Breadcrumb
- Home
- AI Hub For Education
- Research Study Repository
- Evidence Explorer
- AI Tools
- Custom-built Research Tool
The Repository
Search and Filter
Custom-built research tool
Studies tagged with AI from their full text or abstract; see each study for details.
Showing 331 - 360 of 531 studies
Learning by Teaching: Engaging Students as Instructors of Large Language Models in Computer Science Education
Xinming Yang, Haasil Pujara, Jun Li. (08/2025). arXiv.
Undergraduate students using the learning-by-teaching system achieved statistically significant improvements on course assignments (86.4% vs. 82.9%) and projects (85.4% vs. 78.4%) compared to historical cohorts.
Dean of LLM Tutors: Exploring Comprehensive and Automated Evaluation of LLM-generated Educational Feedback via LLM Feedback Evaluators
Keyang Qian, Yixin Cheng, Rui Guan, Wei Dai, Flora Jin, Kaixun Yang, Sadia Nawaz, Zachari Swiecki, Guanliang Chen, Lixiang Yan, Dragan Gašević. (08/2025). arXiv.
Fine-tuning GPT-4.1 on human-labelled scores without rationales achieved the highest agreement with expert reviewers, reaching 79.8% accuracy and a Cohen's kappa of 0.686.
Building Effective Safety Guardrails in AI Education Tools
Hannah-Beth Clark, Laura Benton, Emma Searle, Margaux Dowland, Matthew Gregory, Will Gayne, John Roberts. (08/2025). arXiv.
Oak National Academy deployed four guardrails—pedagogical prompt engineering, input threat detection, an Independent Asynchronous Content Moderation Agent, and teacher-in-the-loop oversight—to protect pupils aged 5 to 16.
Beyond Automation: Socratic AI, Epistemic Agency, and the Implications of the Emergence of Orchestrated Multi-Agent Learning Architectures
Peer-Benedikt Degen, Igor Asanov. (08/2025). arXiv.
Students using the Socratic tutor reported significantly higher motivation for independent thinking compared to those using the uninstructed chatbot.
Personalized Knowledge Transfer Through Generative AI: Contextualizing Learning to Individual Career Goals
Ronja Mehlan, Claudia Hess, Quintus Stierstorfer, Kristina Schaaff. (08/2025). arXiv.
Learners receiving career-tailored scenarios reported slightly higher satisfaction ratings (9.02 versus 8.84 out of 10) than those receiving standard content.
InqEduAgent: Adaptive AI Learning Partners with Gaussian Process Augmentation
Tian-Fang Zhao, Wen-Xi Yang, Guan Liu, Liang Yang. (08/2025). arXiv.
The localized Pareto matching model achieved the highest overall accuracy gain of 30.47 percent, showing a 4.47 percentage point improvement over baseline.
When AIs Judge AIs: The Rise of Agent-as-a-Judge Evaluation for LLMs
Fangyi Yu. (08/2025). arXiv.
Multi-agent debate frameworks such as ChatEval improve alignment with human judgments by 10% to 16% over single-agent prompts on question answering and dialogue benchmarks.
Assessing the Reliability and Validity of Large Language Models for Automated Assessment of Student Essays in Higher Education
Andrea Gaggioli, Giuseppe Casaburi, Leonardo Ercolani, Francesco Collova, Pietro Torre, Fabrizio Davide. (08/2025). arXiv.
Human–model scoring agreement was consistently near zero across all models (Quadratic Weighted Kappa between -0.04 and 0.02) and lacked statistical significance.
A Mixed User-Centered Approach to Enable Augmented Intelligence in Intelligent Tutoring Systems: The Case of MathAlde app
Guilherme Guerino, Luiz Rodrigues, Luana Bianchini, Mariana Alves, Marcelo Marinho, Thomaz Veloso, Valmir Macario, Diego Dermeval, Thales Vieira, Ig Bittencourt, Seiji Isotani. (08/2025). arXiv.
In the classroom study, teachers used the correction feature on 139 of 784 student answers (17.6%) to overturn incorrect AI assessments.
