Apprentice Tutor Builder: A Platform For Users to Create and Personalize Intelligent Tutors

TL;DR

Intelligent tutoring systems offer scalable learning support but are hard to author, requiring specialized programming and tutor design expertise. The Apprentice Tutor Builder provides a drag-and-drop tutor interface paired with interactive training of HTN-based expert agents through demonstrations, feedback, and labels. The work reports a user study with 14 instructors that demonstrates workable tutor creation and expert-model training and yields actionable design recommendations for teacher-friendly authoring tools. The findings suggest ATB can enable personalized tutoring at scale by lowering the technical burden for educators and by exposing intuitive, interactive AI authoring workflows.

Abstract

Intelligent tutoring systems (ITS) are effective for improving students' learning outcomes. However, their development is often complex, time-consuming, and requires specialized programming and tutor design knowledge, thus hindering their widespread application and personalization. We present the Apprentice Tutor Builder (ATB) , a platform that simplifies tutor creation and personalization. Instructors can utilize ATB's drag-and-drop tool to build tutor interfaces. Instructors can then interactively train the tutors' underlying AI agent to produce expert models that can solve problems. Training is achieved via using multiple interaction modalities including demonstrations, feedback, and user labels. We conducted a user study with 14 instructors to evaluate the effectiveness of ATB's design with end users. We found that users enjoyed the flexibility of the interface builder and ease and speed of agent teaching, but often desired additional time-saving features. With these insights, we identified a set of design recommendations for our platform and others that utilize interactive AI agents for tutor creation and customization.

Figures (6)

• Figure 1: Apprentice Tutor Builder: (a) the interface builder starts from a blank canvas, and (b) users can drag and drop components from the right component box to the draggable pane area.
• Figure 2: A sample working memory for the fraction arithmetic task. The facts of the tutor consist of the field names and their values.
• Figure 3: A sample HTN decomposition for fraction multiplication. The blue squares represent tasks, the green circles are methods, and the red squares are operators.
• Figure 4: Interactive dialogs for training the expert model: (A) the system asking the user to provide a demonstration, (B) The agent requesting a label for a previously provided demosntration, (C) the agent requesting feedback on the correctness after taking a step, and (D) the agent asking if the problem has been correctly completed.
• Figure 5: Experience levels of participants in four categories (1) artificial intelligence, (2) mathematics, (3) programming/coding, and (4) teaching (top) and instruction levels taught (bottom)