Competency Framework & Knowledge Graph — Requirements Specification

EduOne AI Platform | Version 1.0

Field	Value
Document ID	CF-RS-001
Version	1.0
Date	2026-07-17
Status	Draft
Owner	EduOne AI Platform Team

---

Table of Contents

1. Purpose & Scope
2. Framework Reference Architecture
3. Competency Hierarchy
4. Skill Structure
5. Skill Dependency Graph
6. Proficiency Levels
7. Root-Cause Diagnosis
8. End-to-End Example
9. Hackathon MVP Scope
10. Functional Requirements

---

1. Purpose & Scope

1.1 What is a Competency Framework & Knowledge Graph

A Competency Framework is a structured, hierarchical model that defines the capabilities a learner or professional needs to demonstrate in a given domain. It organizes knowledge into domains, competency areas, competencies, skills, and measurable indicators, providing a shared language for educators, learners, and AI systems.

A Knowledge Graph extends the framework by encoding the relationships between skills — prerequisites, co-requisites, conceptual foundations, and transfer pathways — as a directed graph. Each node represents a skill with its proficiency levels, and each edge represents a typed dependency with strength, confidence, and diagnostic metadata.

Together, the Competency Framework & Knowledge Graph form a reasoning substrate that AI systems can traverse to:

• Diagnose why a learner is failing at a particular task
• Trace the failure back to its root cause in foundational skills
• Recommend the minimum remediation path to close the gap
• Predict which skills are at risk based on upstream weaknesses

graph LR
    A["Static Skill List"] -->|"+ hierarchy"| B["Competency Framework"]
    B -->|"+ relationships"| C["Knowledge Graph"]
    C -->|"+ AI reasoning"| D["Diagnostic Engine"]
    D -->|"outputs"| E["Root-Cause Diagnosis"]
    D -->|"outputs"| F["Remediation Path"]
    D -->|"outputs"| G["Mastery Prediction"]

1.2 Why It Matters

Traditional learning platforms treat skills as flat, independent items. When a learner fails a quiz on "Neural Network Training," the system can only report the failure — it cannot explain why the learner failed or what foundational gaps caused it.

The Competency Framework & Knowledge Graph transforms static skill taxonomies into a reasoning graph that AI can traverse for root-cause diagnosis. This enables:

Capability	Without Knowledge Graph	With Knowledge Graph
Failure reporting	"You scored 40% on Neural Networks"	"Your failure is likely caused by a gap in Linear Algebra (matrix multiplication)"
Remediation	"Review Neural Networks chapter"	"Complete these 3 prerequisite exercises in Linear Algebra first"
Learning path	Fixed, one-size-fits-all sequence	Adaptive path based on individual gap analysis
Teacher insight	Aggregate class scores	"60% of students who fail ML also lack probability foundations"
Curriculum design	Expert intuition only	Data-driven dependency validation

1.3 Scope for EduOne

EduOne adopts a unified framework that combines and harmonizes the following international reference frameworks:

• UNESCO AI Competency Framework — for AI literacy dimensions and progression levels
• ISTE Standards — for computational thinking pillars
• SFIA 9 — for professional skill levels and responsibility mapping
• WEF Education 4.0 & Global Skills Taxonomy — for future-of-work skill classification

The unified approach ensures that EduOne's competency model is:

1. Internationally grounded — mapped to recognized standards
2. Locally adaptable — can be customized for national curricula (e.g., Vietnam MOET)
3. Vertically integrated — covers K-12, higher education, and professional development
4. AI-native — designed from the ground up for machine reasoning, not just human browsing

graph TD
    UNESCO["UNESCO AI<br/>Competency Framework"]
    ISTE["ISTE Standards"]
    SFIA["SFIA 9"]
    WEF["WEF Education 4.0"]

    UNESCO --> UF["EduOne Unified<br/>Competency Framework"]
    ISTE --> UF
    SFIA --> UF
    WEF --> UF

    UF --> KG["Knowledge Graph"]
    KG --> DX["Diagnostic Engine"]
    KG --> LP["Learning Paths"]
    KG --> CA["Curriculum Analytics"]

---

2. Framework Reference Architecture

2.1 Framework Types Overview

The following table defines the seven framework types that together form a complete competency architecture. EduOne's Knowledge Graph integrates all seven.

#	Framework Type	Question It Answers	Example	Role in EduOne
1	Competency Framework	What capabilities does a person need?	AI Literacy, Computational Thinking, Data Ethics	Defines the what — the target capabilities
2	Skill Taxonomy	How are skills defined and classified?	Programming → Python → Functions → Recursion	Provides the structure — hierarchical classification
3	Proficiency Framework	How skilled is the learner?	Foundation → Applied → Advanced → Expert	Defines the how much — levels of mastery
4	Curriculum Framework	What content and in what order?	Grade 6: Scratch, Grade 8: Python, Grade 10: Data Science	Defines the when — learning sequence
5	Assessment Framework	How to assess competency?	Quiz, coding task, project defense, peer review	Defines the how to measure — evidence types
6	Skill Dependency Graph	What must be learned first?	Linear Algebra → Neural Networks, Statistics → Hypothesis Testing	Defines the why before — prerequisite relationships
7	Role-Skill Mapping	What skills does a role need?	Data Analyst → SQL, Statistics, Visualization; ML Engineer → Python, TensorFlow, MLOps	Defines the who needs what — career alignment

graph TB
    CF["1. Competency<br/>Framework"] --> ST["2. Skill<br/>Taxonomy"]
    ST --> PF["3. Proficiency<br/>Framework"]
    CF --> CuF["4. Curriculum<br/>Framework"]
    CuF --> AF["5. Assessment<br/>Framework"]
    ST --> SDG["6. Skill Dependency<br/>Graph"]
    CF --> RSM["7. Role-Skill<br/>Mapping"]

    SDG --> KG["Knowledge Graph<br/>(Unified)"]
    PF --> KG
    AF --> KG
    RSM --> KG

2.2 UNESCO AI Competency Framework for Students

The UNESCO AI Competency Framework for Students (2024) defines AI literacy across 4 dimensions and 3 progression levels.

4 Dimensions:

Dimension	Description	Key Competencies
1. AI Mindset	Understanding what AI is, its capabilities and limitations	AI awareness, critical evaluation of AI outputs, growth mindset toward AI
2. AI Use & Application	Using AI tools effectively and responsibly	Prompt engineering, AI tool selection, workflow integration
3. AI Ethics & Social Impact	Understanding societal implications	Bias detection, privacy awareness, responsible AI use
4. AI Foundations	Technical understanding of how AI works	Data representation, ML concepts, pattern recognition

3 Progression Levels:

Level	Name	Description	Typical Age
Level 1	Explore	Awareness and basic interaction with AI	Ages 6-12
Level 2	Practice	Guided application and critical evaluation	Ages 12-16
Level 3	Create	Independent creation and responsible deployment	Ages 16-18+

Strengths:

• Internationally recognized and widely adopted
• Comprehensive coverage of both technical and ethical dimensions
• Age-appropriate progression with clear milestones

Limitations:

• Primarily focused on K-12; lacks professional development levels
• Does not define fine-grained skill dependencies
• Assessment criteria are descriptive, not machine-readable

2.3 UNESCO AI Competency Framework for Teachers

The UNESCO AI Competency Framework for Teachers defines competencies across 5 professional groups:

Group	Focus Area	Example Competencies
Group 1	Human-centred mindset	Critical thinking about AI, AI awareness for educators
Group 2	AI ethics	Responsible AI integration in teaching, bias awareness in educational AI
Group 3	AI foundations & applications	Understanding ML basics, selecting AI teaching tools
Group 4	AI pedagogy	Designing AI-enhanced lessons, assessment with AI, differentiated instruction
Group 5	AI for professional development	Using AI for teacher growth, collaborative AI practices, research with AI

NOTE

EduOne uses the Teacher Framework to define competency requirements for educator roles in the Role-Skill Mapping, ensuring that teachers using the platform can also be assessed and supported.

