# 4. Grading

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

---

## 4. Grading

### 4.1 Instant Results
**What it does:** Delivers instant results for assessments: as soon as a learner submits an assessment (or completes it, for auto-graded formats), the platform grades the objective questions and presents the result immediately. The Super Administrator configures the instant-results behavior so the learner gets immediate feedback on their performance, without waiting for a manual grade.

**Sub-features:**
- Instant grading: the objective questions (MCQ, fill-in-the-blank, true/false) graded on submission
- Immediate result: the score presented to the learner right after submission
- Result detail: the score, the pass/fail, and the per-question breakdown
- Auto-grade scope: the question types that are auto-graded (the objective ones)
- Manual-grade handoff: the subjective questions (short answer, essay) routed for grading
- Result recording: the result recorded against the learner and the assessment
- Audit logging of the result recording

**Super Administrator User Journey:**
1. For an assessment, Super Admin confirms the instant-results behavior: the objective questions are graded on submission, and the result is presented immediately.
2. Sets the auto-grade scope: the question types that are auto-graded (multiple choice, fill-in-the-blank, true/false).
3. Sets the manual-grade handoff: the subjective questions (short answer, essay) are routed for grading (AI or manual, per the configuration).
4. A learner submits the assessment; the platform grades the objective questions instantly and presents the score, the pass/fail, and the per-question breakdown.
5. For the subjective questions, the result is pending (the grading in progress); the learner sees the objective result now and the full result once the subjective grading completes.
6. The result is recorded against the learner and the assessment; it is available in the learner's history and the assessment's analytics.
7. For a result that is disputed (a learner challenges a question's grading), Super Admin reviews the question and the grading; a correction is applied and recorded.
8. The result recording is audit-logged with the learner, the assessment, the result, and the timestamp.

**Rules & Edge Cases:**
- Instant results apply to the objective questions; they are graded on submission and presented immediately.
- The subjective questions are routed for grading (AI or manual); the full result is complete once they are graded.
- The result includes the score, the pass/fail, and the per-question breakdown; it is specific, not just a number.
- A result is recorded against the learner and the assessment; it is in the history and the analytics.
- A disputed result is reviewed; a correction is applied and recorded (the original is not silently overwritten).
- Result recording is audit-logged with the learner, assessment, result, and timestamp.

### 4.2 AI Grading for Short Answer and Essay Questions
**What it does:** Grades the subjective questions (short answer and essay) using the platform's in-house AI: the AI evaluates the learner's response against the marking criteria and produces a grade with a rationale. The Super Administrator configures the AI grading (the criteria, the scale) and can review/override the AI's grades, so the subjective grading is consistent and scalable, with human oversight.

**Sub-features:**
- AI grading: the short answer/essay responses graded by the in-house AI
- Marking criteria: the criteria the AI grades against (the rubric)
- Grade scale: the scale the AI uses (the marks, the bands)
- Rationale: the AI's reasoning for the grade (the feedback)
- Human review: the AI grades reviewed (a sample or all, per the configuration)
- Override: a human override of an AI grade (with the reason)
- Audit logging of the AI grades and overrides

**Super Administrator User Journey:**
1. For an assessment with subjective questions, Super Admin configures the AI grading: the marking criteria (the rubric) and the grade scale (the marks/bands).
2. Sets the human review: the AI grades reviewed (a sample for a high-volume assessment, all for a high-stakes one), per the configuration.
3. A learner submits the short answer/essay; the AI grades the response against the criteria and produces the grade with a rationale (the feedback).
4. For the reviewed grades, Super Admin (or the designated reviewer) checks the AI's grade against the response and the criteria.
5. For a grade that is off (the AI missed a point, or was too lenient/strict), the reviewer overrides it with the correct grade and the reason; the override is recorded.
6. For a grade that is sound, the reviewer confirms it; the AI grade stands.
7. Reviews the AI grading quality: the override rate (a high override rate signals a criteria issue), so the criteria are refined.
8. The AI grades and overrides are audit-logged with the response, the grade, the actor, and the timestamp.

**Rules & Edge Cases:**
- The AI grades the subjective questions against the marking criteria; the criteria are the rubric, not an implicit standard.
- The AI grade includes a rationale (the feedback); the learner sees why they got the grade.
- The human review is per the configuration (a sample or all); a high-stakes assessment is reviewed more thoroughly.
- An override is recorded with the reason; the AI's original grade is preserved (not destroyed).
- The override rate is a quality signal; a high rate triggers a criteria review.
- AI grades and overrides are audit-logged with the response, grade, actor, and timestamp.

### 4.3 Detailed Explanations per Question
**What it does:** Provides a detailed explanation for each question in the result: the correct answer, why it is correct, and why the other options (or the learner's answer) are not. The Super Administrator ensures the explanations are present and accurate, so the result is a learning moment — the learner understands not just their score, but the reasoning behind each question.

**Sub-features:**
- Per-question explanation: the explanation attached to each question
- Correct answer detail: the correct answer and the reasoning for it
- Wrong-answer detail: why the learner's answer (or the other options) is not correct
- Explanation coverage: the explanations present for the assessment's questions
- Explanation accuracy: the explanations checked for correctness
- Result presentation: the explanations shown in the result (per question)
- Audit logging of explanation changes

**Super Administrator User Journey:**
1. For an assessment, Super Admin confirms the explanation coverage: each question has a detailed explanation (the correct answer, the reasoning, and why the alternatives are not correct).
2. For a question missing an explanation, adds it (the correct answer and the reasoning); the coverage is complete.
3. Checks the explanation accuracy: the explanations are correct (a wrong explanation is as bad as a wrong answer); an inaccurate one is corrected.
4. A learner completes the assessment and reviews the result: per question, the explanation shows the correct answer, why it is correct, and why their answer was not.
5. For a question the learner got wrong, the explanation clarifies the concept (the learning moment); the learner can revisit the topic.
6. For a question the learner got right, the explanation reinforces the reasoning (why the correct answer is correct).
7. Reviews the explanations across the assessment: the coverage is complete and the accuracy is sound, so the result is a genuine learning tool.
8. The explanation changes are audit-logged with the question, the explanation, and the timestamp.

**Rules & Edge Cases:**
- Each question has a detailed explanation; the coverage is complete (a missing explanation is a gap).
- The explanation covers the correct answer (the reasoning) and the wrong answer (why it is not correct); it is specific.
- The explanation accuracy is checked; an inaccurate explanation is corrected (it is as important as the answer).
- The explanations are shown in the result, per question; the result is a learning moment, not just a score.
- An explanation change is recorded; the current explanation is the accurate one.
- Explanation changes are audit-logged with the question, explanation, and timestamp.