Prerna Ravi

Understanding how ideas develop and flow in small-group conversations is critical for analyzing collaborative learning. A key structural feature of these interactions is threading, the way discourse talk naturally organizes into interwoven topical strands that evolve over time. While threading has been widely studied in asynchronous text settings, detecting threads in synchronous spoken dialogue remains challenging due to overlapping turns and implicit cues. At the same time, large language models (LLMs) show promise for automating discourse analysis but often struggle with long-context tasks that depend on tracing these conversational links. In this paper, we investigate whether explicit thread linkages can improve LLM-based coding of relational moves in group talk. We contribute a systematic guidebook for identifying threads in synchronous multi-party transcripts and benchmark different LLM prompting strategies for automated threading. We then test how threading influences performance on downstream coding of conversational analysis frameworks, that capture core collaborative actions such as agreeing, building, and eliciting. Our results show that providing clear conversational thread information improves LLM coding performance and underscores the heavy reliance of downstream analysis on well-structured dialogue. We also discuss practical trade-offs in time and cost, emphasizing where human-AI hybrid approaches can yield the best value. Together, this work advances methods for combining LLMs and robust conversational thread structures to make sense of complex, real-time group interactions.

Data science is emerging as a crucial 21st-century competence, influencing professional practices from citing evidence when advocating for social change to developing artificial intelligence (AI) models. For middle and high school students, data science can put formerly decontextualized subjects into real-world scenarios. Many existing curricula, however, lack authenticity and personal relevance for students. A critique of data science courseware cites the lack of "author proximity," in which students do not contribute to the data's production or see their personal experiences reflected in the data. This paper introduces a novel data science curriculum to scaffold middle and high school students in undertaking real-world data science practices. Through project-based learning modules, the curriculum engages students in investigating solutions to community-based problems through visualization and analysis of live sensor data and public data sets. Materials include formative assessments to help educators (especially those from non-math and computing backgrounds) measure their students' abilities to identify statistical patterns, critically evaluate data biases, and make predictions. As we pilot and co-design with teachers, we will look closely at whether the curriculum's resources can successfully support non-technical practitioners engaging in an integrated curriculum.

App Planner is an interactive support tool for K-12 students, designed to assist in creating mobile applications. By utilizing generative AI, App Planner helps students articulate the problem and solution through guided conversations via a chat-based interface. It assists them in brainstorming and formulating new ideas for applications, provides feedback on those ideas, and stimulates creative thinking. Here we report usability tests from our preliminary study with high-school students who appreciated App Planner for aiding the app design process and providing new viewpoints on human aspects especially the potential negative impact of their creation.