2.4 ISTE Standards

The International Society for Technology in Education (ISTE) Standards define 4 pillars of Computational Thinking (CT):

Pillar	Definition	Example in AI Context
Decomposition	Breaking complex problems into smaller, manageable parts	Breaking "build a chatbot" into: data collection, model selection, training, evaluation, deployment
Pattern Recognition	Identifying similarities, trends, and regularities in data	Recognizing that spam emails share common word patterns
Abstraction	Focusing on important information while ignoring irrelevant details	Representing a student as a feature vector [GPA, attendance, scores] for prediction
Algorithm Design	Developing step-by-step instructions to solve a problem	Designing a decision tree algorithm for student placement

Integration with EduOne:

• ISTE CT pillars map directly to Competency Area C (Computational Thinking) in the EduOne hierarchy
• Each pillar becomes a competency with its own skill tree and proficiency levels
• ISTE's emphasis on "CT for all subjects" aligns with EduOne's cross-domain dependency edges

2.5 SFIA 9

The Skills Framework for the Information Age (SFIA) version 9 provides a professional skills framework with 7 levels of responsibility:

Level	Name	Autonomy	Influence	Complexity	Business Skills
1	Follow	Works under supervision	Minimal	Routine tasks	Learning
2	Assist	Works under routine direction	Individual	Straightforward	Assisting
3	Apply	Works under general direction	Colleagues	Moderately complex	Applying
4	Enable	Works with broad direction	Team/project	Complex	Enabling
5	Ensure, Advise	Works with defined accountability	Organization unit	Significant	Advising
6	Initiate, Influence	Defines organizational strategy	Organization-wide	Unpredictable	Influencing
7	Set Strategy	Shapes industry practices	Industry/domain	Strategic	Leading

Integration with EduOne:

• SFIA levels inform the Professional Responsibility dimension of EduOne's proficiency model
• SFIA skill definitions are mapped to EduOne competencies via external reference IDs
• SFIA's responsibility model is used for Role-Skill Mapping in professional/higher-ed contexts

2.6 WEF Education 4.0 & Global Skills Taxonomy

The World Economic Forum's Education 4.0 framework and Global Skills Taxonomy define future-of-work skills across categories:

Category	Skills	Relevance to EduOne
Problem-Solving	Analytical thinking, creative thinking, systems thinking	Maps to Competency Areas A, B, C
Technology	AI/ML, programming, data analysis, cybersecurity	Maps to Competency Areas D, E, F, G
Self-Management	Resilience, flexibility, lifelong learning, curiosity	Cross-cutting skills in proficiency progression
Working with People	Leadership, social influence, teamwork, empathy	Maps to Competency Area J

WEF's Top 10 Skills for 2025-2030:

1. Analytical thinking
2. Creative thinking
3. AI and big data
4. Leadership and social influence
5. Resilience, flexibility, and agility
6. Curiosity and lifelong learning
7. Technological literacy
8. Design and user experience
9. Motivation and self-awareness
10. Empathy and active listening

2.7 Conclusion: Unified Approach

No single framework is sufficient for EduOne's needs. Each framework contributes a critical dimension:

graph LR
    UNESCO_S["UNESCO Students"] -->|"AI literacy<br/>dimensions"| UF["EduOne Unified<br/>Framework"]
    UNESCO_T["UNESCO Teachers"] -->|"Educator<br/>competencies"| UF
    ISTE["ISTE Standards"] -->|"CT pillars"| UF
    SFIA["SFIA 9"] -->|"Professional<br/>levels"| UF
    WEF["WEF Education 4.0"] -->|"Future-of-work<br/>skills"| UF

    UF -->|"Hierarchy"| H["6-Level<br/>Competency Tree"]
    UF -->|"Dependencies"| D["Skill Dependency<br/>Graph"]
    UF -->|"Proficiency"| P["Dual Proficiency<br/>Model"]
    UF -->|"Mappings"| M["External<br/>Reference IDs"]

EduOne's unified approach:

1. Adopts UNESCO's AI literacy dimensions as the primary organizing structure for AI competencies
2. Integrates ISTE's CT pillars as a dedicated competency area
3. Uses SFIA's 7-level responsibility model for professional proficiency
4. Maps WEF skills to competency areas for career relevance
5. Extends all frameworks with machine-readable dependency edges for AI reasoning

---

3. Competency Hierarchy

3.1 Hierarchy Structure

The EduOne Competency Framework uses a 6-level hierarchy from broadest to most specific:

Domain → Competency Area → Competency → Skill → Indicator → Evidence

graph TD
    D["Domain<br/><i>e.g., STEM & AI Education</i>"]
    D --> CA["Competency Area<br/><i>e.g., C. Computational Thinking</i>"]
    CA --> C["Competency<br/><i>e.g., C3. Algorithmic Design</i>"]
    C --> S["Skill<br/><i>e.g., gradient_descent</i>"]
    S --> I["Indicator<br/><i>e.g., Can explain convergence conditions</i>"]
    I --> E["Evidence<br/><i>e.g., Quiz score ≥ 80%, Code submission</i>"]

Level	Description	Typical Count	Example
Domain	Broadest classification	1-3	STEM & AI Education
Competency Area	Major thematic grouping	10 per domain	Computational Thinking
Competency	Specific capability cluster	6-11 per area	Algorithmic Design
Skill	Atomic, assessable unit	3-8 per competency	Gradient Descent
Indicator	Observable behavior at a proficiency level	2-5 per skill per level	"Can implement gradient descent from scratch"
Evidence	Artifact that demonstrates the indicator	1-3 per indicator	Code submission, quiz answer, project deliverable

3.2 Competency Areas and Competencies

The EduOne framework defines 10 Competency Areas under the STEM & AI Education domain, containing a total of 86 competencies.

---

A. Scientific Thinking (6 Competencies)

ID	Competency	Description
A1	Observation & Questioning	Formulating clear, testable questions based on careful observation of phenomena
A2	Hypothesis Formulation	Constructing evidence-based hypotheses that can be tested experimentally
A3	Experimental Design	Designing controlled experiments with appropriate variables and procedures
A4	Data Collection & Analysis	Systematically gathering, organizing, and analyzing experimental data
A5	Evidence-Based Reasoning	Drawing conclusions supported by evidence and logical reasoning
A6	Scientific Communication	Presenting findings clearly using appropriate scientific conventions and formats

---

B. Mathematical Thinking (8 Competencies)

ID	Competency	Description
B1	Number Sense & Operations	Understanding number systems, magnitudes, and arithmetic operations
B2	Algebraic Reasoning	Using variables, expressions, and equations to represent relationships
B3	Statistical Reasoning	Collecting, summarizing, and interpreting data using statistical measures
B4	Probabilistic Reasoning	Understanding chance, randomness, and probability distributions
B5	Geometric & Spatial Reasoning	Analyzing shapes, transformations, and spatial relationships
B6	Logical Reasoning & Proof	Constructing and evaluating logical arguments and formal proofs
B7	Mathematical Modeling	Translating real-world problems into mathematical representations
B8	Linear Algebra & Matrix Operations	Understanding vectors, matrices, transformations, and linear systems