A Theory of Adaptive Scaffolding for LLM-Based Pedagogical Agents
Clayton Cohn, Surya Rayala, Namrata Srivastava, Joyce Horn Fonteles, Shruti Jain, Xinying Luo, Divya Mereddy, Naveeduddin Mohammed, Gautam Biswas. (08/2025). arXiv.
Inquizzitor's scoring accuracy matched or surpassed human agreement across all three formative assessments, achieving quadratic-weighted kappas of 86.63, 94.12, and 90.74.
WIP: Enhancing Game-Based Learning with AI-Driven Peer Agents
Chengzhang Zhu, Cecile H. Sam, Yanlai Wu, Ying Tang. (08/2025). arXiv.
Students completing agent-supported modules with language model inquiry improved test scores from 39.2% to 45.7%, whereas passive video instruction module scores dropped from 50.0% to 28.3%.
AI-Educational Development Loop (AI-EDL): A Conceptual Framework to Bridge Al Capabilities with Classical Educational Theories
Ning Yu, Jie Zhang, Sandeep Mitra, Rebecca Smith, Adam Rich. (08/2025). arXiv.
Student performance improved significantly between their initial and revised submissions following AI feedback, with teacher-assigned grades increasing from a mean of 1.56 to 1.77.
Beyond Agreement: Rethinking Ground Truth in Educational AI Annotation
Danielle R. Thomas, Conrad Borchers, Kenneth R. Koedinger. (07/2025). arXiv.
Comparative judgment improved raters' accuracy by 13% and inter-rater agreement from 0.66 to 0.80 on reading short answers, matching or exceeding expert consistency.
Personalized Education with Ranking Alignment Recommendation
Haipeng Liu, Yuxuan Liu, Ting Long. (07/2025). arXiv.
Across all five simulated environments, both Ranking Alignment Recommendation variants achieved the highest or second-highest learning effects compared to baseline methods.
Automated Feedback on Student-Generated UML and ER Diagrams Using Large Language Models
Sebastian Gürtl, Gloria Schimetta, David Kerschbaumer, Michael Liut, Alexander Steinmaurer. (07/2025). arXiv.
All six participants highlighted the tool's ability to provide instant, iterative, and scalable feedback to support self-directed student learning without increasing tutor workload.
AI Literacy as a Key Driver of User Experience in AI-Powered Assessment: Insights from Socratic Mind
Meryem Yilmaz Soylu, Jeonghyun Lee, Jui-Tse Hung, Christopher Zhang Cui, David A. Joyner. (07/2025). arXiv.
AI literacy significantly predicted perceived usability, satisfaction, and engagement, but it did not directly predict perceived learning effectiveness.
A Comprehensive Review of AI-based Intelligent Tutoring Systems: Applications and Challenges
Meriem Zerkouk, Miloud Mihoubi, Belkacem Chikhaoui. (07/2025). arXiv.
Intelligent tutoring systems can improve student performance by 20%, showing learning gains comparable to human tutoring in structured domains like mathematics and physics.
Students' Feedback Requests and Interactions with the SCRIPT Chatbot: Do They Get What They Ask For?
Andreas Scholl, Natalie Kiesler. (07/2025). arXiv.
Chatbot responses matched or exceeded student-requested feedback types in 75% of question-response pairs, with exact matches in 47% and over-matches in 28%.
LEKIA: A Framework for Architectural Alignment via Expert Knowledge Injection
Boning Zhao, Yutong Hu. (07/2025). arXiv.
The full LEKIA system improved overall performance by 14.8% over the baseline model that only used theoretical guidelines.
EduThink4AI: Translating Educational Critical Thinking into Multi-Agent LLM Systems
Xinmeng Hou, Zhouquan Lu, Wenli Chen, Hai Hu, Qing Guo. (07/2025). arXiv.
On a GPT-3.5-Turbo backbone, MetaCrit reached 94.12% accuracy on TruthfulQA, 84% on CIAR, and zero toxicity on BOLD and HONEST.
