Peer agents can supplement real-time collaborative learning in asynchronous online courses. Constructive Controversy (CC) theory suggests that humans deepen their understanding of a topic by confronting and resolving controversies. This study explores whether CC’s benefits apply to LLM-based peer agents, focusing on the impact of agents’ disputatious behaviors and disclosure of agents’ behavior designs on the learning process. In our mixed-method study (n=144), we compare LLMs that follow detailed CC guidelines (regulated) to those guided by broader goals (unregulated) and examine the effects of disclosing the agents’ design to users (transparent vs. opaque). Findings show that learners’ values influence their agent interaction: those valuing control appreciate unregulated agents’ willingness to cease push-back upon request, while those valuing intellectual challenges favor regulated agents for stimulating creativity. Additionally, design transparency lowers learners’ perception of agents’ abilities. Our findings lay the foundation for designing effective collaborative peer agents in isolated educational settings.

Exposure to diverse perspectives is helpful for bursting the filter bubble in online public video platforms. The recent advancement of Large Language Models (LLMs) illuminates the potential of creating a debate chatbot that prompts users to critically examine their stances on a topic formed by watching videos. However, whether the viewer is influenced by the chatbot may depend on its persona. In this paper, we investigated the effect of two relevant persona attributes - social identity and rhetorical styles - on critical thinking. In a mixed-methods study (n=36), we found that chatbots with outgroup (vs. ingroup) identity (t(33)=-2.33, p=0.03) and persuasive (vs. eristic) rhetoric (t(44)=1.98, p=0.05) induced critical thinking most effectively, making participants re-examine their arguments. However, participants’ stances remain largely unaffected, likely due to the chatbot’s lack of contextual knowledge and human touch. Our paper provides empirical groundwork for designing chatbot persona for remedying filter bubbles in online communities.

Videos are convenient resources for asynchronous learning, but they lack interpersonal interactions found in synchronous classrooms. Due to missed social connectedness, the isolated video-based learners experience low emotional, behavioral, and cognitive engagement. This work presents "Scripted Vicarious Dialogues" (SVD), a technique for engaging students in a pseudo-social experience of witnessing scripted dialogues between virtual characters (teaching assistants and students) around a video. We conducted a participatory design study to derive design guidelines for SVD. The findings indicate the need to distinguish the virtual components and to give students control of the dialogue’s pace. We then implemented an interactive prototype of SVD and evaluated it (N=40) against a non-social, direct-learning baseline. The results show that the preference for SVD versus the baseline is polarized (25 of 40 preferred SVD; no neutral preferences), and those who preferred SVD had significantly higher emotional and behavioral engagement with SVD compared to the baseline.

This article investigates a cross-modal retrieval problem in which a user would like to retrieve a passage of music from a MIDI file by taking a cell phone picture of several lines of sheet music. This problem is challenging for two reasons: it has a significant runtime constraint since it is a user-facing application, and there is very little relevant training data containing cell phone images of sheet music. To solve this problem, we introduce a novel feature representation called a bootleg score which encodes the position of noteheads relative to staff lines in sheet music. The MIDI representation can be converted into a bootleg score using deterministic rules of Western musical notation, and the sheet music image can be converted into a bootleg score using classical computer vision techniques for detecting simple geometrical shapes. Once the MIDI and cell phone image have been converted into bootleg scores, we can estimate the alignment using dynamic programming. The most notable characteristic of our system is that it has no trainable weights at all - only a set of about 40 hyperparameters. With a training set of just 400 images, we show that our system generalizes well to a much larger set of 1600 test images from 160 unseen musical scores. Our system achieves a test F measure score of 0.89, has an average runtime of 0.90 seconds, and outperforms baseline systems based on music object detection and sheet-audio alignment. We provide extensive experimental validation and analysis of our system.

Previous research has documented correlations between spatial reasoning ability and success in STEM fields [7-9]. While connection between spatial reasoning and other STEM fields like physics and calculus may seem obvious, there are no theories to explain this correlation of CS (computer science) performance and spatial reasoning. [5,7,8]. We aim to better understand this correlation, specifically between CS performance in an introductory course and spatial reasoning by observing the common characteristics between students' strategies in solving CS and spatial reasoning problems. We conducted interviews with eight students who have prior experience in CS but have taken only two introductory CS courses. In the interview, we asked the participants to solve a total of four problems; two CS problems and two spatial reasoning problems. [1,6]. The CS problems were code-reading problems in Python, which focus on basic programming concepts, including nested loops and functions. We analyzed the participants' problem-solving strategies and recorded their explanation for each problem. We observed for both types of problem, participants appeared to first observe the problem for its fundamental structures then decomposes the problem into smaller sub-problems. We conjecture that problem decomposition may be a required skill to solve both CS and spatial reasoning problems, and thus, could be a possible factor that contributes to the correlation between the two fields. Further study into utilization of problem decomposition in CS and spatial reasoning may provide more insights into the correlation between success in CS and spatial reasoning ability.