Prompt Anatomy (TRACI Framework):
Placeholders to Replace:
<course_topic> — the subject area of the assessment<assessment_context> — the type of assessment, when it occurs in the program, and what it measures<learner_responses> — the actual learner responses to analyze<performance_criteria> — what "good" looks like for each question or taskExample Input:
Workplace Safety and Hazard Recognition | End-of-module knowledge check, 10 scenario-based questions, taken after a 2-hour eLearning module | [Paste learner response data] | Correct hazard identification, appropriate response selection, regulatory citation accuracy
Tip: For the most useful output, include the correct answers and scoring rubric alongside the learner responses — this allows the AI to provide precise gap analysis rather than generic feedback.
You are an instructional designer and assessment specialist with expertise in formative feedback design, Bloom's taxonomy, and evidence-based corrective instruction. Your task is to analyze learner assessment responses and generate immediate, constructive feedback that helps learners understand their performance and improve.
Course Topic:
<course_topic>
[INSERT THE SUBJECT AREA — e.g., Workplace Safety and Hazard Recognition, Project Management Fundamentals, Data Privacy and GDPR Compliance]
</course_topic>
Assessment Context:
<assessment_context>
[INSERT THE TYPE OF ASSESSMENT, WHEN IT OCCURS, AND WHAT IT MEASURES — e.g., End-of-module knowledge check with 10 scenario-based questions, taken after a 2-hour eLearning module; Mid-program skills demonstration assessed against a 5-point rubric; Pre/post assessment measuring knowledge gain across 4 competency areas]
</assessment_context>
Learner Responses:
<learner_responses>
[PASTE OR ATTACH LEARNER RESPONSES HERE — include the questions/tasks, the learner's answers, and the correct answers or scoring rubric if available]
</learner_responses>
Performance Criteria:
<performance_criteria>
[INSERT WHAT "GOOD" LOOKS LIKE — e.g., Correct hazard identification with appropriate response selection; Demonstration of all 5 coaching behaviors in the role-play; Score of 80% or above with no critical safety items missed]
</performance_criteria>
Before generating feedback, analyze the responses inside <response_analysis> tags:
1. For each question or task, determine whether the learner's response is correct, partially correct, or incorrect — and identify the specific nature of any error (factual misunderstanding, procedural gap, application failure, or reasoning error).
2. Look for patterns across responses — does the learner consistently struggle with a particular concept, skill area, or question type? A pattern of errors reveals a systematic gap rather than a random mistake.
3. Identify the Bloom's taxonomy level of each question and whether the learner's error suggests they have not mastered a prerequisite level (e.g., failing an application question may indicate a knowledge-level gap).
4. Assess the learner's overall performance relative to the performance criteria — are they close to meeting the standard or significantly below it?
5. Determine the appropriate feedback tone — a learner who scored 78% on an 80% threshold needs different feedback than one who scored 35%.
After your analysis, produce the following:
## Real-Time Feedback Report: [Course Topic]
### Overall Performance Summary
- **Score:** [X / Total] ([Percentage]%)
- **Performance threshold:** [Passing standard]
- **Result:** Met / Not Yet Met
- **Overall assessment:** [2-3 sentences summarizing performance — what the learner demonstrated well and where the primary gaps are]
### Question-by-Question Feedback
For each question or task:
**Question [#]: [Brief question description]**
- **Learner's response:** [What they answered]
- **Result:** ✓ Correct / △ Partially Correct / ✗ Incorrect
- **Feedback:** [2-3 sentences explaining why the answer is correct or incorrect — not just stating the right answer but explaining the underlying concept or reasoning. Reference the specific knowledge or skill being tested.]
- **If incorrect — Misconception identified:** [What the learner likely misunderstood or confused]
- **If incorrect — Remediation guidance:** [Specific action: review a particular section, practice a specific skill, revisit a concept with a recommended resource]
### Pattern Analysis
- **Strongest areas:** [Concepts or skills where the learner performed consistently well]
- **Primary gap:** [The single most significant knowledge or skill gap revealed by the assessment]
- **Secondary gaps:** [Additional areas needing attention, if any]
- **Bloom's level analysis:** [At what cognitive level does the learner perform well, and where do they break down? e.g., "Strong on recall and comprehension; struggles with application-level scenarios"]
### Recommended Next Steps
| Priority | Action | Rationale | Resource/Activity |
|----------|--------|-----------|-------------------|
| 1 | [Specific action] | [Why this addresses the primary gap] | [Specific module, practice exercise, or reference] |
| 2 | [Specific action] | [Why this matters] | [Resource] |
| 3 | [Specific action] | [Why this matters] | [Resource] |
### Motivational Close
[1-2 sentences that acknowledge effort, frame gaps as growth opportunities, and set a clear expectation for what the learner should do next — growth mindset framing without empty praise]
## Evaluation
After generating the feedback, evaluate it inside <feedback_evaluation> tags:
### 1. Diagnostic Precision Over Binary Scoring
Evaluate whether the feedback goes beyond "correct/incorrect" to diagnose the specific nature of each error. There are at least four distinct error types: factual misunderstanding (the learner believes something false), procedural gap (the learner knows the concept but cannot execute the steps), application failure (the learner knows the rule but cannot apply it to a new scenario), and reasoning error (the learner follows a logical chain that breaks at a specific point). Each error type requires a different remediation approach. Feedback that treats all errors the same — "review the material and try again" — fails to help the learner understand what went wrong.
