A Framework for Analyzing the Causes of Student Laziness Based on Knowledge Graph and Multi-Agent System

This article introduces the student-laziness-kg-multi-agent project released by hungvv109 on GitHub (release date: 2026-06-06, link: https://github.com/hungvv109/student-laziness-kg-multi-agent). This project combines Neo4j knowledge graph, LangGraph multi-agent workflow, and Google Gemini to build an innovative framework for systematically analyzing the root causes of student laziness, addressing the problem that traditional methods struggle to capture complex correlations and deep-seated causal relationships.

Student laziness is a complex issue in education, with causes spanning multiple dimensions such as psychology, environment, physiology, and society. Traditional methods (questionnaires, simple statistics) struggle to capture complex correlations and deep-seated causal relationships between factors. This project proposes a solution combining knowledge graph and multi-agent system, which can identify individual factors and their interaction networks, providing precise basis for educational interventions.

Neo4j Knowledge Graph

• Nodes: Entities like students, behavioral performance, psychological/environmental/physiological factors
• Relationships: Associations such as 'causes', 'influences', 'belongs to'
• Attributes: Features like age, gender, academic performance

LangGraph Multi-Agent Division of Labor

• Data Collection Agent: Collects academic/behavioral/psychological/environmental data, cleans and preprocesses it
• Pattern Recognition Agent: Analyzes data to identify laziness patterns and key indicators
• Causal Reasoning Agent: Core component, explores causal chains on the knowledge graph
• Recommendation Generation Agent: Generates personalized intervention suggestions
• Verification Agent: Verifies the accuracy of results and effectiveness of suggestions

Google Gemini

Provides text understanding, knowledge extraction (to expand the graph), reasoning assistance, and report generation capabilities.

1. Data Ingestion and Preprocessing: Collect multi-source data (academic/behavioral/psychological/environmental), clean and standardize it
2. Knowledge Graph Construction: Entity recognition, relationship extraction, graph modeling, data import into Neo4j
3. Multi-Agent Collaborative Analysis: Cross-verify pattern recognition and causal reasoning results
4. Root Cause Identification and Recommendation Generation: Identify root causes and generate personalized intervention plans
5. Result Verification and Feedback: Track intervention effects and optimize system strategies

• Educational Personalization: Develop tutoring strategies based on students' unique causes
• Early Warning: Identify early signs of laziness to prevent academic failure
• Educational Research: Accumulate data to support research on educational laws
• Intelligent Tutoring Systems: Integrate into systems to provide 24/7 personalized support

• Data Privacy: Protect sensitive data through desensitization, access control, and encrypted storage
• Uncertainty in Causal Relationships: Use probabilistic graph models and uncertainty reasoning to quantify confidence levels
• Dynamic Update of Knowledge Graph: Design an incremental update mechanism to ensure timeliness

This project demonstrates the innovative application of knowledge graph and multi-agent systems in the education field. Combining Neo4j, LangGraph, and Gemini provides new ideas for analyzing complex educational issues. The architecture can be extended to scenarios such as learning difficulty diagnosis and career planning, and will play a more important role in education in the future.