---

C. Computational Thinking (8 Competencies)

ID	Competency	Description
C1	Decomposition	Breaking complex problems into smaller, manageable sub-problems
C2	Pattern Recognition	Identifying regularities, similarities, and trends in data or processes
C3	Abstraction	Identifying essential information and ignoring irrelevant details
C4	Algorithmic Design	Designing step-by-step procedures to solve problems efficiently
C5	Logical Thinking	Applying Boolean logic, conditionals, and systematic reasoning
C6	Evaluation & Optimization	Assessing solution efficiency and improving performance
C7	Generalization & Transfer	Applying solutions from one context to new, similar contexts
C8	Debugging & Troubleshooting	Systematically identifying and resolving errors in processes or code

---

D. Digital Technology (8 Competencies)

ID	Competency	Description
D1	Digital Literacy	Using digital tools and platforms effectively for learning and work
D2	Information Management	Searching, evaluating, organizing, and curating digital information
D3	Computer Systems & Architecture	Understanding hardware components, operating systems, and networks
D4	Cloud Computing & Services	Using cloud platforms, storage, and computing resources
D5	Cybersecurity Fundamentals	Understanding threats, protections, and secure practices
D6	Networking & Internet Protocols	Understanding how data moves through networks and the internet
D7	Digital Media & Content Creation	Creating, editing, and publishing digital media content
D8	Human-Computer Interaction	Designing and evaluating user interfaces and user experiences

---

E. Programming & Software Creation (9 Competencies)

ID	Competency	Description
E1	Programming Fundamentals	Understanding variables, data types, control flow, and basic I/O
E2	Data Structures	Using arrays, lists, dictionaries, trees, graphs, and hash maps
E3	Functions & Modular Design	Writing reusable functions, modules, and clean interfaces
E4	Object-Oriented Programming	Designing with classes, inheritance, polymorphism, and encapsulation
E5	Algorithm Implementation	Implementing sorting, searching, graph, and dynamic programming algorithms
E6	Software Development Practices	Using version control, testing, code review, and CI/CD pipelines
E7	Web Development	Building front-end and back-end web applications
E8	Database Design & SQL	Designing schemas, writing queries, and managing relational databases
E9	API Design & Integration	Creating and consuming RESTful and GraphQL APIs

---

F. Data Literacy & Data Science (9 Competencies)

ID	Competency	Description
F1	Data Awareness & Representation	Understanding what data is, types of data, and how it is represented
F2	Data Collection & Preparation	Gathering, cleaning, transforming, and validating datasets
F3	Exploratory Data Analysis	Summarizing data characteristics through statistics and visualization
F4	Data Visualization & Storytelling	Creating effective charts, dashboards, and data narratives
F5	Statistical Inference	Drawing conclusions from samples using hypothesis tests and confidence intervals
F6	Predictive Modeling	Building and evaluating models that predict future outcomes
F7	Feature Engineering	Creating, selecting, and transforming variables to improve model performance
F8	Big Data & Data Pipelines	Processing large-scale data using distributed computing and ETL pipelines
F9	Data Governance & Quality	Ensuring data accuracy, consistency, lineage, and compliance

---

G. Artificial Intelligence (11 Competencies)

ID	Competency	Description
G1	AI Awareness & Concepts	Understanding what AI is, its types, capabilities, and limitations
G2	Machine Learning Fundamentals	Understanding supervised, unsupervised, and reinforcement learning paradigms
G3	Training & Optimization	Training models using loss functions, gradient descent, and hyperparameter tuning
G4	Model Evaluation & Validation	Assessing model performance using metrics, cross-validation, and error analysis
G5	Deep Learning & Neural Networks	Understanding architectures: CNNs, RNNs, Transformers, and their applications
G6	Natural Language Processing	Processing and understanding human language with AI techniques
G7	Computer Vision	Analyzing and interpreting visual data using AI models
G8	Generative AI & LLMs	Understanding and using generative models, prompt engineering, and LLM applications
G9	AI System Design & Deployment	Designing end-to-end AI systems, MLOps, and production deployment
G10	Human-AI Interaction	Designing effective interactions between humans and AI systems
G11	AI Safety & Alignment	Ensuring AI systems are safe, reliable, and aligned with human values

---

H. Engineering & Making (9 Competencies)

ID	Competency	Description
H1	Design Thinking	Applying empathy, ideation, prototyping, and iteration to solve problems
H2	Prototyping & Fabrication	Building physical and digital prototypes using various tools and materials
H3	Electronics & Circuits	Understanding circuits, sensors, actuators, and electronic components
H4	Robotics & Automation	Designing, building, and programming robots and automated systems
H5	IoT & Embedded Systems	Connecting devices, sensors, and microcontrollers to networks
H6	3D Modeling & Printing	Creating 3D designs and manufacturing physical objects
H7	Systems Engineering	Managing complexity in multi-component engineered systems
H8	Project Management	Planning, executing, and delivering engineering projects on time and budget
H9	Quality Assurance & Testing	Ensuring products meet specifications through systematic testing

---

I. Responsible Technology (9 Competencies)

ID	Competency	Description
I1	Digital Citizenship	Behaving responsibly, ethically, and safely in digital environments
I2	Data Privacy & Protection	Understanding and applying data privacy principles and regulations
I3	AI Ethics & Fairness	Identifying and mitigating bias, ensuring fairness in AI systems
I4	Intellectual Property	Understanding copyright, licensing, and attribution in digital contexts
I5	Environmental Impact	Assessing and reducing the environmental footprint of technology
I6	Accessibility & Inclusion	Designing technology that is usable by people of all abilities
I7	Digital Well-being	Managing screen time, mental health, and healthy technology habits
I8	Regulatory Compliance	Understanding and adhering to technology regulations and standards
I9	Ethical Decision-Making	Applying ethical frameworks to technology decisions and trade-offs

---

J. Technology Collaboration & Innovation (9 Competencies)

ID	Competency	Description
J1	Team Collaboration	Working effectively in diverse teams using collaborative tools and practices
J2	Communication & Presentation	Presenting technical ideas clearly to both technical and non-technical audiences
J3	Creative Problem-Solving	Generating novel solutions through divergent thinking and innovation methods
J4	Entrepreneurial Thinking	Identifying opportunities, evaluating feasibility, and creating value
J5	Cross-Disciplinary Integration	Connecting knowledge and methods from multiple disciplines
J6	Mentoring & Knowledge Sharing	Teaching, guiding, and sharing expertise with peers and juniors
J7	Community Building	Contributing to open-source, professional, and learning communities
J8	Continuous Learning	Maintaining and updating skills through self-directed learning
J9	Innovation Management	Organizing, evaluating, and scaling innovative ideas and processes

---

4. Skill Structure

4.1 Skill Data Model

Each skill in the Knowledge Graph is a richly structured node containing identification, classification, proficiency definitions, prerequisites, and observable indicators.