| Excellent | Adequate | Needs Revision |
|-----------|----------|----------------|
| Each incorrect response is diagnosed with a specific error type (factual, procedural, application, reasoning); the feedback explains what the learner likely misunderstood and why that misunderstanding leads to the observed error; different error types receive different remediation guidance | Errors are identified and the correct answer is explained, but the specific nature of the misunderstanding is not diagnosed; all errors receive similar remediation advice ("review the content") | Feedback states whether each answer is correct or incorrect and provides the right answer without explaining why the learner's response was wrong or what they misunderstood |
### 2. Pattern Recognition Across Responses
Evaluate whether the feedback identifies systematic patterns across multiple responses rather than treating each question as an isolated event. A learner who misses three application-level scenario questions while answering all recall questions correctly has a different gap than one who misses questions randomly across all types. The pattern analysis should connect individual errors to an overarching gap and prioritize the remediation accordingly — fixing the root pattern is more efficient than addressing each error independently.
| Excellent | Adequate | Needs Revision |
|-----------|----------|----------------|
| Explicit pattern analysis connects errors across questions to identify systematic gaps; the primary gap is clearly distinguished from random errors; remediation targets the pattern rather than individual questions; Bloom's level analysis reveals where the learner's cognitive processing breaks down | Some patterns are noted but the analysis does not clearly distinguish systematic gaps from isolated mistakes; remediation addresses individual questions rather than the underlying pattern | Each question is analyzed independently with no cross-question pattern analysis; no attempt to identify systematic gaps or cognitive level breakdowns |
### 3. Bloom's Taxonomy Alignment in Remediation
Evaluate whether the remediation guidance is calibrated to the Bloom's taxonomy level of the gap. If a learner fails an application-level question, the appropriate remediation is not "re-read the definition" (knowledge level) but rather "practice applying the concept to three different scenarios" (application level). The feedback should identify at what cognitive level the learner demonstrates competence and at what level they break down, then target remediation at the transition point — not at a level they have already mastered.
| Excellent | Adequate | Needs Revision |
|-----------|----------|----------------|
| Each remediation recommendation is calibrated to the Bloom's level of the identified gap; the feedback explicitly states the cognitive level where the learner breaks down and targets practice at that level; remediation for application failures differs from remediation for recall failures | Remediation recommendations are present but not explicitly calibrated to cognitive levels; some recommendations target the right level by coincidence rather than design | All remediation recommendations are the same type regardless of the nature of the gap (e.g., "review the material" for both recall and application failures); no Bloom's level analysis |
### 4. Feedback Tone and Growth Mindset Framing
Evaluate whether the feedback maintains a growth mindset tone throughout — especially in the Needs Improvement scenarios. The feedback should frame gaps as "not yet" rather than "failed," describe what the learner needs to DO rather than what they ARE, and avoid both empty praise ("Great effort!") and discouraging language ("You clearly didn't understand this"). The motivational close should set a specific, achievable next step rather than a vague encouragement. For high performers, the feedback should challenge them to go deeper rather than simply praising their score.
| Excellent | Adequate | Needs Revision |
|-----------|----------|----------------|
| Feedback uses "not yet" framing consistently; gaps are described as actions to take rather than labels; high performers receive stretch challenges; low performers receive specific next steps that feel achievable; no empty praise or discouraging language | Tone is generally positive but some phrases slip into fixed mindset territory ("you struggled with") or empty praise ("great job"); high performers receive praise without stretch goals | Feedback tone is either discouraging ("you failed to understand") or artificially positive ("amazing effort!") without substance; gaps are described as personal deficiencies; no differentiation in tone based on performance level |
### 5. Remediation Specificity and Resource Targeting
Evaluate whether the recommended next steps point to specific, actionable resources or activities — not vague directives. "Review Module 3" is not helpful if Module 3 is 45 minutes long and the learner's gap is in one specific concept covered in a 3-minute segment. The remediation should specify what to review (a specific section, concept, or skill), how to practice (a specific exercise type or scenario), and how to verify improvement (a self-check or follow-up assessment). Each recommendation should be proportional to the size of the gap.
| Excellent | Adequate | Needs Revision |
|-----------|----------|----------------|
| Each recommendation specifies the exact content to review, the practice activity to complete, and the verification method to confirm improvement; recommendations are proportional to gap size; the #1 priority action is the single most efficient path to meeting the performance threshold | Recommendations reference specific modules or topics but lack practice activities or verification methods; some recommendations are too broad for the size of the gap | Recommendations are generic ("study more," "review the material," "try again") with no specific content, practice, or verification guidance |
### 6. Timeliness and Cognitive Load Calibration
Evaluate whether the feedback is structured for immediate use — not a research paper the learner needs to study. Real-time feedback should lead with the most important information (overall result and primary gap), provide question-level detail that can be scanned quickly, and close with no more than three prioritized actions. A feedback report that provides exhaustive analysis of every question equally buries the signal in noise. The structure should use visual hierarchy (headers, bold text, icons) to enable rapid scanning, and the total volume should be proportional to the learner's needs — high performers get brief confirmation and a stretch challenge; struggling learners get focused remediation on the primary gap.
| Excellent | Adequate | Needs Revision |
|-----------|----------|----------------|
| Feedback leads with overall result and primary gap; question-level detail is scannable with clear visual hierarchy; recommended actions are limited to three prioritized items; feedback volume is calibrated to performance level (brief for high performers, focused for low performers) | Feedback is organized but all questions receive equal detail regardless of importance; recommended actions are not clearly prioritized; feedback length does not vary by performance level | Feedback is a wall of text with no visual hierarchy; every question receives exhaustive analysis regardless of whether the learner answered correctly; no clear prioritization of actions; the learner would need to study the feedback to find the key takeaway |