Teaching university-level courses benefits from a consistent effort to heighten students' learning engagement and offering tailored learning experiences. In this endeavor, motivating students to engage with textbook material stands as a crucial yet challenging objective. Inspired by the recent advancements in expansive language models and human-centric AI, we leverage cutting-edge large language models and innovative text visualization techniques when developing an interactive tool that assists students’ reading and learning experience. Our web-based tool helps individualize how students navigate, explore, and learn from textbooks, powered by visualization models. Our solution encompasses conceptual visualizations derived from the textbook’s content, alongside an AI-powered chatbot adept at addressing
student inquiries and guiding their reading journey. We plan to evaluate the tool within a controlled study in real-world teaching contexts in Fall 2024.
What is IRead?
IRead is an interactive textbook exploration tool that combines knowledge visualization and natural language processing techniques to provide a new and individualized learning experience for students. The tool is accessed through a webpage interface using computers or tablets. Figure 1 shows a potential mock-up interface of IRead. In the example in Figure 1, the student has searched the text for “Forklift” after instruction from the teacher. To the left, the tool shows the relevant pages to read. To the right, the tool shows relations among the concept “forklift” (centered) and other related concepts in the textbook. The bubble closeness and size represent the strengths and scope of the relationship. The bubbles can be zoomed into for further nested concepts and details. The concept visualization helps the student (1) understand relationships across the book’s chapters and content and (2) navigate easily to the next recommended reading. The tool remembers the students’ covered material and dynamically suggests what to read next. At the bottom right a chatbot generates study questions for the student, prompting them to read further.
Description Figure 1:
The above example is solely based on the book’s content. We also enhance the book with external sources. We integrate illustrations and generative AI-generated (or selected sources) explanations from outside the textbook.
Why do we develop IRead? In all teaching, it is important to motivate students to actively immerse themselves in the course material. Traditionally, the textbook has played a central role in this setting. A textbook however represents the authors’ structured summary of an entire field or area. By a book’s physical design, it provides a linear presentation of the course material and is not tailored to students’ individual backgrounds and learning strategies. Consider for example these extracted examples from the LET course evaluation of 363-0445-00S Production and Operations Management fall semester 2022:
- “Reading the chapters of the book also takes a lot of time, especially for slow-readers like I am” (Student).
- “On certain topics, I would have liked more information and was not able to find them in the book. I spend a lot of time googling things.” (Student)
- “Maybe adding so many readings on top is a bit of an overkill, honestly,
I did not make time to read everything, it takes too much time.” (Student) Diverse learning preferences and backgrounds of students can pose obstacles to uniform engagement strategies. The textbook format traditionally doesn’t cater well to students’ different needs. Recent advances in large language models and human-centric AI bring promising solutions to this long-standing challenge.
How do we assess IRead?
After the development and implementation, we evaluate the quality and performance of the tool scientifically. We invite ETH students and undertake comparative user studies with them. These studies occurred either in person or via online video conference sessions. During the study sessions, participants engaged in predefined tasks and responded to questions centered around the textbook content. We organized the students into two groups: one group will utilize conventional PDF reading with a plain textbook, while the other will follow the workflow and utilize the tool we have developed. We captured their mouse movements and video and audio recordings of their interactions and commentary throughout the session. In addition, some participants were invited to address additional interview questions based on their exploratory actions. Subsequently, we conducted qualitative and quantitative analyses based on our collected feedback.
As a future plan, we will test the beta version of the system in 363-0445-00L Production and Operations Management and can regress the tool usage time on course grading. The tool will be tested with the 800-page textbook “Introduction to Manufacturing: An Industrial Engineering and Management Perspective” (Baudin and Netland, 2022; Routledge) shown in Figure 3. Using Netland’s book enables access and use rights, and since this book is used in courses at ETH Zurich it also enables experimentation and feedback.
Methods, tools or strategies did to encourage student engagement for learning success
Including the student perspective is crucial to the success of this project. We provide two distinct student involvement opportunities to address both development and evaluation needs. First, we will involve students as our target users to understand user requirements and evaluate our system in multiple ways:
First, we will involve students as our target users to understand user requirements and evaluate our system in multiple ways:
I. We invite ETH students to participate in our user studies and we collect their feedback on concepts and module developments.
II. We will provide the beta version of the tool that we develop for the textbook “Introduction to Manufacturing: An Industrial Engineering and Management Perspective” (Baudin and Netland, 2022) to students of the D-MTEC course Production and Operations Management (~150 students) in the fall term of 2024.
III. In the post-project implementation, the tool will be continuously revised based on user feedback and expanded to encompass a wider range of textbooks and courses throughout ETH and elsewhere.
Second, we offered the opportunity for semester course projects and Bachelor’s thesis projects to students with backgrounds in computer science or related fields. This involvement engages them directly in the developmental phase of the system.
Innovative elements of the project
There are three major innovative elements of our project:
I. For ETH students, we introduce a new reading tool that holds the potential to enhance motivation and learning dynamics when engaging with textbooks. Contrasting the traditional textbook, the tool allows personalization. Such individualization can take a detailed form where each student logs in and the tool remembers the parts that have been read and helps students navigate to the next step based on past readings and interests.