4.2 Full Skill Example

{
  "id": "skill_gradient_descent",
  "name": "Gradient Descent",
  "domain": "STEM & AI Education",
  "competencyArea": {
    "id": "G",
    "name": "Artificial Intelligence"
  },
  "competency": {
    "id": "G3",
    "name": "Training & Optimization"
  },
  "definition": "Understanding and applying gradient descent optimization to iteratively minimize a loss function by updating model parameters in the direction of steepest descent.",
  "version": "1.0",
  "status": "published",
  "externalReferences": [
    {
      "framework": "UNESCO_AI_Students",
      "dimension": "AI Foundations",
      "level": "Create"
    },
    {
      "framework": "SFIA_9",
      "skill": "Machine Learning",
      "level": 3
    }
  ],
  "proficiencyLevels": [
    {
      "level": "L1",
      "name": "Recognize",
      "description": "Can identify gradient descent as an optimization technique and recognize when it is being used.",
      "indicators": [
        "Identifies gradient descent in a list of optimization methods",
        "Recognizes the purpose of a learning rate parameter",
        "Matches gradient descent to the concept of minimizing a function"
      ]
    },
    {
      "level": "L2",
      "name": "Explain",
      "description": "Can explain how gradient descent works, including the role of gradients, learning rate, and convergence.",
      "indicators": [
        "Explains the intuition behind following the negative gradient",
        "Describes the effect of learning rate on convergence speed and stability",
        "Differentiates between batch, stochastic, and mini-batch gradient descent",
        "Explains what convergence means and how to detect it"
      ]
    },
    {
      "level": "L3",
      "name": "Apply",
      "description": "Can implement gradient descent from scratch and apply it to train simple models.",
      "indicators": [
        "Implements vanilla gradient descent in Python for linear regression",
        "Selects appropriate learning rate through experimentation",
        "Plots and interprets the loss curve during training",
        "Applies gradient descent to logistic regression"
      ]
    },
    {
      "level": "L4",
      "name": "Diagnose",
      "description": "Can diagnose training issues related to gradient descent and select appropriate variants or fixes.",
      "indicators": [
        "Diagnoses vanishing or exploding gradient problems",
        "Selects between SGD, Adam, RMSProp based on problem characteristics",
        "Implements learning rate scheduling strategies",
        "Analyzes convergence behavior and adjusts hyperparameters accordingly"
      ]
    },
    {
      "level": "L5",
      "name": "Design",
      "description": "Can design custom optimization strategies and contribute to novel optimization research.",
      "indicators": [
        "Designs custom learning rate schedules for specific architectures",
        "Implements and benchmarks novel optimization algorithms",
        "Writes technical documentation explaining optimization design decisions",
        "Mentors others on optimization best practices"
      ]
    }
  ],
  "prerequisites": [
    {
      "skillId": "skill_derivatives_partial",
      "relationshipType": "HARD_PREREQUISITE",
      "requiredLevel": "L3",
      "rationale": "Gradient descent requires computing partial derivatives of the loss function"
    },
    {
      "skillId": "skill_linear_algebra_matrix_ops",
      "relationshipType": "HARD_PREREQUISITE",
      "requiredLevel": "L2",
      "rationale": "Parameter updates involve matrix operations in multi-dimensional spaces"
    },
    {
      "skillId": "skill_loss_functions",
      "relationshipType": "CONCEPTUAL_FOUNDATION",
      "requiredLevel": "L2",
      "rationale": "Understanding what is being minimized is essential context"
    },
    {
      "skillId": "skill_python_programming",
      "relationshipType": "PROCEDURAL_FOUNDATION",
      "requiredLevel": "L3",
      "rationale": "Implementation requires Python proficiency including NumPy"
    }
  ],
  "indicators": [
    {
      "id": "ind_gd_001",
      "level": "L1",
      "text": "Identifies gradient descent as an optimization technique",
      "assessmentMethods": ["multiple_choice", "matching"],
      "evidenceTypes": ["quiz_response"]
    },
    {
      "id": "ind_gd_002",
      "level": "L2",
      "text": "Explains the role of learning rate in convergence",
      "assessmentMethods": ["short_answer", "oral_explanation"],
      "evidenceTypes": ["written_response", "video_recording"]
    },
    {
      "id": "ind_gd_003",
      "level": "L3",
      "text": "Implements gradient descent for linear regression",
      "assessmentMethods": ["coding_task", "code_review"],
      "evidenceTypes": ["code_submission", "unit_test_results"]
    },
    {
      "id": "ind_gd_004",
      "level": "L4",
      "text": "Diagnoses and fixes vanishing gradient problems",
      "assessmentMethods": ["debugging_task", "case_study"],
      "evidenceTypes": ["code_submission", "written_analysis"]
    },
    {
      "id": "ind_gd_005",
      "level": "L5",
      "text": "Designs custom optimization strategy for a novel architecture",
      "assessmentMethods": ["project", "technical_report"],
      "evidenceTypes": ["project_deliverable", "technical_document"]
    }
  ],
  "metadata": {
    "createdAt": "2026-07-01T00:00:00Z",
    "updatedAt": "2026-07-15T00:00:00Z",
    "createdBy": "curriculum_team",
    "tags": ["optimization", "machine-learning", "deep-learning", "calculus"]
  }
}

4.3 Skill Schema Summary

Field	Type	Description
id	string	Unique identifier (e.g., skill_gradient_descent)
name	string	Human-readable skill name
domain	string	Top-level domain classification
competencyArea	object	Parent competency area reference
competency	object	Parent competency reference
definition	string	Clear, concise definition of what the skill entails
proficiencyLevels	array	5 levels (L1-L5) with descriptions and indicators
prerequisites	array	List of prerequisite skills with relationship types
indicators	array	Observable behaviors with assessment methods
externalReferences	array	Mappings to UNESCO, ISTE, SFIA, WEF
metadata	object	Timestamps, authorship, tags

---

5. Skill Dependency Graph

5.1 Dependency Types

The Knowledge Graph supports 7 types of dependencies between skills, each with distinct semantics for diagnosis and remediation:

#	Dependency Type	Semantics	Strength	Example
1	HARD_PREREQUISITE	Cannot learn target without source; fundamental blocker	0.9 - 1.0	Partial Derivatives → Gradient Descent
2	SOFT_PREREQUISITE	Easier with source but not mandatory; alternative paths exist	0.5 - 0.8	Linear Algebra → PCA (can learn PCA intuitively without full LA)
3	CONCEPTUAL_FOUNDATION	Source provides the mental model needed to understand target	0.6 - 0.9	Probability → Bayesian Inference
4	PROCEDURAL_FOUNDATION	Source provides procedural skills needed to implement target	0.5 - 0.8	Python Programming → Implement Neural Network
5	TRANSFER_SUPPORT	Source helps transfer learning to a new context	0.3 - 0.6	Statistics → A/B Testing in Product Management
6	CO_REQUISITE	Should be learned in parallel; mutual reinforcement	0.4 - 0.7	Linear Algebra ↔ Multivariable Calculus
7	COMMON_MISCONCEPTION_SOURCE	Lack of source causes specific misconceptions in target	0.6 - 0.9	Correlation vs. Causation → Interpreting ML Feature Importance

5.2 Dependency Edge Structure

Each dependency edge is a richly annotated connection in the Knowledge Graph:

{
  "dependencyId": "dep_partial_deriv_to_grad_desc",
  "sourceSkillId": "skill_derivatives_partial",
  "targetSkillId": "skill_gradient_descent",
  "relationshipType": "HARD_PREREQUISITE",
  "requiredSourceLevel": "L3",
  "strength": 0.95,
  "confidence": 0.90,
  "rationale": "Gradient descent computes the gradient (vector of partial derivatives) of the loss function. Without the ability to compute partial derivatives (L3 Apply), a learner cannot understand or implement the parameter update step.",
  "diagnosticSignals": [
    {
      "signal": "Learner cannot explain why parameters move in the negative gradient direction",
      "indicatesGap": true,
      "gapLikelihood": 0.85
    },
    {
      "signal": "Learner confuses gradient with loss value",
      "indicatesGap": true,
      "gapLikelihood": 0.90
    },
    {
      "signal": "Learner cannot compute the gradient for a simple quadratic function",
      "indicatesGap": true,
      "gapLikelihood": 0.95
    }
  ],
  "remediationCandidates": [
    {
      "skillId": "skill_derivatives_partial",
      "targetLevel": "L3",
      "estimatedEffort": "4 hours",
      "suggestedActivities": [
        "Practice computing partial derivatives of multivariable functions",
        "Implement symbolic differentiation in Python",
        "Complete gradient computation exercises"
      ]
    }
  ],
  "metadata": {
    "source": "expert_defined",
    "definedBy": "Dr. Smith, Curriculum Expert",
    "validatedAt": "2026-06-15T00:00:00Z",
    "evidenceCount": 0,
    "lastReviewedAt": "2026-07-01T00:00:00Z"
  }
}

