0% found this document useful (0 votes)
2 views4 pages

Chapter Proposal

The document is a proposal for a book chapter on the use of Agentic AI in personalized learning, focusing on a multi-agent system designed for higher education. It outlines objectives such as creating a scalable framework, a dynamic student profiling engine, and a safe pedagogical interaction model. The chapter aims to contribute to the field by presenting an educational design pattern, cognitive modeling techniques, and an ethical execution framework.

Uploaded by

germancharan
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
2 views4 pages

Chapter Proposal

The document is a proposal for a book chapter on the use of Agentic AI in personalized learning, focusing on a multi-agent system designed for higher education. It outlines objectives such as creating a scalable framework, a dynamic student profiling engine, and a safe pedagogical interaction model. The chapter aims to contribute to the field by presenting an educational design pattern, cognitive modeling techniques, and an ethical execution framework.

Uploaded by

germancharan
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Agentic AI and the Future of Education Series Call for Book Chapters

Book Chapter Proposal


Edited Volume: Agentic AI and the Future of Education
Publisher: Wiley / ISTE (Education 5.0 Series)

Designing Multi-Agent Systems for Personalized Learning:


Cognitive Modeling, Adaptive Curricular Orchestration, and
Academic Mentorship

Target Topic: Personalized Learning Through Agentic AI


Submitter Name: [Your Name]
Contact Email: [Your Email Address]
Submission Date: July 17, 2026

1 Proposal Abstract
In the transition from traditional e-learning to AI-enabled instruction, static adaptive learn-
ing pathways are being replaced by dynamic, self-reflective systems. This chapter proposes a
comprehensive multi-agent architectural framework designed to provide autonomous, person-
alized, and human-centered learning at scale. While previous AI tutoring systems relied on
single-prompt Large Language Models (LLMs) prone to hallucination and linear paths, our
framework leverages a specialized group of collaborating agents—including a Learner Profiler,
Curriculum Adaptor, Virtual Mentor, and Diagnostic Assessor—orchestrated through a cen-
tralized governance layer.

1.1 Objectives
The primary objective of this chapter is to bridge the gap between cognitive load theory and
agentic systems design. Specifically, the chapter aims to:
1. Define a scalable, modular multi-agent blueprint integrated with LTI v1.3 and xAPI
educational standards.
2. Formulate a dynamic student profiling engine utilizing Bayesian state updates to com-
pute cognitive load and concept mastery from telemetry inputs (pauses, error patterns,
reading speed).
3. Establish a safe, hallucination-resistant pedagogical interaction model using Socratic dia-
logue prompts and autonomous retrieval-augmented generation (RAG).
4. Propose actionable guidelines for cost and latency containment using a hybrid local-SLM
and cloud-LLM routing topology.

1.2 Scope
The scope of this work is focused on Higher Education and professional STEM instruction,
where abstract conceptual reasoning, coding exercises, and mathematical problem-solving de-
mand high-fidelity adaptive tutoring. We constrain our system design to environments utilizing

Page 1
Agentic AI and the Future of Education Series Call for Book Chapters

modern LLMs integrated with vector databases (curricula and text repositories) and real-time
interaction logs. The chapter discusses the technical architecture, database schemas, prompt-
chaining patterns, and safety guardrails, rather than focusing purely on hardware or broad
education policy.

1.3 Contribution
The chapter offers three key contributions to the field of Agentic AI in education:

ˆ The Educational Multi-Agent Design Pattern: A formal, reusable architectural de-


sign pattern that separates concerns between pedagogy (Mentor Agent), profiling (Profiler
Agent), content delivery (Curriculum Agent), and grading (Assessor Agent).

ˆ Telemetry-Driven Cognitive Modeling: A mathematical formulation using state es-


timation equations to update student cognitive load (CL ) and topic understanding (UL )
based on behavioral telemetry.

ˆ An Ethical and Safe Execution Framework: A twin-engine safety pipeline featuring


a “Governance Agent” with retry budgets that prevents deadlocks, content hallucinations,
and out-of-scope interactions before they reach the learner.

Page 2
Agentic AI and the Future of Education Series Call for Book Chapters

2 System Architecture & Collaborative Flow


2.1 Orchestration Architecture
The framework divides the educational interaction loop into distinct functional domains, ensur-
ing that grading remains isolated from mentoring, and tutoring outputs are filtered for safety.

Profile DB Vector DB

Profiler Curriculum
Agent Agent

Student LMS / UI Router

Assessor Mentor Governance


Agent Agent Agent

Figure 1: Multi-Agent Educational Orchestration Layout

2.2 Hybrid SLM/LLM Routing Layer


To address latency and token costs, high-frequency grading and telemetry evaluations are routed
to local Small Language Models (SLMs), while Socratic mentoring is delegated to cloud LLMs.

Page 3
Agentic AI and the Future of Education Series Call for Book Chapters

Syntax Validation
Local SLM (8B)
Gaze Logs
High-Speed/Low-Cost
Initial Guardrails

Student Telemetry Router /


& Answer Submission Classifier

Socratic Scaffold
Frontier LLM (32B+)
Cognitive Profiling
High-Cognitive
Path Adaptation

Figure 2: Hybrid SLM/LLM Routing and Cost Containment

3 Proposed Chapter Outline

Section Core Contents and Focus Areas


I. Introduction Shift from legacy adaptive LMS systems to Agentic AI. Limitations
of monolithic tutoring (hallucinations, context window exhaustion,
lack of concerns separation).
II. Theoretical Foundation Structural mapping of Vygotsky’s Zone of Proximal Development
(ZPD) and Sweller’s Cognitive Load Theory to prompt-chain engi-
neering patterns.
III. Multi-Agent Ecosystem Engineering specifications of Router, Learner Profiler, Curriculum
Adaptor, Virtual Mentor, and Diagnostic Assessor agents.
IV. Technical Implementation Prompt templates for Socratic dialogue scaffolding, memory-buffering
strategies, and Textbook vector retrieval (RAG).
V. Safety & Ethics Governance retry budget limits to prevent agent deadlocks, student
data privacy controls, and telemetry shielding against side-channels.
VI. Case Study & Directions Evaluation results of a university computer science pilot study, mul-
timodal gaze-tracking extensions, and conclusion.

Table 1: Proposed Book Chapter Structural Blueprint

Page 4

You might also like