Affective Computing-Based Intelligent Tutoring System for Enhancing Conceptual Understanding in Computer Education Using Machine Learning

Project Abstract

This study presents an affective computing-based intelligent tutoring system (AC-ITS) designed to enhance conceptual understanding in computer education by leveraging machine learning to model and respond to students' emotional and cognitive states in real time. The system integrates multimodal affect sensing (facial expressions, voice prosody, interaction patterns) with domain-specific conceptual diagnostics to create adaptive feedback strategies that align instructional support with learners’ affective states and knowledge gaps. A mixed-methods approach guides the design, development, and evaluation, combining data-driven modeling with pedagogical theory on conceptual change, cognitive load, and motivation. The research employs a modular architecture comprising an affective layer, a domain understanding layer, and an adaptive tutoring engine. The affective layer uses supervised and unsupervised learning to infer emotions such as confusion, frustration, curiosity, and engagement from audiovisual streams and interaction traces, while the domain understanding layer constructs a dynamic model of learners’ conceptual network in core computer science topics (programming constructs, data structures, algorithms, and software engineering principles) using Bayesian knowledge tracing and concept maps. The tutoring engine integrates instructional strategies grounded in social-cognitive learning theory, cognitive load theory, and multimedia learning principles, delivering adaptive hints, scaffolded problems, reflective prompts, and practice tasks calibrated to the learner’s current affective and cognitive state. The study also introduces a personalized assessment framework that triangulates diagnostic tests, trace data, and self-reports to classify learners into proficiency archetypes and emotion profiles, guiding tailored remediation paths. A pilot study with diverse learners across introductory computer science courses informs iterative refinements, followed by a larger quasi-experimental trial measuring conceptual mastery, transfer performance, engagement, and affective experience. Outcomes are evaluated using pre/post concept inventories, problem-solving transfer tasks, and validated affective state scales, complemented by server-side analytics to monitor system adaptability and user satisfaction. The anticipated contributions include (1) a robust multimodal affect detection pipeline integrated with LDCS-based knowledge tracing for real-time adaptation; (2) an evidence-based repertoire of affect-aware instructional interventions that reduce cognitive overload while promoting conceptual coherence; (3) a scalable framework for deploying AC-ITS in standard computer education settings with minimal overhead on instructors; and (4) empirical insights into how affective states mediate learning gains in conceptual domains and how machine learning-driven personalization affects intrinsic motivation and persistence. Ethical considerations address privacy, data security, and transparency in affect inference, with a design that offers opt-in participation, data minimization, and clear user control over teaching strategies. The study aims to demonstrate that aligning affect-aware instructional support with precise conceptual diagnostics can accelerate the development of robust foundational knowledge in computer science, yielding higher learning gains, improved transfer, and enhanced learner autonomy.

Project Overview

What This Project Is About

---

A simple student-friendly overview of how emotions and computer learning tools can work together to improve understanding in computer science topics. The project builds a system that senses students’ feelings or engagement levels and adapts learning content to help them grasp concepts more clearly, using basic machine learning ideas to tailor feedback.

---

---

The Problem It Addresses

---

Many students struggle to stay engaged or misunderstand core computer science ideas. Without guidance that fits their mood or attention, learning can be slow or frustrating. This project aims to reduce these gaps by making tutoring more responsive and personalized.

---

---

Objectives of the Project

---

1. Identify typical student struggles in core computer science topics.
2. Build a simple system that detects student engagement or emotion from reactions or input.
3. Adapt explanations and practice tasks based on detected states to improve understanding.
4. Evaluate whether the system helps students learn faster or remember concepts longer.

---

---

What You Will Do Step by Step

---

1) Review basic ideas of tutoring systems and emotion-friendly learning. 2) Collect data from volunteers using simple prompts or activities to gauge engagement. 3) Develop a small model that predicts when a student needs a hint or a different example. 4) Create learning activities and feedback that adapt to the predicted state. 5) Test with a group of students and compare learning outcomes. 6) Analyze results to see if adaptation helps. 7) Revise based on feedback and prepare a simple report.

---

---

---

---

Expected Outcome

Expect a prototype tutoring system that adjusts lessons based on student engagement to improve conceptual understanding, with initial evidence of better retention and satisfaction among users.