5.3 Sources of Dependencies

Dependencies in the Knowledge Graph originate from three sources, each with different confidence levels and validation requirements:

Source 1: Expert-Defined

• Origin: Curriculum experts, domain specialists, and instructional designers
• Process: Manual definition based on pedagogical expertise and subject-matter knowledge
• Confidence: High (0.8-1.0) — based on expert consensus
• Validation: Peer review by at least 2 domain experts

Source 2: Curriculum-Derived

• Origin: Extracted from existing learning sequences, course structures, and textbook orderings
• Process: Automated analysis of curriculum materials to identify implied prerequisites
• Confidence: Medium (0.5-0.8) — curriculum order may reflect convention rather than true dependency
• Validation: Expert review to confirm pedagogical validity

Source 3: Evidence-Derived

• Origin: Learned from learner performance data — patterns of failure and success
• Process: Statistical analysis of learner trajectories to discover hidden dependencies
• Confidence: Variable (0.3-0.9) — depends on data volume and statistical significance
• Validation Workflow:

stateDiagram-v2
    [*] --> Detected: Statistical pattern found
    Detected --> Reviewed: Expert reviews evidence
    Reviewed --> Validated: Expert confirms dependency
    Reviewed --> Rejected: Expert rejects (spurious correlation)
    Validated --> Published: Added to production graph
    Rejected --> [*]
    Published --> [*]

Status	Description	Visibility
Detected	Statistical pattern found in learner data; awaiting expert review	Internal only
Reviewed	Expert has examined the evidence and supporting data	Internal only
Validated	Expert confirms the dependency is pedagogically valid	Available for staging
Rejected	Expert determines the pattern is spurious or coincidental	Archived
Published	Dependency is active in the production Knowledge Graph	Available for diagnosis

5.4 Example Dependency Graphs

Machine Learning Branch

Number Sense (B1)
└── Ratio & Proportion (B1)
    └── Algebraic Reasoning (B2)
        ├── Functions & Equations (B2)
        │   └── Derivatives & Calculus (B2)
        │       └── Partial Derivatives (B2)
        │           └── Gradient Descent (G3)
        │               └── Training Neural Networks (G5)
        │                   └── Model Evaluation (G4)
        │                       └── ML Workflow (G2)
        └── Statistical Reasoning (B3)
            └── Probability (B4)
                ├── Probability Distributions (B4)
                │   └── Bayesian Inference (B4)
                │       └── Naive Bayes Classifier (G2)
                └── Hypothesis Testing (F5)
                    └── Model Evaluation (G4)

graph TD
    NS["Number Sense<br/>(B1)"] --> RP["Ratio & Proportion<br/>(B1)"]
    RP --> AR["Algebraic Reasoning<br/>(B2)"]
    AR --> FE["Functions & Equations<br/>(B2)"]
    AR --> SR["Statistical Reasoning<br/>(B3)"]
    FE --> DC["Derivatives & Calculus<br/>(B2)"]
    DC --> PD["Partial Derivatives<br/>(B2)"]
    PD --> GD["Gradient Descent<br/>(G3)"]
    GD --> TNN["Training Neural Networks<br/>(G5)"]
    TNN --> ME["Model Evaluation<br/>(G4)"]
    ME --> MLW["ML Workflow<br/>(G2)"]
    SR --> PROB["Probability<br/>(B4)"]
    PROB --> PBD["Probability Distributions<br/>(B4)"]
    PBD --> BI["Bayesian Inference<br/>(B4)"]
    BI --> NBC["Naive Bayes Classifier<br/>(G2)"]
    PROB --> HT["Hypothesis Testing<br/>(F5)"]
    HT --> ME

Neural Networks Branch

Linear Algebra (B8)
├── Matrix Multiplication (B8)
│   └── Forward Propagation (G5)
│       └── Activation Functions (G5)
│           └── Backpropagation (G5)
│               └── Training Deep Networks (G5)
└── Vector Spaces (B8)
    └── Embeddings & Representations (G6)
        └── Word2Vec / Transformers (G6)

Partial Derivatives (B2)
└── Chain Rule (B2)
    └── Backpropagation (G5)

---

6. Proficiency Levels

6.1 Learning Proficiency (for Learners)

The EduOne framework defines 6 learning proficiency levels (L0-L5) that describe a learner's mastery progression for any skill:

Level	Name	Description	Cognitive Level	Observable Behaviors	Example (Gradient Descent)
L0	Unexposed	Has not encountered the skill; no awareness	None	Cannot recognize the concept	Has never heard of gradient descent
L1	Recognize	Can identify and recall the concept but cannot explain it	Remember	Selects correct definition from options; matches terms to descriptions	Identifies gradient descent in a list of optimization methods
L2	Explain	Can explain the concept in own words and compare with alternatives	Understand	Explains how it works; compares variants; draws diagrams	Explains why learning rate matters; differentiates SGD from Adam
L3	Apply	Can use the skill to solve standard problems	Apply	Implements the skill in standard contexts; follows established procedures	Implements gradient descent for linear regression in Python
L4	Analyze / Adapt	Can diagnose issues, adapt the skill to non-standard situations	Analyze	Debugs failures; selects best variant for context; modifies approach	Diagnoses vanishing gradients; selects appropriate optimizer
L5	Create / Transfer	Can create novel solutions and transfer the skill to entirely new domains	Create	Designs new approaches; teaches others; publishes work	Designs custom optimization strategy; writes research paper

graph LR
    L0["L0<br/>Unexposed"] --> L1["L1<br/>Recognize"]
    L1 --> L2["L2<br/>Explain"]
    L2 --> L3["L3<br/>Apply"]
    L3 --> L4["L4<br/>Analyze/Adapt"]
    L4 --> L5["L5<br/>Create/Transfer"]

    style L0 fill:#f5f5f5,stroke:#999
    style L1 fill:#e3f2fd,stroke:#1976d2
    style L2 fill:#e8f5e9,stroke:#388e3c
    style L3 fill:#fff3e0,stroke:#f57c00
    style L4 fill:#fce4ec,stroke:#c62828
    style L5 fill:#f3e5f5,stroke:#7b1fa2

6.2 Professional Responsibility (SFIA-Referenced)

For professional development and higher education contexts, EduOne maps skills to SFIA 9 responsibility levels:

Level	Name	Autonomy	Influence Scope	Key Characteristics
1	Follow	Works under close supervision	Own work	Applies basic concepts; follows instructions; learns on the job
2	Assist	Works under routine direction	Immediate colleagues	Performs defined tasks; contributes to team; developing judgment
3	Apply	Works under general guidance	Project/team	Applies skills independently; makes routine decisions; mentors juniors
4	Enable	Works with broad direction	Function/department	Enables others; manages complex work; makes significant decisions
5	Ensure / Advise	Accountable for defined areas	Organizational unit	Sets direction; ensures quality; advises senior management
6	Initiate / Influence	Defines strategy	Organization-wide	Initiates major change; influences industry; shapes organizational direction
7	Set Strategy	Leads at the highest level	Industry/domain	Sets industry standards; shapes policy; recognized thought leader

IMPORTANT

Learning Proficiency (L0-L5) and Professional Responsibility (1-7) are orthogonal dimensions. A learner can be at L3 (Apply) in gradient descent while operating at SFIA Level 2 (Assist) in their professional role. Both dimensions are tracked independently.

