Breadcrumb
- Home
- AI Hub For Education
- Research Study Repository
- Elementary (PK5)
Search and Filter
The Repository
Elementary (PK5)
Research synthesis is AI-generated, human reviewed. Updated 11/2026.
Displaying 1 - 30 of 390
When Evaluators Cry Wolf: Lessons From Production LLM-As-Judge Evaluation In Educational AI
Chris Rohlfs, Rodrigo Vergara Bosse, Priscilla Rain Hopper, Keanon O'Keefe, Patrick Russell. (09/2026). arXiv.
Across 21 deployed evaluators, the calibration strategy cut confirmed false-positive flags by 99% and raised per-flag precision from 0.6% to 48.9%.
Characterizing LLM-Based Family Education Through The Lens Of Activity Theory: A Scoping Review Of The Hci Literature
Lan Luo, Yuqi Liang, Jie Cai, Anqi Wang, Dongyijie Primo Pan, Muzhi Zhou, Chun Yu, Pan Hui. (09/2026). arXiv.
Most reviewed studies focused on child–parent interaction (83%) and centered on language, narrative, AI literacy, and relational development rather than mathematical or scientific inquiry.
"We'Ll Fix It Later": Education, AI, And The Deferral Of Privacy In EdTech
Meghna Manoj Nair, Rachel Greenstadt. (09/2026). arXiv.
One-third of audited platforms made no meaningful AI disclosure despite visible AI features, and 73% provided only generic accountability and breach-response language.
Evaluation Of Pre-Trained Models For Pedagogical Assessment Of Novel AI-Assisted Educational Questions
Michael Lawrence Castanares, Princess Ventures, Allan Tan. (09/2026). arXiv.
Traditional machine learning models and BERT suffered substantial drops in Bloom classification performance on out-of-distribution AI questions, while large language models maintained higher baseline accuracy.
AI-Assisted Social Story Intervention For Special Education: The Design Of Adapted Stories
Buyankhishig Enkhjargal, Himanshi Lalwani, Hanan Salam. (09/2026). arXiv.
AI assistance suggests it can lower the burden of preparing Social Stories while supporting more individualized story creation.
Dose, Not Access: Digital Reading Tutor Use And Early Reading Growth
Nicholas A. Gage. (09/2026). EdArXiv.
High-dose users averaging at least 15 minutes weekly outperformed near-zero users on end-of-year DIBELS reading composite scores by a medium effect size.
Instructional Objectives Completed And Skill-Specific Early Reading Growth In A Digital Reading Tutor
Nicholas A. Gage. (09/2026). EdArXiv.
Each additional 10 objectives completed predicted a 3.15 percentile point gain on composite reading scores (g = 0.10).
I Code Or AI Code: A Comparative Evaluation Of AI-Rated Scores In Classroom Observations
Yasmin Fong, Jane Xiang, Tak-Yue Dickson Chan, Kerry Lee, Eva Yi Hung Lau. (09/2026). arXiv.
AI and human ratings showed moderate agreement across CLASS dimensions (weighted kappa of 0.681), with AI scores typically falling within one scale point of human ratings.
Teaching AI, Robotics, & Community: A Hubs-Based K-12 Education Framework For Reaching Rural Schools
Maxwell J. Jacobson, Gustavo Rodriguez-Rivera, Petros Drineas, Yexiang Xue. (09/2026). arXiv.
A simulation under moderate conditions projected the framework would reach 74% of Indiana's 1,925 public K-12 schools and create 992 robotics programs over 40 years, compared to 161 under natural growth.
Human Resilience In The AI Era -- What Machines Can'T Replace
Shaoshan Liu, Anina Schwarzenbach, Yiyu Shi. (09/2026). arXiv.
Resilience operates across psychological, social, and organizational levels, combining individual adaptation with peer support and institutional recovery processes.
Alternative AI Philosophy: Daoism As Method For AI In Education
Qin Xie. (09/2026). arXiv.
Dao nature provides an alternative epistemic aim that reframes AI's opacity, urging learners to acknowledge cognitive boundaries instead of pursuing calculative mastery.
Tinker Tales: A Tangible Dialogue System For Child-AI Co-Creative Storytelling
Nayoung Choi, Jiseung Hong, Peace Cyebukayire, Ikseon Choi, Jinho D. Choi. (09/2026). arXiv.
Educationally scaffolded questions elicited narrative contributions in 90% to 100% of exchanges, compared to only 37% for general elaboration questions.
Exploring K-12 Teachers' Perceptions Of Students' Relationships With AI Companions: Boundaries, Intervention Strategies, And Design Implications
Qing Xiao, Wenhan Xie, Ziyu Deng, Ruiwei Xiao, Ziyue Feng, Xie He, Shiyu Zhang, John Stamper, Hong Shen, Xinying Hou. (09/2026). arXiv.
Most teachers found AI tutors and study buddies acceptable, while intimate roles like romantic partners, friends, or parents raised concerns about eliminating developmental friction.
Artificial Intelligence Literacy And Sustainable Development: An Ethical Governance And Development Goals Framework
Md. Masudul Islam, Mirza Niaz Morshed, Md. Shafiqul Islam. (09/2026). arXiv.
Governance literacy was the strongest predictor of sustainable development goal nexus awareness (standardized beta = 0.64), followed by ethical literacy and technical literacy.
An Emancipatory Vision For Designing (Generative) AI For Learner Flourishing
Luis Pablo Prieto, Yannis Dimitriadis. (09/2026). arXiv.
Current AI in education often relies on problematic assumptions, including treating agency as a zero-sum game, overestimating predictability in complex systems, and presuming infinite resources for permanent scaffolding.
