Recent Advances in Intelligent Textbooks for Better Learning

Bo Jiang, Meijun Gu, Ying DuOpenalex Unknown 2022 · DOI · OpenAlex

Abstract

Intelligent textbooks are a new form of digital textbooks that provide students with intelligent learning services, such as automatic question answering, adaptive navigation support, automatic linking, and personalized recommendation. Educators are investing much effort into the conversion of their paper-based textbooks and digital textbooks into intelligent textbooks. This paper provides a brief introduction to the development of intelligent textbooks and reviews major adaptive technologies and the usage of intelligent textbooks in the past decade. We first summarize the development of intelligent textbooks and the existing authoring platforms. Then, the key intelligent technologies used in the intelligent textbooks are reviewed, including domain modeling, student modeling, and instruction technologies. Moreover, the empirical studies conducted to evaluate intelligent textbooks are summarized. This review promotes understanding of the current state of the technology and platform development of intelligent textbooks.

Introduction

Traditional digital textbooks, such as standard PDFs, are largely passive resources. While they offer basic tools like highlighting and note-taking, they lack the ability to understand why a student is interacting with the text or where they are struggling. The field of "intelligent textbooks" (iTBs) aims to address this by transforming digital reading materials into active, AI-driven tutoring systems. By integrating Intelligent Tutoring Systems (ITS) with standard reading interfaces, these platforms turn passive reading into an interactive, data-driven learning experience.

Core Technologies in Intelligent Textbooks

The transition from static content to intelligent systems relies on three primary pillars of modeling:

  • Domain Modeling: This involves automatically extracting concepts and relationships from text. Current approaches utilize NLP techniques like "bag-of-words," chapter-based semantic analysis, and latent topic modeling to build knowledge maps without requiring extensive manual expert labor.
  • Student Modeling: Adapting "Knowledge Tracing"—traditionally used in mathematics software—for reading contexts. Since reading does not always provide clear right/wrong answers, these models use proxies such as time-on-page, highlighting patterns, and underlining to predict a student's mastery level.
  • Instructional Design: This is the delivery layer that provides adaptive navigation (suggesting the "next best page"), tutoring dialogs (using chatbots for inquiry-based questions), and error-sensitive feedback on embedded interactive exercises.

Evolution of Intelligent Platforms

Over the last decade, a variety of platforms have emerged, targeting subjects ranging from computer science to biology and mathematics. These platforms vary in their approach, from providing instant feedback on code exercises to using interactive animations for conceptual knowledge.

Summary of major intelligent textbook platforms and their core functionalities.

As shown in the table above, platforms like Inquire Biology and zyBooks have pioneered the use of automated question-answering and interactive simulations, while newer systems like Intextbooks focus on the automated conversion of existing PDF resources into intelligent formats.

Empirical Insights and Key Results

Research into these systems has yielded several critical insights into digital learning behavior:

  • Reading Behavior as a Predictor: Highlighting patterns alone can explain approximately 13% of the variance in observed quiz grades, suggesting that how a student reads is a strong signal for what they are learning.
  • The Social Navigation Effect: Platforms like Reading Mirror use "icicle plots" to show students how their progress compares to peers. While this increases engagement, students often require a "warm-up" period of about three weeks to adapt to social comparison features before seeing learning gains.
  • Learning Gains: Empirical studies show significant promise. For example, Inquire Biology demonstrated significant improvements in homework quiz scores ($p = 0.02$). Large-scale studies on the Runestone platform showed that interactive components significantly increased student motivation and programming gains compared to static PDFs.
  • Differential Impact: Intelligent textbooks seem particularly effective for "at-risk" learners. Systems like ELM-ART significantly improved test scores specifically for students with weak initial programming skills.

Challenges and Future Directions

Despite these advances, several bottlenecks remain. The "authoring bottleneck" makes creating intelligent textbooks incredibly time-consuming and expensive. Furthermore, current AI still struggles to fully "understand" mathematical language and complex formulas, hindering development in certain STEM fields.

Future work must focus on developing robust, automated pipelines to turn "flat" PDF files into structured, semantic knowledge bases. Additionally, "closing the loop" by interconnecting intelligent textbooks with Learning Management Systems (LMS) and formal classroom practices will be essential to creating a unified, data-driven educational ecosystem.

Summary updated

Abstract

Abstract Intelligent textbooks are a new form of digital textbooks that provides students with intelligent learning services, such as automatic question answering, adaptive navigation support, automatic linking, and personalized recommendation. Educators are investing much effort into the conversion of their paper-based textbooks and digital textbooks into intelligent textbooks. This chapter provides a brief introduction to the development of intelligent textbooks and reviews major adaptive technologies and the usage of intelligent textbooks in the past decade. We first summarize the development of intelligent textbooks and the existing authoring platforms. Then, the key intelligent technologies used in the intelligent textbooks are reviewed, including domain modeling, student modeling, and instruction technologies. Moreover, the empirical studies conducted to evaluate intelligent textbooks are summarized. This chapter would promote understanding of the current state of the technology and platform development of intelligent textbooks.
BibTeX
@inproceedings{jiang2022recent,
  author = {Bo Jiang and Meijun Gu and Ying Du},
  title = {{Recent Advances in Intelligent Textbooks for Better Learning}},
  booktitle = {Openalex Unknown},
  year = {2022},
  doi = {10.1007/978-3-031-09687-7\_15},
}
Figure