---

7. Root-Cause Diagnosis

7.1 Algorithm Flow

The root-cause diagnosis algorithm traverses the Knowledge Graph to find the earliest unmastered foundation that explains a learner's failure. The algorithm follows 8 steps:

Step	Action	Description
1	Observed Failure	A learner fails an assessment or exhibits poor performance on a task
2	Identify Target Indicator	Map the failure to a specific indicator (observable behavior) in the framework
3	Locate Target Skill	Identify the skill and proficiency level the indicator belongs to
4	Inspect Prerequisites	Retrieve all prerequisite edges for the target skill from the Knowledge Graph
5	Check Evidence	For each prerequisite, check the learner's evidence and proficiency level
6	Traverse Weak Prerequisites Recursively	For any prerequisite below the required level, recursively inspect its prerequisites
7	Find Earliest Unmastered Foundation	Identify the deepest node(s) in the traversal where the learner's proficiency is below required
8	Recommend Minimum Remediation Path	Generate the shortest path from the root cause to the target skill

flowchart TD
    S1["1. Observed Failure<br/><i>Learner fails Model Evaluation quiz</i>"]
    S2["2. Identify Target Indicator<br/><i>Cannot compute precision/recall</i>"]
    S3["3. Locate Target Skill<br/><i>Model Evaluation (G4) @ L3</i>"]
    S4["4. Inspect Prerequisites<br/><i>Probability, Statistics, ML Fundamentals</i>"]
    S5["5. Check Evidence<br/><i>Probability: L1, Statistics: L2, ML: L2</i>"]
    S6["6. Traverse Weak Prerequisites<br/><i>Probability is below L3 → check its prereqs</i>"]
    S7["7. Find Earliest Gap<br/><i>Ratio & Proportion: L1 (root cause)</i>"]
    S8["8. Recommend Remediation<br/><i>Ratio → Probability → Model Evaluation</i>"]

    S1 --> S2 --> S3 --> S4 --> S5 --> S6 --> S7 --> S8

7.2 Root Cause Score Formula

When multiple prerequisite gaps exist, the system ranks them using a Root Cause Score to prioritize the most impactful remediation:

Root Cause Score = Dependency Strength × Gap Probability × Evidence Confidence × Error Match × Remediation Leverage

Factor	Symbol	Description	Range
Dependency Strength	DS	How strong is the prerequisite relationship?	0.0 - 1.0
Gap Probability	GP	How likely is it that the learner has this gap?	0.0 - 1.0
Evidence Confidence	EC	How confident are we in the gap assessment?	0.0 - 1.0
Error Match	EM	How well does the gap explain the observed error pattern?	0.0 - 1.0
Remediation Leverage	RL	How many downstream skills will benefit from fixing this gap?	0.0 - 1.0

Scoring Example

A learner fails Model Evaluation (G4). The system identifies 4 candidate gaps:

Candidate Gap	DS	GP	EC	EM	RL	Score	Rank
Probability (B4)	0.90	0.85	0.80	0.75	0.90	0.413	1
Ratio & Proportion (B1)	0.95	0.90	0.70	0.60	0.95	0.342	2
Statistical Reasoning (B3)	0.85	0.70	0.90	0.80	0.70	0.300	3
Python Programming (E1)	0.50	0.40	0.95	0.30	0.50	0.029	4

TIP

The highest-scoring gap (Probability) is recommended as the primary remediation target. The system also checks whether fixing Probability would automatically resolve lower-ranked gaps (e.g., if Ratio & Proportion is a prerequisite of Probability).

7.3 Multiple Case Analysis

The power of root-cause diagnosis is that the same observed failure can have different root causes for different learners. Here are 4 cases where learners all fail Model Evaluation (G4) but for different reasons:

Case A: Mathematical Foundation Gap

Attribute	Value
Learner	Student A — Grade 10
Observed Failure	Cannot compute F1-score; confuses precision and recall
Target Skill	Model Evaluation (G4) @ L3
Root Cause	Ratio & Proportion (B1) @ L1 (required: L3)
Diagnostic Signal	Cannot compute ratios from a confusion matrix; treats precision as a count
Remediation	Ratio & Proportion → Fractions & Percentages → Probability → Model Evaluation
Estimated Effort	6 hours

Case B: Probabilistic Reasoning Gap

Attribute	Value
Learner	Student B — University Year 1
Observed Failure	Computes accuracy but cannot interpret it for imbalanced datasets
Target Skill	Model Evaluation (G4) @ L4
Root Cause	Probability Distributions (B4) @ L2 (required: L3)
Diagnostic Signal	Does not understand base rates; cannot explain why 95% accuracy on 95/5 class split is meaningless
Remediation	Probability Distributions → Class Imbalance → Model Evaluation
Estimated Effort	4 hours

Case C: Conceptual Confusion

Attribute	Value
Learner	Student C — Bootcamp participant
Observed Failure	Uses training accuracy to evaluate model; reports 99% accuracy on overfitted model
Target Skill	Model Evaluation (G4) @ L3
Root Cause	Overfitting Concepts (G3) @ L1 (required: L2)
Diagnostic Signal	Does not understand train/test split purpose; no concept of generalization
Remediation	Overfitting Concepts → Cross-Validation → Model Evaluation
Estimated Effort	3 hours

Case D: Procedural Gap

Attribute	Value
Learner	Student D — Self-taught developer
Observed Failure	Understands metrics conceptually but cannot implement evaluation pipeline
Target Skill	Model Evaluation (G4) @ L3
Root Cause	Python Data Manipulation (E1) @ L2 (required: L3)
Diagnostic Signal	Cannot use sklearn.metrics; struggles with DataFrame operations for confusion matrix
Remediation	Python Data Manipulation → sklearn API → Model Evaluation
Estimated Effort	5 hours

---

8. End-to-End Example

Scenario: Junior AI Engineer — Overfitting Failure

8.1 Context

Learner Profile:

• Name: Minh (Junior AI Engineer, 6 months experience)
• Current Role: ML Engineer Intern
• Target: Complete the "Build a Classification Model" project
• Problem: Minh's model achieves 98% training accuracy but only 52% test accuracy

8.2 Observed Failure

Minh submits a classification model for the project assessment. The evaluation system detects:

Metric	Training Set	Test Set	Status
Accuracy	98.2%	52.1%	❌ Severe overfitting
F1-Score	0.97	0.48	❌ Below threshold
Loss Curve	Continuously decreasing	Increases after epoch 5	❌ Classic overfitting pattern

The system triggers root-cause diagnosis.