Findings of MEGA: Maths Explanation with LLMs using the Socratic Method for Active Learning
Tosin Adewumi, Foteini Simistira Liwicki, Marcus Liwicki, Viktor Gardelli, Lama Alkhaled, Hamam Mokayed. (07/2025). arXiv.
University students unanimously agreed more often that the interactive MEGA method supported learning better than step-by-step explanations, particularly for difficult MATH benchmark problems (47.5% versus 26.67%).
Partnering with AI: A Pedagogical Feedback System for LLM Integration into Programming Education
Niklas Scholz, Manh Hung Nguyen, Adish Singla, Tomohiro Nagashima. (07/2025). arXiv.
Teachers rated the generated feedback highly for pedagogical soundness (mean 3.84 out of 5), comprehension (4.04), and learning effectiveness (3.89).
Mitigating Trojanized Prompt Chains In Educational Llm Use Cases: Experimental Findings and Detection Tool Design
Richard M. Charles, James H. Curry, Richard B. Charles. (07/2025). arXiv.
High-risk prompt chains bypassed safety guardrails in 50% of trials, with neither model flagging transitions into semantically risky territory at the third conversational turn.
SentiDrop: A Multi-Modal Machine Learning model for Predicting Dropout in Distance Learning
Meriem Zerkouk, Miloud Mihoubi, Belkacem Chikhaoui. (07/2025). arXiv.
A tuned XGBoost model incorporating student sentiment reached 84% accuracy on unseen next-academic-year data, compared to 82% for the baseline model.
Beyond Classical And Contemporary Models: A Transformative Ai Framework For Student Dropout Prediction In Distance Learning Using RAG, Prompt Engineering, And Cross-Modal Fusion
Miloud Mihoubi, Meriem Zerkouk, Belkacem Chikhaoui. (07/2025). arXiv.
The TRIAD-Drop model achieved 89% accuracy and an F1-score of 0.85, outperforming the strongest classical baseline by seven percentage points.
On the development of an AI performance and behavioural measures for teaching and classroom management
Andreea Ioana Niculescu, Jochen Ehnes, Chen Yi, Du Jiawei, Tay Chiat Pin, Joey Tianyi Zhou, Vigneshwaran Subbaraju, Teh Kah Kuan, Tran Huy Dat, Gi Soong Chee, Kenneth Kwok. (07/2025). arXiv.
A pretrained action recognition model achieved classification accuracies of 83.33% for board writing, 88.57% for board gesturing, and 65.06% for pointing.
SimStep: Chain-of-Abstractions for Incremental Specification and Debugging of AI-Generated Interactive Simulations
Zoe Kaputa, Anika Rajaram, Vryan Almanon Feliciano, Zhuoyue Lyu, Maneesh Agrawala, Hari Subramonyam. (07/2025). arXiv.
Participating educators rated SimStep's usability positively with an overall score of 4.66 out of 6, reporting it felt intuitive and natural to use.
Enhancing Essay Cohesion Assessment: A Novel Item Response Theory Approach
Bruno Alexandre Rosa, Hilário Oliveira, Luiz Rodrigues, Eduardo Araujo Oliveira, Rafael Ferreira Mello. (07/2025). arXiv.
Combining BERT models with item response theory score adjustments achieved the highest overall agreement on both Brazilian high school and 5th-grade narrative essay datasets.
Conversational Education at Scale: A Multi-LLM Agent Workflow for Procedural Learning and Pedagogic Quality Assessment
Jiahuan Pei, Fanghua Ye, Xin Sun, Wentao Deng, Koen Hindriks, Junxiao Wang. (07/2025). arXiv.
The workflow achieved a 95.16% completion rate and strong topical relevance across models, but active learner question generation varied widely.
AGACCI : Affiliated Grading Agents for Criteria-Centric Interface in Educational Coding Contexts
Kwangsuk Park, Jiwoong Yang. (07/2025). arXiv.
AGACCI achieved 12 percentage points higher average rubric-level accuracy than the single-model baseline across 360 graduate-level programming submissions.
