Co-designing Large Language Model Tools for Project-Based Learning with K-12 Educators

Prerna Ravi, John Masla, Gisella Kakoti, Grace C. Lin, Emma Anderson, Matt Taylor, Anastasia K. Ostrowski, Cynthia Breazeal, Eric Klopfer, Hal Abelson · 2025

Read Co-designing Large Language Model Tools for Project-Based Learning with K-12 Educators by Prerna Ravi, John Masla, Gisella Kakoti, Grace C. Lin, Emma Anderson, Matt Taylor, Anastasia K. Ostrowski, Cynthia Breazeal, Eric Klopfer, Hal Abelson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-09799.txt.

Introduction

Prerna Ravi and nine co-authors report a 2025 study on school teachers and language models. The team worked with teachers across subjects in primary and secondary schools. Their shared focus was project based learning, where students learn by building real projects. Such teaching is rich but hard to plan, run, and grade.

Teachers took part in talks, workshops, and rounds of draft designs. They said software help should handle routine tasks and tailor support to each student. They also wanted tools that grow their own skills without taking over their role. The group raised worries about cost, fairness, and long term effects in class. From this work the authors offer design guidance for future classroom tools.

Catalog record

CREATORPrerna Ravi, John Masla, Gisella Kakoti, Grace C. Lin, Emma Anderson, Matt Taylor, Anastasia K. Ostrowski, Cynthia Breazeal, Eric Klopfer, Hal Abelson
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE103 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-09799