8.3 Dependency Trace

graph TD
    FAIL["❌ FAILURE<br/>Model Evaluation (G4) @ L3<br/><i>98% train / 52% test</i>"]
    
    FAIL --> P1["Overfitting & Regularization (G3)<br/>Required: L3 | Actual: L1 ⚠️"]
    FAIL --> P2["Cross-Validation (G4)<br/>Required: L3 | Actual: L2 ⚠️"]
    FAIL --> P3["Loss Functions (G3)<br/>Required: L2 | Actual: L2 ✅"]
    
    P1 --> P1A["Bias-Variance Tradeoff (G2)<br/>Required: L2 | Actual: L0 🔴"]
    P1 --> P1B["Regularization Techniques (G3)<br/>Required: L2 | Actual: L0 🔴"]
    
    P2 --> P2A["Statistical Sampling (B3)<br/>Required: L2 | Actual: L2 ✅"]
    
    P1A --> P1A1["Statistical Reasoning (B3)<br/>Required: L2 | Actual: L2 ✅"]
    P1A --> P1A2["Model Complexity (G2)<br/>Required: L2 | Actual: L1 ⚠️"]

    style FAIL fill:#ffcdd2,stroke:#c62828
    style P1 fill:#fff3e0,stroke:#f57c00
    style P2 fill:#fff3e0,stroke:#f57c00
    style P3 fill:#e8f5e9,stroke:#388e3c
    style P1A fill:#ffcdd2,stroke:#c62828
    style P1B fill:#ffcdd2,stroke:#c62828
    style P2A fill:#e8f5e9,stroke:#388e3c
    style P1A1 fill:#e8f5e9,stroke:#388e3c
    style P1A2 fill:#fff3e0,stroke:#f57c00

8.4 Diagnosis Result

Priority	Root Cause	Current Level	Required Level	Score
🔴 1	Bias-Variance Tradeoff (G2)	L0 Unexposed	L2 Explain	0.72
🔴 2	Regularization Techniques (G3)	L0 Unexposed	L2 Explain	0.68
⚠️ 3	Model Complexity (G2)	L1 Recognize	L2 Explain	0.45
⚠️ 4	Cross-Validation (G4)	L2 Explain	L3 Apply	0.38

Primary Root Cause: Minh has never been exposed to the bias-variance tradeoff. Without understanding this fundamental concept, he cannot diagnose overfitting, does not know why regularization is needed, and cannot evaluate model complexity appropriately.

8.5 Remediation Path

The system generates a minimum remediation path — the shortest sequence of skills to close the gap:

graph LR
    R1["Step 1<br/>Model Complexity<br/>(G2) → L2<br/><i>2 hours</i>"]
    R2["Step 2<br/>Bias-Variance<br/>Tradeoff (G2) → L2<br/><i>3 hours</i>"]
    R3["Step 3<br/>Regularization<br/>Techniques (G3) → L2<br/><i>3 hours</i>"]
    R4["Step 4<br/>Cross-Validation<br/>(G4) → L3<br/><i>2 hours</i>"]
    R5["Step 5<br/>Retry Project<br/>Assessment<br/><i>1 hour</i>"]

    R1 --> R2 --> R3 --> R4 --> R5

Recommended Learning Activities:

Step	Skill	Target Level	Activities	Estimated Time
1	Model Complexity (G2)	L2 Explain	Read "Underfitting vs. Overfitting" tutorial; complete concept quiz	2 hours
2	Bias-Variance Tradeoff (G2)	L2 Explain	Interactive visualization of bias-variance; explain in own words	3 hours
3	Regularization Techniques (G3)	L2 Explain	L1/L2 regularization tutorial; apply dropout in code exercise	3 hours
4	Cross-Validation (G4)	L3 Apply	Implement k-fold CV on the project dataset; compare results	2 hours
5	Retry Assessment	L3 Apply	Resubmit classification model with regularization and CV	1 hour

Total Estimated Remediation Time: 11 hours

8.6 Mastery Update

After completing the remediation path, the system updates Minh's skill profile:

Skill	Before	After	Change
Model Complexity (G2)	L1 Recognize	L2 Explain	+1
Bias-Variance Tradeoff (G2)	L0 Unexposed	L2 Explain	+2
Regularization Techniques (G3)	L0 Unexposed	L2 Explain	+2
Cross-Validation (G4)	L2 Explain	L3 Apply	+1
Model Evaluation (G4)	L2 (failing)	L3 Apply	+1

Minh resubmits the project and achieves:

Metric	Training Set	Test Set	Status
Accuracy	87.5%	84.2%	✅ Healthy gap
F1-Score	0.86	0.83	✅ Above threshold
Loss Curve	Converges	Converges	✅ No overfitting

---

9. Hackathon MVP Scope

9.1 Focus Area

The Hackathon MVP focuses on a single vertical slice of the full framework to demonstrate the end-to-end value proposition:

Dimension	MVP Scope	Full Scope
Domain	STEM & AI Education	Multiple domains
Competency Area	G. Artificial Intelligence (AI Literacy subset)	All 10 areas
Competencies	5 competencies (G1-G5)	86 competencies
Skills	15-20 skills	300+ skills
Indicators	30-40 indicators	500+ indicators
Dependency Edges	25-40 edges	1000+ edges
Proficiency Levels	L0-L5 (Learning Proficiency only)	Learning + Professional

9.2 MVP Competency Slice

The MVP covers the following 5 competencies from the AI area:

graph TD
    G["G. Artificial Intelligence"]
    G1["G1. AI Awareness<br/>& Concepts"]
    G2["G2. ML<br/>Fundamentals"]
    G3["G3. Training<br/>& Optimization"]
    G4["G4. Model Evaluation<br/>& Validation"]
    G5["G5. Deep Learning<br/>& Neural Networks"]

    G --> G1
    G --> G2
    G --> G3
    G --> G4
    G --> G5

    G1 -->|"prereq"| G2
    G2 -->|"prereq"| G3
    G2 -->|"prereq"| G4
    G3 -->|"prereq"| G5
    G4 -->|"co-req"| G3

9.3 MVP Demo Loop

The hackathon demo follows a 6-step loop that demonstrates the full diagnostic cycle:

flowchart LR
    D1["1. Failure<br/><i>Learner fails<br/>assessment</i>"]
    D2["2. Model Update<br/><i>Update learner's<br/>skill profile</i>"]
    D3["3. Graph Traversal<br/><i>Walk dependency<br/>edges</i>"]
    D4["4. Gap Detection<br/><i>Find root cause<br/>skills</i>"]
    D5["5. Remediation<br/><i>Generate learning<br/>path</i>"]
    D6["6. Mastery Update<br/><i>Update after<br/>completion</i>"]

    D1 --> D2 --> D3 --> D4 --> D5 --> D6
    D6 -.->|"retry"| D1

Demo Scenario: Use the Minh (Junior AI Engineer) scenario from Section 8 as the primary demo case.

9.4 MVP Technical Requirements

Component	Technology	Status
Competency data store	Supabase (PostgreSQL)	To be implemented
Knowledge Graph storage	PostgreSQL with adjacency list + recursive CTEs	To be implemented
Diagnosis engine	Python service (FastAPI)	To be implemented
Graph visualization	D3.js or Mermaid (read-only)	To be implemented
Learner skill profile	JSON document in Supabase	To be implemented
API layer	REST API via FastAPI	To be implemented

