Teacher-AID is an AI-powered classroom analytics system designed to help teachers understand how students are reasoning about problems, not just whether answers are correct.
The system analyzes student submissions, identifies misconceptions and reasoning patterns, and generates actionable teaching insights at both the individual student level and the entire class level.
Instead of grading alone, Teacher-AID acts as an AI teaching assistant that surfaces learning patterns, highlights common errors, and suggests targeted reteaching strategies.
Traditional educational systems track:
- Correct vs incorrect answers
- Scores
- Completion rates
Teacher-AID goes deeper by identifying:
- why a student made an error
- what concept they misunderstood
- which misconceptions are spreading in a class
The system continuously builds a learning memory of the classroom and uses it to generate insights for teachers.
When a student submits an answer, the system:
- Evaluates correctness
- Classifies the bug category or reasoning error
- Extracts the misconception behind the mistake
- Generates evidence and reasoning patterns
Example insight:
Misconception: Python range() includes the stop value
Concept: loops
Bug Category: misunderstanding_range_bounds
This allows the system to track patterns across multiple students.
Teacher-AID maintains a persistent memory of student misconceptions.
For each student, the system tracks:
- recurring misconceptions
- reasoning patterns
- evidence examples
- confidence trends
- timestamps of occurrences
This allows teachers to see how a student's understanding evolves over time.
The system also aggregates errors across the entire class.
This enables:
- identifying class-wide misconceptions
- detecting emerging vs persistent issues
- measuring how many students are affected
Example insight:
Top misconception:
"Python range() excludes the stop value"
Students affected: 5
Trend: persistent
The system generates teacher-friendly summaries of what is happening in the classroom.
Example:
The class is struggling with understanding that Python's range() function excludes the stop value.
This is leading to off-by-one errors in loops.
A short reteaching segment with visual tracing examples is recommended.
These summaries help teachers quickly decide what to reteach and when.
For individual students, the system suggests targeted interventions.
Example:
Suggested Intervention:
Give this student a short tracing exercise on Python range() boundaries.
These suggestions are generated automatically based on detected misconceptions.
The system generates AI alerts when patterns emerge.
Examples:
- A misconception becomes class-wide
- A student repeatedly struggles with a concept
- A concept trend becomes persistent
Example event:
Event Type: class_insight
Message:
"Misunderstanding of Python range() boundaries is emerging as a class-wide issue."
The frontend dashboard visualizes classroom insights.
Displays:
- top misconceptions
- concept difficulty trends
- students affected
- teacher summaries
- suggested reteaching topics
Possible visualizations:
- misconception frequency charts
- concept heatmaps
- class alerts
Displays:
- recurring misconceptions
- reasoning patterns
- intervention suggestions
- teacher summary
- evidence examples
This allows teachers to quickly understand where a student is struggling.
Shows detected events such as:
- emerging misconceptions
- recurring student errors
- class learning signals
Built with:
- FastAPI
- LLM reasoning engine (DeepSeek)
- Vector memory system (Moorcheh)
Responsibilities:
- classify student reasoning
- store misconception memory
- generate teacher insights
- detect learning events
An LLM analyzes student reasoning to extract:
- bug category
- misconception
- reasoning pattern
- supporting evidence
Embeddings are stored to allow:
- similarity search
- long-term learning memory
- pattern retrieval
This allows the system to recall prior classroom events.
GET /insights/class/{classroom_id}
Returns:
- top misconceptions
- reteach suggestions
- teacher summary
GET /insights/student/{student_id}
Returns:
- recurring misconceptions
- suggested interventions
- teacher summary
GET /events/class/{classroom_id}
Returns detected learning events.
GET /events/student/{student_id}
Returns events related to a specific student.
GET /memory/class/{classroom_id}
Returns aggregated misconception memory.
GET /memory/student/{student_id}
Returns a student's misconception history.
The repository includes a compact demo dataset representing a Grade 10 programming class.
Dataset characteristics:
- 12 students
- ~200 submissions
- multiple programming concepts
- correct and incorrect answers
Concepts covered:
- loops
- conditionals
- variables
- lists
The dataset allows the system to demonstrate:
- class-wide misconceptions
- student progression
- AI-generated teacher insights
Classroom Insight
Concept: loops
Misconception: range() includes the stop value
Trend: persistent
Students affected: 4
Recommendation:
Demonstrate loop tracing using a visual number line.
Planned enhancements include:
- timeline analysis of misconceptions
- automated curriculum alignment
- adaptive question generation
- longitudinal student learning profiles
Teachers often lack visibility into how students are thinking.
Teacher-AID helps bridge this gap by providing:
- explainable AI feedback
- actionable classroom insights
- student reasoning analysis
The goal is to help teachers teach more effectively using AI-powered learning analytics.
This project was developed for an AI education hackathon demonstration and research prototype.