LearnMate : Enhancing Online Education with LLM-Powered Personalized Learning Plans and Support
Xinyu Jessica Wang, Christine P Lee, Bilge Mutlu · 2025
Read LearnMate : Enhancing Online Education with LLM-Powered Personalized Learning Plans and Support by Xinyu Jessica Wang, Christine P Lee, Bilge Mutlu (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-2503-13340.txt.
Introduction
Xinyu Jessica Wang, Christine Lee, and Bilge Mutlu present LearnMate in a 2025 paper on personal online learning. Online courses serve wide mixes of learners, and one size plans fit few of them well. Large language models offer a new way to tailor help to each person.
Guides drawn from past personal learning research ground the model in sound practice. LearnMate turns those guides into a working system that builds a plan per learner. It also gives live support as the learner works through the plan. The design aims to move past the fixed path of classic online courses. The authors close with what the system means for course design and where it should head next.
Catalog record
| CREATOR | Xinyu Jessica Wang, Christine P Lee, Bilge Mutlu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-13340
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
LearnMate : Enhancing Online Education with LLM-Powered Personalized Learning Plans and Support Xinyu Jessica Wang, Christine P Lee, Bilge Mutlu Abstract . With the increasing prevalence of online learning, adapting education to diverse learner needs remains a persistent challenge. Recent advancements in artificial intelligence (AI), particularly large language models (LLMs), promise powerful tools and capabilities to enhance personalized learning in online educational environments. In this work, we explore how LLMs can improve personalized learning experiences by catering to individual user needs toward enhancing the overall quality of online education. We designed personalization guidelines based on the growing literature on personalized learning to ground LLMs in generating tailored learning plans. To operationalize these guidelines, we implemented LearnMate , an LLM-based system that generates personalized learning plans and provides users with real-time learning support. We discuss the implications and future directions of this work, aiming to move beyond the traditional one-size-fits-all approach by integrating LLM-based personalized support into online learning environments. Conference: Extended Abstracts of the CHI Conference on Human Factors in Computing Systems; April 26-May 1, 2025; Yokohama, Japan Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA ’25), April 26-May 1, 2025, Yokohama, Japan DOI: 10.1145/3706599.3719857 ISBN: 979-8-4007-1395-8/2025/04 CCS: Human-centered computing Interactive systems and tools CCS: Applied computing Interactive learning environments CCS: Computing methodologies Natural language processing 1. Introduction Online learning has become ubiquitous, making online learning courses and platforms accessible to a wide range of learners. Learners include people who lack access to traditional educational tools, children interested in extracurricular activities, and adults who seek continuing education. Massive Open Online Courses (MOOCs) have become particularly popular due to their ability to bring high-quality, low-cost education to large numbers of students ( Alseddiqi et al., 2023 ) . As online learning becomes increasingly accessible to diverse populations, there has been growing emphasis on personalizing learning plans to optimize educational outcomes ( Basham et al., 2016 ; Bernacki et al., 2021 ; Bray and McClaskey, 2013 ) . Personalized learning involves several approaches including (1) technology-based personalization, such as intelligent tutoring systems that adapt to student performance; (2) classroom-based personalization, exemplified by self-paced learning with teacher support; and (3) system-level personalization, which enables multiple pathways for knowledge demonstration ( Bernacki et al., 2021 ) . The need for personalized learning is well-documented in existing literature. For example, Bernacki et al. (2021) identify several driving factors: the growing diversity of learners; the limitations of traditional “factory model” approaches that fail to address individual variability; recent educational policies calling for increased personalization; and the specific needs of students with disabilities whose complex learning profiles demand individualized approaches. However, effectively supporting personalized learning in online learning platforms is still at its early development stage, lacking standardization in implementation and coherent theoretical frameworks ( Bernacki et al., 2021 ; Basham et al., 2016 ; Bray and McClaskey, 2013 ; Murtaza et al., 2022 ) . Moreover, giving users sufficient control to accommodate their diverse personalized needs and providing accurate and real-time support based on these needs remain open challenges ( Bernacki et al., 2021 ; Zhang et al., 2023 ) . Recent advancements in AI have catalyzed growing interest in leveraging AI for personalized learning experiences ( Pataranutaporn et al., 2021 ; Ayeni et al., 2024 ; Jian, 2023 ; Davuluri, 2021 ) . Following the emergence of LLMs, researchers have begun exploring their specific applications in personalized learning and individualized support ( Park et al., 2024 ; Wen et al., 2024 ; Laak and Aru, 2024 ; Nam et al., 2024 ; Hsu et al., 2024 ) . Recent LLMs ( Achiam et al., 2023 ; Anil et al., 2023 ; Anthropic, 2024 ) demonstrate remarkable capabilities across diverse tasks, such as common sense reasoning, adaptation to user-specific needs, and mathematical and visuolinguistic problem-solving ( Kwon et al., 2024 ; Krause and Stolzenburg, 2023 ; Lee et al., 2025 ; Didolkar et al., 2024 ; Cherian et al., 2025 ) . With carefully crafted prompting strategies, LLMs show particular promise in generating personalized learning plans in online education ( Ng and Fung, 2024 ; Brown et al., 2020 ; Yao et al., 2024 ; Ma et al., 2024 ) . Additionally, AI-powered conversational services like ChatGPT ( Achiam et al., 2023 ) have emerged as powerful tools to address persistent challenges in MOOCs, especially in delivering personalized feedback and support to students. Recent studies indicate that these models can effectively facilitate various educational tasks, from providing subject-specific guidance to fostering student engagement ( Alseddiqi et al., 2023 ; Kayalı et al., 2023 ; Adetayo et al., 2024 ) . In this paper, we investigate how LLM capabilities can facilitate personalized learning and user support in online educational environments. Our investigation is guided by two research questions: (1) How should LLMs provide personalized self-paced learning methods and resources based on user needs and preferences; and (2) how can LLMs effectively support user learning by offering real-time contextual assistance based on their different needs? To address the first research question, we develop personalization guidelines for LLMs to use when generating learning plans, drawing from the growing literature on personalized learning ( Graham et al., 2019 ; Bernacki et al., 2021 , e.g. ,) . These guidelines focus on four key dimensions: goals , time , pace , and learning paths . To operationalize these guidelines and answer the second research question, we present an LLM-based system, called LearnMate , that generates personalized learning plans and provides real-time support to users by monitoring learning progress, tracking study materials, and offering contextual responses to learner queries. Finally, we discuss the potential implications and future directions of this work. 2. Related Works Personalization in online learning continues to grow in popularity, with numerous successful system implementations ( Bernacki et al., 2021 ; Tapalova and Zhiyenbayeva, 2022 ; Alamri et al., 2021 ) . For example, platforms like Knewton Alta create individualized learning pathways by analyzing student performance and adjusting content delivery in real-time ( Tapalova and Zhiyenbayeva, 2022 ) . Systems like PLATO and McGraw-Hill’s SmartBook demonstrate how adaptive technologies can analyze learning patterns, progress data, and engagement metrics to customize educational experiences, resulting in improved student retention rates from 70.1% to 89.9% ( Alamri et al., 2021 ) . Despite technological advances, personalized learning faces significant challenges in personalized guidance, real-time support, and companionship for students to maintain engagement in online environments ( Anshari et al., 2016 ; García-Morales et al., 2021 ; Murtaza et al., 2022 ) . Moreover, giving users enough control to accurately reflect their diverse personalized needs, and for online learning platforms to provide accurate and real-time support based on these needs is often a challenge ( Bernacki et al., 2021 ) . Recent research has explored LLMs for enhancing personalized learning through intelligent tutoring systems. For example, models like PLPP and ALEDU demonstrate significant potential in creating adaptive learning interfaces that generate dynamic, contextually-appropriate educational content ( Wen et al., 2024 ; Ng and Fung, 2024 ) . By leveraging natural language understanding and generation capabilities, LLMs can provide personalized, pedagogically sound learning paths tailored to individual student understanding levels ( Ng and Fung, 2024 ) , improving