---

10. Functional Requirements

10.1 Requirements Table

ID	Category	Description	Priority	Acceptance Criteria
FR-CF-001	CRUD - Domain	The system shall allow authorized users to create, read, update, and delete domains.	Must	A domain can be created with a unique ID, name, and description. The domain appears in the domain list. Updates are reflected immediately. Deletion cascades to child competency areas or is blocked if children exist.
FR-CF-002	CRUD - Competency Area	The system shall allow authorized users to create, read, update, and delete competency areas within a domain.	Must	A competency area can be created with ID, name, description, and parent domain reference. The area appears under its parent domain. CRUD operations maintain referential integrity.
FR-CF-003	CRUD - Competency	The system shall allow authorized users to create, read, update, and delete competencies within a competency area.	Must	A competency can be created with ID, name, description, and parent competency area reference. Deletion is blocked if child skills exist.
FR-CF-004	CRUD - Skill	The system shall allow authorized users to create, read, update, and delete skills within a competency.	Must	A skill can be created with the full schema defined in Section 4, including proficiency levels, prerequisites, and indicators. All fields are validated against schema constraints.
FR-CF-005	CRUD - Indicator	The system shall allow authorized users to create, read, update, and delete indicators within a skill at a specific proficiency level.	Must	An indicator can be created with ID, text, proficiency level, assessment methods, and evidence types. Indicators are unique within a skill-level combination.
FR-CF-006	Dependency - Create	The system shall allow authorized users to create dependency edges between skills with all required metadata.	Must	A dependency edge can be created with source skill, target skill, relationship type, required source level, strength, confidence, and rationale. The system validates that both skills exist and no circular dependency is created.
FR-CF-007	Dependency - Validate	The system shall validate new dependency edges for circular dependencies, duplicate edges, and consistency.	Must	Creating an edge that would form a cycle is rejected with a clear error message. Duplicate edges (same source, target, type) are rejected. Strength and confidence values must be between 0.0 and 1.0.
FR-CF-008	Dependency - Query Paths	The system shall support querying shortest paths, all paths, and prerequisite chains between any two skills.	Must	Given a source and target skill, the system returns the shortest path (minimum edges) and optionally all paths up to a configurable maximum depth. Paths include all edge metadata.
FR-CF-009	Diagnosis - Root Cause	The system shall implement the root-cause diagnosis algorithm defined in Section 7.1 to identify the earliest unmastered foundation for a learner's failure.	Must	Given a learner ID and a failed indicator, the system traverses the dependency graph and returns a ranked list of root-cause candidates with scores computed per Section 7.2.
FR-CF-010	Diagnosis - Scoring	The system shall compute Root Cause Scores using the formula: Dependency Strength × Gap Probability × Evidence Confidence × Error Match × Remediation Leverage.	Must	Scores are computed correctly for all candidates. Candidates are ranked by score in descending order. Each factor is documented in the response.
FR-CF-011	Remediation - Path Generation	The system shall generate minimum remediation paths from root-cause skills to the target skill.	Must	The remediation path includes ordered skills, target proficiency levels, estimated effort, and suggested activities. The path respects dependency ordering (prerequisites before dependents).
FR-CF-012	Remediation - Activity Suggestion	The system shall suggest specific learning activities for each skill in the remediation path.	Should	Each skill in the remediation path includes at least 2 suggested activities with estimated time. Activities are appropriate for the target proficiency level.
FR-CF-013	Knowledge Evolution - Detection	The system shall detect potential new dependencies from learner performance data using statistical analysis.	Should	When a statistically significant pattern is detected (e.g., 80%+ of learners who fail skill X also have gaps in skill Y), a candidate dependency is created with status "Detected."
FR-CF-014	Knowledge Evolution - Review	The system shall support an expert review workflow for evidence-derived dependencies.	Should	Detected dependencies can be reviewed, validated, or rejected by authorized experts. The review decision and rationale are recorded. Validated dependencies can be published to the production graph.
FR-CF-015	Knowledge Evolution - Publication	The system shall allow validated evidence-derived dependencies to be published to the production Knowledge Graph.	Should	Published dependencies become active in the diagnosis engine. Publication includes a confidence score and evidence count. Published edges are distinguishable from expert-defined edges.
FR-CF-016	Visualization - Graph Display	The system shall provide a visual representation of the Knowledge Graph showing skills as nodes and dependencies as edges.	Should	The visualization displays skills with their proficiency levels, dependency edges with types and strengths, and supports zoom/pan/filter.
FR-CF-017	Visualization - Learner Overlay	The system shall overlay a learner's proficiency levels on the Knowledge Graph visualization.	Should	Each skill node is color-coded by the learner's current proficiency level. Gaps (below required level) are highlighted. The remediation path is visually distinguished.
FR-CF-018	Mapping - External References	The system shall support mapping EduOne skills to external framework references (UNESCO, ISTE, SFIA, WEF).	Should	Each skill can have multiple external references with framework name, dimension/category, and level. External references are queryable (e.g., "show all skills mapped to SFIA Level 3").
FR-CF-019	Mapping - Import/Export	The system shall support importing and exporting the competency framework and Knowledge Graph in JSON format.	Should	The full framework (domains, areas, competencies, skills, indicators, dependencies) can be exported as a single JSON file. The same JSON can be imported to create or update the framework.
FR-CF-020	CRUD - Bulk Operations	The system shall support bulk create, update, and delete operations for skills and dependencies.	Could	Bulk operations accept an array of items and process them in a single transaction. Partial failures are reported with per-item error details.
FR-CF-021	Diagnosis - Multiple Root Causes	The system shall support identifying and ranking multiple concurrent root causes for a single failure.	Must	When multiple prerequisite gaps exist, all are scored and ranked. The system indicates which gaps are independent and which are related (one is a prerequisite of another).
FR-CF-022	Proficiency - Tracking	The system shall track learner proficiency levels for all skills over time.	Must	Each learner has a skill profile with current proficiency levels. Proficiency updates are timestamped and the history is preserved. The system supports querying proficiency at any point in time.
FR-CF-023	Proficiency - Evidence Linking	The system shall link proficiency assessments to specific evidence artifacts.	Should	Each proficiency level change is linked to the evidence that triggered it (quiz result, code submission, project assessment). Evidence artifacts are retrievable from the proficiency history.
FR-CF-024	Search - Skill Discovery	The system shall support searching and filtering skills by name, competency area, tags, and proficiency level.	Should	Full-text search returns relevant skills ranked by relevance. Filters can be combined (e.g., "skills in Competency Area G tagged with 'deep-learning' at L3 or above").
FR-CF-025	Analytics - Gap Analysis	The system shall provide aggregate gap analysis across learner cohorts to identify common weaknesses.	Could	For a given cohort (class, school, program), the system reports the most common skill gaps, most frequent root causes, and recommended curriculum adjustments. Results are exportable as a report.

10.2 Requirements Summary by Priority

Priority	Count	IDs
Must	12	FR-CF-001 through FR-CF-011, FR-CF-021, FR-CF-022
Should	10	FR-CF-012 through FR-CF-019, FR-CF-023, FR-CF-024
Could	3	FR-CF-020, FR-CF-025

10.3 Non-Functional Requirements

ID	Category	Description	Target
NFR-CF-001	Performance	Root-cause diagnosis shall complete within 2 seconds for a graph with up to 500 skills	< 2s response time
NFR-CF-002	Performance	Graph traversal (shortest path) shall complete within 500ms for a graph with up to 1000 edges	< 500ms response time
NFR-CF-003	Scalability	The system shall support up to 10,000 skills and 50,000 dependency edges	10K skills, 50K edges
NFR-CF-004	Availability	The competency framework API shall have 99.5% uptime	99.5% availability
NFR-CF-005	Data Integrity	All CRUD operations shall maintain referential integrity across the hierarchy	Zero orphaned records
NFR-CF-006	Security	Framework modifications shall require authentication and role-based authorization	RBAC enforced

---

Appendix A: Glossary

Term	Definition
Competency	A cluster of related knowledge, skills, and attitudes that enable effective performance
Skill	An atomic, assessable unit of knowledge or ability
Indicator	An observable behavior that demonstrates mastery of a skill at a specific proficiency level
Evidence	An artifact (quiz result, code submission, project) that demonstrates an indicator
Dependency Edge	A directed relationship between two skills indicating prerequisite or supporting relationship
Root Cause	The earliest unmastered foundational skill that explains a learner's failure
Remediation Path	An ordered sequence of skills to learn in order to close a gap
Knowledge Graph	A graph data structure where nodes are skills and edges are typed dependencies
Proficiency Level	A measure of mastery from L0 (Unexposed) to L5 (Create/Transfer)

Appendix B: Document History

Version	Date	Author	Changes
1.0	2026-07-17	EduOne AI Platform Team	Initial release

---

NOTE

This document is a living specification. As the EduOne AI Platform evolves, this document will be updated to reflect new competency areas, dependency types, and diagnostic capabilities. All changes will be tracked in Appendix B.