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Co-designing Large Language Model Tools for Project-Based Learning with K-12 Educators Prerna Ravi, John Masla, Gisella Kakoti, Grace C. Lin, Emma Anderson, Matt Taylor, Anastasia K. Ostrowski, Cynthia Breazeal, Eric Klopfer, Hal Abelson Abstract . The emergence of generative AI, particularly large language models (LLMs), has opened the door for student-centered and active learning methods like project-based learning (PBL). However, PBL poses practical implementation challenges for educators around project design and management, assessment, and balancing student guidance with student autonomy. The following research documents a co-design process with interdisciplinary K-12 teachers to explore and address the current PBL challenges they face. Through teacher-driven interviews, collaborative workshops, and iterative design of wireframes, we gathered evidence for ways LLMs can support teachers in implementing high-quality PBL pedagogy by automating routine tasks and enhancing personalized learning. Teachers in the study advocated for supporting their professional growth and augmenting their current roles without replacing them. They also identified affordances and challenges around classroom integration, including resource requirements and constraints, ethical concerns, and potential immediate and long-term impacts. Drawing on these, we propose design guidelines for future deployment of LLM tools in PBL. Conference: CHI Conference on Human Factors in Computing Systems; April 26-May 1, 2025; Yokohama, Japan CHI Conference on Human Factors in Computing Systems (CHI ’25), April 26-May 1, 2025, Yokohama, Japan DOI: 10.1145/3706598.3713971 ISBN: 979-8-4007-1394-1/25/04 CCS: Human-centered computing Participatory design CCS: Computing methodologies Artificial intelligence CCS: Applied computing Education CCS: Human-centered computing Empirical studies in HCI † † cc-license: by 1. INTRODUCTION Project-based learning (PBL) has gained prominence as a K-12 educational approach that immerses students in meaningful, real-world tasks, fostering deeper learning experiences ( Boaler, 1998 ; Panasan and Nuangchalerm, 2010 ; Schneider et al., 2002 ; Chen and Yang, 2019 ; PBLWorks, 2019 ) . Unlike traditional instructional methods, PBL emphasizes student-centered pedagogy, where learners actively construct knowledge through exploration, collaboration, and reflection ( Condliffe, 2017 ; Noguera et al., 2015 ; Peterson, 2012 ) . By providing students with opportunities to explore topics of personal and cultural relevance, PBL supports diverse learning needs and backgrounds, making education more inclusive and accessible. While PBL offers significant benefits to students, it also necessitates a fundamental shift in educators’ roles from knowledge providers to learning facilitators ( Ertmer and Simons, 2006 ; Savery, 2015 ) . This requires educators to adopt new pedagogical strategies, manage complex classroom dynamics and provide ongoing, formative assessments that support student learning ( Sahoo et al., 2013 ; Albanese and Hinman, 2019 ) . Challenges to PBL implementation include designing standards-aligned projects that meet diverse student needs, managing planning time, transitioning students to active roles ( Simons et al., 2004 ; Chen and Yang, 2019 ; Thomas, 1998 ; Zheng et al., 2024b ; Viro et al., 2020 ; Ertmer and Simons, 2006 ; Gallagher, 1997 ) , creating authentic driving questions and assessing interpersonal skills, which traditional assessment methods often fail to capture ( Markula and Aksela, 2022 ; Blumenfeld et al., 1991 ; Mentzer et al., 2017 ; Zheng et al., 2024b ; Aksela et al., 2019 ; Colley, 2006 ; Brinkerhoff and Glazewski, 2004 ) . Such challenges emphasize the need for nuanced instructional strategies for PBL that better capture the scope of student learning ( Zheng et al., 2024b ) . To address these challenges, the Human-Computer Interaction (HCI) community has shown a rising interest in generative artificial intelligence (GenAI) systems for education ( Muller et al., 2022 ; Sun et al., 2024 ; Weisz et al., 2024 ; Han et al., 2024 ; Zheng et al., 2024b ; Cai et al., 2024 ; Shaer et al., 2024 ) . Large language models (LLMs) in particular are uniquely suited to address the inherent complexities of PBL ( Gustafson et al., 2025 ; Zheng et al., 2024a ; Nikolic and Beckman, [n. d.] ) due to their capacity to personalize learning and offer real-time custom feedback with the potential to support differentiated instruction in PBL contexts ( Asrifan et al., 2024 ; Kong et al., 2024 ; Alam, 2023 ) . By automating routine tasks and augmenting project workflows, these tools could help streamline classroom management, thereby supporting teachers in delivering more engaging learning experiences ( Zheng et al., 2024b ; Krajcik and Blumenfeld, 2006 ) . Given the potential of LLMs to transform PBL, it is essential that these tools are designed prioritizing educators’ needs and experiences, with teachers actively involved in the design process ( Penuel et al., 2007 ; Bakah et al., 2012 ; Wan et al., 2021 ; Handelzalts, 2019 ; McKenney et al., 2016 ) . 1.1. Our Contribution To the best of our knowledge, this is the first co-design study with K-12 teachers aimed specifically at incorporating LLMs into PBL pedagogy. Given the unprecedented and rapidly advancing nature of GenAI and its ethical considerations, our research is particularly timely. Our research objectives (RO) include: (1) Explore Demands and Challenges in PBL: We investigated specific challenges educators encounter when implementing PBL in their classrooms, focusing on how they assess student learning outcomes in areas such as teamwork, creativity, and problem-solving. (2) Co-design GenAI Tools for PBL: Next, we explored how GenAI-powered tools, co-designed with interdisciplinary educators, could address diverse needs and challenges of PBL environments. Our co-design workshops examined how these tools could alleviate teachers’ workloads while supporting their current practices, pedagogical goals, and professional growth. We constructed wireframes embodying teachers’ values and received iterative feedback, from teachers with diverse expertise, on their potential integration into PBL environments. (3) Outline Design Considerations for LLM Tools in PBL: Finally, we propose a set of design guidelines for LLM tools that educational technology designers and educators may find valuable for assessing student progress, setting learning goals, providing iterative and personalized feedback, and managing resources, timelines, and artifacts across different student projects and milestones. Our findings provide valuable insights into shaping the future use of LLMs in PBL, guiding and encouraging its responsible implementation in classrooms. It should be noted that although we used the term GenAI for the study framing with our participants, they focused on a particular type of GenAI: LLMs for their proposed solutions. Because of this duality, both these terms are used in this paper. 2. RELATED WORK 2.1. Project-Based Learning Rooted in constructivist learning theories, project-based learning (PBL) places students at the center of the learning process, challenging conventional education practices and redefining the roles of teaching, learning, and school organization ( Fitzgerald, 2020 ; Guo et al., 2023 ; Kokotsaki et al., 2016 ; Woods et al., 2024 ) . Unlike problem-based learning, which promotes promotes deductive reasoning ( Condliffe, 2017 ; Thomas et al., 1999 ) , and involves structured evaluation based on how well a student’s solution addresses a defined problem (e.g., designing a more effective trash can to improve adoption or conducting participatory action research on a social issue), project-based learning engages students in complex, real-world tasks to create a broader range of artifacts (e.g. reports, models, and presentations) that demonstrate learning ( Krajcik and Blumenfeld, 2006 ; Oguz-Unver and Arabacioglu, 2014 ; Thomas, 2010 ; Chen and Yang, 2019 ; Barron and Darling-Hammond, 2008 ; Thomas et al., 1999 ) (e.g., using

Read the complete text