II. For faculty members, we provide a more attractive and effective method to immerse students in course content with constrained teaching resources, fostering tailored and personalized learning experiences as explained above.
III. For the entire degree programme, we offer a versatile new tool that all ETH courses could potentially utilize in the future to enhance the digitalization and student engagement across the curriculum.
What effect did the innovative elements have on student learning?
Our individualized reading tool can always offer different concepts to explore next, to allow for dynamic exploration-based learning based on the student’s curiosity and interests. The personalization can also be at a less granular level: for example, in the course Production and Operations Management, one could provide different learning paths for Bachelor’s students (BScI, Master’s of Science students, and Master of Advanced Studies (MAS) students. This will further empower students to actively engage with the course material, resulting in deeper comprehension and a more pleasant learning experience. Ultimately, this should improve their learning outcomes. Our tool could also enable educators to cater to individual learning styles, optimize instructional strategies, and ensure the delivery of foundational knowledge to students. Pedagogically, this project is grounded in the guided discovery learning approach [1, 2, 3], which strikes a balance between didactic teaching and unassisted discovery learning. On one hand, we plan to facilitate discovery learning [4] and enhance student engagement by developing an interactive textbook exploration tool. In particular, our approach aligns with constructivist-based education principles[ 5], where students actively construct their understanding and knowledge by integrating new information with their prior knowledge, rather than passively receiving it. This theory finds its roots in the work of Swiss developmental psychologist Jean Piaget. On the other hand, we recognize that «free exploration of a highly complex environment may overload working memory and hinder learning», as described by the cognitive load theory [6]. To mitigate this, we incorporate inquiry-based learning theory [7] to create a guided discovery learning experience, fostering a deeper understanding of the material with reduced cognitive load. To achieve this, we combine teacher expertise with large language models to develop a facilitator that guides students through their learning adventures. Leveraging the teacher’s input of a suggested reading trajectory and the students› reading history, the large language model generates personalized questions based on the upcoming content. This approach encourages students to seek answers through further exploration of the textbook, resulting in iterative rounds of discovery and constructivist-based learning.
How did you ensure (continues) feedback on student learning progress?
After the development and implementation, we evaluate the quality and performance of the tool scientifically. We invite ETH students and undertake comparative user studies with them. These studies occurred either in person or via online video conference sessions. During the study sessions, participants engaged in predefined tasks and responded to questions centered around the textbook content. We organized the students into two groups: one group will utilize conventional PDF reading with a plain textbook, while the other will follow the workflow and utilize the tool we have developed. We captured their mouse movements and video and audio recordings of their interactions and commentary throughout the session. In addition, some participants were invited to address additional interview questions based on their exploratory actions. Subsequently, we conducted qualitative and quantitative analyses based on our collected feedback. As a future plan, we will test the beta version of the system in 363-0445-00L Production and Operations Management and can regress the tool usage time on course grading. The tool will be tested with the 800-page textbook “Introduction to Manufacturing: An Industrial Engineering and Management Perspective” (Baudin and Netland, 2022; Routledge) shown in Figure 3. Using Netland’s book enables access and use rights, and since
Which elements of your project would you recommend to others?
I. A new learning paradigm that effectively directs students in actively and collaboratively delving into specific subject matter within a textbook. II. An AI-powered visualization tool that seamlessly incorporates the aforementioned workflow, enhancing the process of active engagement and deeper comprehension while reading a textbook. III. The practical application of the aforementioned outcomes using the textbook «Introduction to Manufacturing: An Industrial Engineering and Management Perspective”. Within the context of the «363-0445- 00S Production and Operations Management» course led by
What discussion points are you particularly interested in when exchanging with other lecturers?
- How to Guide students to use AI tools correctly and ethically for their learning?
- How to balance the introduction of new interactive technology to avoid distracting students from reading?
- How to guarantee accuracy to safeguard the AI’s outcomes?
- The role of the textbook in future teaching V. The role of Generative AI in future teaching
References
[1] Brown, Ann L., and Joseph C. Campione. Guided discovery in a community of learners. The MIT Press, 1994.
[2] Brown, Ann L. Design experiments: Theoretical and methodological challenges in creating complex interventions in classroom settings. The journal of the learning sciences 2.2 (1992): 141-178.
[3] Palincsar, Annemarie Sullivan. Collaborative Research and Development of Reciprocal Teaching. Educational leadership 46.4 (1989): 37-40.
[4] Dewey, John. «The university elementary school: History and character.» University Record 2 (1897): 72-5.
[5] Matthews, Michael R., ed. Constructivism in science education: A philosophical examination. Springer Science & Business Media, 1998.
[6] Kirschner, Paul A., John Sweller, and Richard E. Clark. «Why minimal guidance during instruction does not work: An analysis of the failure of constructivist, discovery, problem-based, experiential, and inquiry-based teaching.» Educational psychologist 41.2 (2006): 75-86.
[7] Bruner, Jerome S. «The act of discovery.» Harvard Educational Review (1961).