eTutor: Online Learning for Personalized Education

Cem Tekin, Mihaela van der Schaar

Abstract: Given recent advances in information technology and artificial intelligence, web-based education systems have became complementary and, in some cases, viable alternatives to traditional classroom teaching. The popularity of these systems stems from their ability to make education available to a large demographics (see MOOCs). However, existing systems do not take advantage of the personalization which becomes possible when web-based education is offered: they continue to be one-size-fits-all. In this paper, we aim to provide a first systematic method for designing a personalized web-based education system. Personalizing education is challenging: (i) students need to be provided personalized teaching and training depending on their contexts (e.g. classes already taken, methods of learning preferred, etc.), (ii) for each specific context, the best teaching and training method (e.g type and order of teaching materials to be shown) must be learned, (iii) teaching and training should be adapted online, based on the scores/feedback (e.g. tests, quizzes, final exam, likes/dislikes etc.) of the students. Our personalized online system, e-Tutor, is able to address these challenges by learning how to adapt the teaching methodology (in this case what sequence of teaching material to present to a student) to maximize her performance in the final exam, while minimizing the time spent by the students to learn the course (and possibly dropouts). We illustrate the efficiency of the proposed method on a real-world eTutor platform which is used for remedial training for a Digital Signal Processing (DSP) course.

Introduction

While Massive Open Online Courses (MOOCs) have democratized access to education, they suffer from a significant structural flaw: they are largely static. In current systems, every student—regardless of whether they are a struggling novice or a bored expert—typically encounters the same content in the same sequence. This "one-size-fits-all" approach often leads to high dropout rates, as experts disengage when the material lacks challenge, and novices quit when they become overwhelmed.

To address this, we identify three core challenges in personalizing education: Contextual Diversity (varying student backgrounds), Methodological Uncertainty (the unknown optimal teaching path for specific concepts), and Dynamic Adaptation (the need to react to real-time performance feedback). As shown in the comparison below, eTutor is the first framework to systematically address all these dimensions while providing theoretical guarantees on performance.

Comparison of eTutor with existing educational methodologies.

The eTutor Framework

We formalize the tutoring process as an online learning problem. The eTutor algorithm is designed to minimize "regret"—defined as the performance gap between the eTutor and an "ideal" benchmark tutor that possesses a priori knowledge of the optimal teaching strategy.

High-level architecture of the eTutor system.

Key Algorithmic Components

  • Exploration vs. Exploitation: eTutor maintains sample mean reward estimates for various teaching sequences. If a specific path has not been sufficiently tested (tracked via a logarithmic counter), the system "explores" it. Otherwise, it "exploits" the best-known path to maximize the student's score.
  • Teaching Cost ($c$): Unlike models that treat student time as infinite, eTutor incorporates a cost parameter into its reward function. The algorithm aims for the highest score relative to time spent, allowing it to skip materials that offer only marginal improvements at a high time cost.
  • Context-Based Learning: To accelerate the learning process, eTutor groups students by context (e.g., prior knowledge). This allows the system to apply insights gained from previous students to new learners with similar backgrounds.

Experimental Evaluation

The system was evaluated through a real-world deployment in a remedial Digital Signal Processing (DSP) course focusing on Discrete Fourier Transforms (DFT). The "pool of materials" consisted of text-based explanations and questions of varying difficulty. We compared eTutor against a Random Rule (RR) and a Fixed Rule (FR), the latter of which presents all materials in a predetermined order.

Key Results

The experiment, involving 500 students, yielded the following insights:

  1. Score Improvement: eTutor achieved a 15.7% improvement in average final exam scores compared to the Random Rule.
  2. Efficiency Gains: While eTutor’s scores were comparable to the Fixed Rule (a 1.1% increase), it achieved these results with 50% less time spent by students. On average, students using eTutor spent 8.5 minutes on the material, compared to 17 minutes for those under the Fixed Rule.
  3. Theoretical Convergence: We mathematically proved that as the student population grows, eTutor’s performance converges to the optimal strategy, ensuring the system improves over time rather than fluctuating randomly.

Conclusion and Future Work

eTutor demonstrates that personalized online learning can significantly improve educational efficiency without sacrificing mastery. By treating time as a finite resource and leveraging student context, the system successfully identifies which materials are essential for individual success.

Future research will focus on scaling the system to full-semester courses with larger material pools, which increases the algorithmic search space. Additionally, we aim to explore adapting the type of material (e.g., video vs. text) and personalizing the "cost" parameter ($c$) to reflect the varying time constraints of different learner demographics.

Summary updated

Abstract

Given recent advances in information technology and artificial intelligence, web-based education systems have became complementary and, in some cases, viable alternatives to traditional classroom teaching. The popularity of these systems stems from their ability to make education available to a large demographics (see MOOCs). However, existing systems do not take advantage of the personalization which becomes possible when web-based education is offered: they continue to be one-size-fits-all. In this paper, we aim to provide a first systematic method for designing a personalized web-based education system. Personalizing education is challenging: (i) students need to be provided personalized teaching and training depending on their contexts (e.g. classes already taken, methods of learning preferred, etc.), (ii) for each specific context, the best teaching and training method (e.g type and order of teaching materials to be shown) must be learned, (iii) teaching and training should be adapted online, based on the scores/feedback (e.g. tests, quizzes, final exam, likes/dislikes etc.) of the students. Our personalized online system, e-Tutor, is able to address these challenges by learning how to adapt the teaching methodology (in this case what sequence of teaching material to present to a student) to maximize her performance in the final exam, while minimizing the time spent by the students to learn the course (and possibly dropouts). We illustrate the efficiency of the proposed method on a real-world eTutor platform which is used for remedial training for a Digital Signal Processing (DSP) course.
BibTeX
@inproceedings{tekin2015etutor,
  author = {Cem Tekin and Mihaela van der Schaar},
  title = {{eTutor: Online Learning for Personalized Education}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2015},
  url = {https://openreview.net/forum?id=_VynAsx3GI},
}
Figure