Breadcrumb
- Home
- AI Hub For Education
- Research Study Repository
- Evidence Explorer
- Methods
- Multilevel / Growth Modeling
The Repository
Search and Filter
Multilevel / growth modeling
Studies tagged with AI from their full text or abstract; see each study for details.
Showing 1 - 18 of 18 studies
Evaluating AI Tutoring At The Speed Of Innovation: Practitioner-Led Micro-Randomised Trials Of An AI Tutoring Platform In Gcse Science
Wayne Harrison, Rahil Khowaja, Emma Dobson, Germaine Uwimpuhwe, Steve Higgins. (09/2026). arXiv.
Students allocated to the AI tutoring platform scored significantly higher on science post-tests than those doing business-as-usual revision, with an effect size of 0.33.
Llama Lima: A Living Meta-Analysis On The Effects Of Generative AI On Learning Mathematics
Anselm Strohmaier, Samira Bodefeld, Oliver Straser, Frank Reinhold. (09/2026). arXiv.
Generative AI interventions showed an overall positive effect on mathematics learning outcomes (Hedges' g = 0.57), with no evidence of publication bias.
Semantic Variability Of LLM-Generated Replies Across LLMs: Implications For Designing Conversation-Based Assessment
Jiangang Hao. (08/2026). arXiv.
Models generated replies more semantically similar to their own outputs (similarity 0.715 to 0.795) than to other models (0.443 to 0.604).
Making AI-Generated Feedback Matter: A Large-Scale Study Of Feedback Workflows And Student Enactment
Omar Alsaiari, Nilufar Baghaei, Jason M. Lodge, Naomi Winstone, Dragan Gašević, Hassan Khosravi. (08/2026). arXiv.
Enacted Feedback was associated with significantly higher feedback uptake (estimated probability of 26.2%) compared with Directed Feedback (14.1%) and Self-Directed Feedback (0.1%).
Knowledge without Wisdom: Measuring Misalignment between LLMs and Intended Impact
Michael Hardy, Yunsung Kim. (04/2026). arXiv.
Language models agreed more strongly with one another than with human experts, but their ratings often negatively aligned with actual student learning gains.
Comparing The Impact Of Pedagogy-Informed Custom And General-Purpose Gai Chatbots On Students' Science Problem-Solving Processes And Performance Using Heterogeneous Interaction Network Analysis
Hanyu Su, Huilin Zhang, Shihui Feng. (04/2026). arXiv.
Students showed significantly higher interaction intensity and greater cognitive interaction diversity when using the custom Socratic chatbot compared to the general-purpose chatbot.
Evaluating A Data-Driven Redesign Process For Intelligent Tutoring Systems
Qianru Lyu, Conrad Borchers, Meng Xia, Karen Xiao, Paulo F. Carvalho, Kenneth R. Koedinger, Vincent Aleven. (03/2026). arXiv.
Overall test-score learning gains did not differ significantly between students using the redesigned tutor and those using the original tutor.
From School AI Readiness To Student AI Literacy: A National Multilevel Mediation Analysis Of Institutional Capacity And Teacher Capability
Xiu Guan, Mingmin Zheng, Dragan Gašević, Wenxin Guo, Yingqun Liu, Xibin Han, Danijela Gašević, Ruiling Ma, Qi Wu, Lixiang Yan. (03/2026). arXiv.
Overall school AI readiness was positively associated with student AI literacy after adjusting for institutional and regional characteristics (beta = 0.016).
Llama Lima: A Living Meta-Analysis On The Effects Of Generative AI On Learning Mathematics Version 2, 03/26
Anselm Strohmaier, Samira Bodefeld, Oliver Straser, Frank Reinhold. (03/2026). arXiv.
Instructional interventions using generative AI show an overall positive effect on mathematics learning outcomes (Hedges’ g = 0.57).
Multilingual Large Language Models Do Not Comprehend All Natural Languages To Equal Degrees
Natalia Moskvina, Raquel Montero, Masaya Yoshida, Ferdy Hubers, Paolo Morosi, Walid Irhaymi, Jin Yan, Tamara Serrano, Elena Pagliarini, Fritz Gunther, Evelina Leivada. (02/2026). arXiv.
Models fell behind human baselines in all languages except Spanish for GPT-4o and Italian for DeepSeek-V3, where model accuracy was comparable to humans.
Are They Just Delegating? Cross-Sample Predictions On University Students' & Teachers' Use Of AI
Fabian Albers, Sebastian Straub, Nikol Rummel, Nils Kobis. (01/2026). arXiv.
Students reported using AI more frequently and delegating tasks to AI to a greater extent than teachers across five of six academic tasks, with translation being the only exception.
Personalized Multimodal Feedback Using Multiple External Representations: Strategy Profiles And Learning In High School Physics
Natalia Revenga-Lozano, Karina E. Avila, Steffen Steinert, Matthias Schweinberger, Clara Gómez-Pérez, Jochen Kuhn, Stefan Küchemann. (01/2026). arXiv.
Using elaborated multirepresentational feedback was positively associated with post-test scores in both student cohorts, regardless of prior knowledge or confidence.
The Social Blindspot In Human-Ai Collaboration: How Undetected Ai Personas Reshape Team Dynamics
Lixiang Yan, Xibin Han, Yu Zhang, Samuel Greiff, Inge Molenaar, Yizhou Fan, Roberto Martinez-Maldonado, Linxuan Zhao, Xinyu Li, Yueqiao Jin, Dragan Gašević. (12/2025). arXiv.
Participants showed limited ability to detect AI teammates, correctly identifying them only 30.8% of the time, while misclassifying human partners as AI in 27.6% of cases.
Measuring Teaching With Llms
Michael Hardy. (10/2025). arXiv.
Three sentence-embedding models achieved human-level performance, with correlations exceeding 0.65 for general teaching practices and surpassing average human rater correlations for math instruction.
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.
Detecting LLM-Generated Short Answers and Effects on Learner Performance
Shambhavi Bhushan, Danielle R. Thomas, Conrad Borchers, Isha Raghuvanshi, Ralph Abboud, Erin Gatz, Shivang Gupta, Kenneth R. Koedinger. (06/2025). arXiv.
A fine-tuned GPT-4o model detected AI-generated short answers with 80% accuracy, outperforming the commercial tool GPTZero, which reached 70% accuracy.
Evaluating Trust in AI, Human, and Co-produced Feedback Among Undergraduate Students
Audrey Zhang, Yifei Gao, Wannapon Suraworachet, Tanya Nazaretsky, Mutlu Cukurova. (04/2025). arXiv.
Under blind evaluation, students rated AI and co-produced feedback significantly higher than human teaching assistant feedback for usefulness and objectivity.
Event Segmentation Applications In Large Language Model Enabled Automated Recall Assessments
Ryan A. Panela, Alexander J. Barnett, Morgan D. Barense, Bjornn Herrmann. (02/2025). arXiv.
GPT-4 with lower randomness parameters identified narrative event boundaries that aligned more closely with human responses than humans agreed among themselves.