The Policy Deficit In AI X Social-Emotional Learning Research
Tran Van Cuong, Liu Yihan, Nguyen Van Tuong. (08/2026). arXiv.
Nearly three-quarters of reviewed studies (72%) did not mention any substantive consequences or implications for policy governance.
Methodologies For Improving The Quality Of AI Tutoring In K-12 Education
Tushar Udeshi, Anna Khazenzon, Kabir Khan, Nick Breen, RJ Corwin, Chris DiGiano, Kodi Weatherholtz, Marek Zaluski. (08/2026). arXiv.
Migrating the main Exercise AI tutor from GPT-4o to GPT-4.1 reduced giving away final answers by 96.66% and increased behavioral engagement by 9.93%.
Rethinking Generative AI Literacy: An Integrative, Developmental, And Dialectical Framework For K-12 Teacher Education
Shahin Hossain, Sima Ahmadi, Leqi Li, Idowu David Awoyemi, Wei Huang, Chenxi Zhou, Jujia Li, Samaa Haniya, Shapla Khanam, Tasbirun Mashreka Subaha. (08/2026). arXiv.
Most reviewed educational AI literacy studies focus on knowledge (97%) and skills (96%), whereas equity (73%) and agency (60%) receive comparatively thin treatment.
Eduzone: A Framework For Evaluating LLM Safety For K-12 Students And Teachers
Junyeong Park, Jieun Han, Haneul Yoo, So-Yeon Ahn, Jinsung Yoon, Alice Oh. (08/2026). arXiv.
Models were more vulnerable to education-specific risks than conventional harms, with average attack success rates reaching 80.3% for student academic misconduct and 73.0% for excessive cognitive load.
When LLM Essays Outscore Student Essays: What A Korean Writing Rubric Rewards And Where Readers Disagree
Shinwoo Park, Yo-Sub Han. (08/2026). arXiv.
Large language model essays outscored student essays by an average of 18.35 points across a 100-point rubric, exceeding student averages across all four evaluated models.
Virtual Tutoring with Computer-Assisted Learning: An Experiment in Take-Up and Learning
Philip Oreopoulos, Ruochong Dong, Nina Low. (08/2026). NBER.
Offering the tutoring raised Khan Academy practice by about 10 minutes weekly and led to an average math assessment score gain of 0.055 to 0.06 standard deviations.
Human-LLM Collaborative Inductive Coding For Conceptualizing K-12 Educator AI Use
Alex Liu, Min Sun, Lief Esbenshade, Michael Xiao, Victor Tian, Zachary Zhang, Kevin He. (07/2026). arXiv.
Collaborative inductive coding produced a final hierarchical codebook of 72 items across 19 categories and six domains, after human coders identified five missing codes.
Agreement Is Not Quality: Blind Expert Verification Of Human And LLM Qualitative Coding When Human Consensus Is Not Ground Truth
Alex Liu, Lief Esbenshade, Michael Xiao, Victor Tian, Zachary Zhang, Kevin He, Min Sun. (07/2026). arXiv.
A blind expert preferred human and model coding at statistically indistinguishable rates (51.5% versus 48.5%), despite models having lower agreement with human coders than humans had with each other.
Beyond Direct Answering: Aligning Educational LLMs As Socratic Guides Via Heuristic Reinforcement Learning
Xiaokun Wang, Siyu Song, Wentao Liu, Xiaodong Zou. (07/2026). arXiv.
Reinforcement learning without a directness penalty raised scaffolding effectiveness to 63.3% and cut answer keyword leakage to 13.3% on 30 test questions.
AI-Integrated Scientific Inquiry: A Practice-Centered Vision For Science Education
Arne Bewersdorff, Matias Rojas, Xiaoming Zhai. (07/2026). arXiv.
Artificial intelligence can be integrated into science education as simplified, pedagogically bounded scientific tools mapped directly onto established Next Generation Science Standards practices.
Generative AI And Agency In Education: A Critical Scoping Review And Thematic Analysis
Jasper Roe, Mike Perkins. (07/2026). arXiv.
Literature on educational generative AI focuses on three primary themes: control in digital spaces, variable engagement and access, and changing notions of human-AI agency.
Faster Completion, Less Learning: Generative AI Reduced Study Time On Math Problems And The Knowledge They Build
Sina Rismanchian, Hasan Uzun, Jeffrey Matayoshi, Eric Cosyn, Eyad Kurd-Misto. (07/2026). arXiv.
Learning time on AI-susceptible math word problems dropped cumulatively by 31.3% for high schoolers and 26.9% for college students after ChatGPT's release, with no change in Grade 5.
L2-Bench: An Evaluation Benchmark For Measuring LLM Capabilities In Second Language Education
James Edgell, Wm. Matthew Kennedy, Ben Knight, Danielle Carvalho, Martin Ku, Isaac Pattis. (07/2026). arXiv.
Across 1,000 tasks, Claude Opus 4.7 achieved the highest overall score at 85.5%, followed by GPT 5.4 at 84.1% and Gemini 3.1 Pro at 83.4%.
AIED's Unfinished Mission: Centering Agency And Motivation In The Age Of Effortless Bypass
H. Chad Lane. (07/2026). arXiv.
Freely available generative AI creates an effortless bypass dilemma by allowing learners to outsource cognitive tasks, which risks undermining self-regulated learning and metacognitive habits.
Why Teaching Resists Automation In An AI-Inundated Era: Human Judgment, Non-Modular Work, And The Limits Of Delegation
Songhee Han. (07/2026). arXiv.
Teaching resists automation because instructional tasks are inherently non-modular and derive their value from interconnected professional judgment across learners, time, and pedagogical contexts.
