated and reviewed).
3. Adds new questions to keep the bank current (covering the updated content, filling the coverage gaps).
4. Rotates the question set: the questions used in assessments are rotated, so the assessments are not repetitive or memorizable.
5. Reviews the refreshed bank for quality and coverage: the new questions meet the bar, and the coverage is adequate (no gap introduced by the refresh).
6. For a question that is over-used (appearing in too many assessments), retires it from active rotation (it is not deleted, but it is not reused).
7. Confirms the refresh outcome: the bank is current, the coverage is adequate, and the assessments built from it are fresh.
8. The refresh actions are audit-logged with the questions, the action (replace/add/rotate/retire), and the timestamp.

**Rules & Edge Cases:**
- The refresh is a regular cycle; it is not a one-time event.
- An outdated question is replaced (retired and a current one added); it is not silently left stale.
- Rotation prevents repetitiveness; an over-used question is retired from active rotation (not deleted).
- The refresh is reviewed for quality and coverage; a gap introduced by the refresh is closed.
- A retired question is preserved (not deleted); it is out of active rotation.
- Refresh actions are audit-logged with the questions, action, and timestamp.