Breadcrumb
- Home
- AI Hub For Education
- Research Study Repository
- Outcomes – Numeracy
Search and Filter
The Repository
Outcomes – Numeracy
Research synthesis is AI-generated, human reviewed. Updated 11/2026.
Displaying 1 - 30 of 259
Not Faster, But Better: Generative AI And Teacher Skill In Peru
Carla Z. Glave, Carolina Lopez, Ezequiel Molina, Roman Andres Zarate. (10/2026). World Bank.
Providing AI licenses, training, and follow-up sharply increased AI adoption but saved teachers no time on any of their core tasks.
Teacher-Mediated Multimodal Platform And Math Achievement: Evidence From Goblins' Yearlong District Implementation
Anika Alam, Nicolas Dominguez-Mel, Jing Liu. (10/2026). EdWorking Papers.
Using the Goblins platform is associated with a 0.11 standard deviation increase in MAP math scores within the district.
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.
Studentbench: AI And Human Tutoring Yield Equivalent Gre Learning Gains
Curtis Northcutt, Inaara Hasmani, Kevin Feng, Trevor Khangi, Andreas Plesner, Jonas Mueller. (09/2026). arXiv.
AI tutoring produced learning gains statistically equivalent to expert human tutoring on the combined GRE assessment (p=.015).
When AI Tutors Speak: Evidence From A Randomized Field Experiment
Shihao Yang, Marshall Van Alstyne, Chrysanthos Dellarocas. (09/2026). arXiv.
Students assigned to the structured AI tutor gained 6.63 more points on a 55-point assessment than peers in the holdout group.
Little Impact Of ChatGPT On High School Test Scores
Nick Huntington-Klein. (09/2026). arXiv.
Higher educational ChatGPT use had an insignificant effect on high school test scores that was close to zero, bounded below roughly 0.04 standard deviations.
From Content Generation To Learning Support: Pedagogy-Guided Generative Video Tutors For STEM Learning
Xinchen Ma, Shuimu Wang, Gaole He, Yanbin Zhang, Chunyang Wang, Yunshi Lan, Weining Qian. (09/2026). arXiv.
The PIVOT framework generates educational videos with pedagogically aligned content, clear and engaging presentation, and coherent instructional flow across four STEM domains.
A Brief AI Literacy Intervention Does Not Significantly Reduce Over-Reliance And Increases Under-Reliance On ChatGPT: A Randomized Study
Brett Puppart, Jaan Aru. (09/2026). arXiv.
A brief AI literacy text did not significantly reduce over-reliance on incorrect ChatGPT advice, with incorrect recommendations accepted in 48.7% of intervention trials versus 55.9% in control trials.
Mathematics Graduate Training In The Age Of AI
David Glickenstein. (09/2026). arXiv.
Generative AI shifts the conditions under which graduate mathematics is learned, written, assessed, and defended.
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.
Statistical Theory In The Age Of Machine-Assisted Mathematics: Rethinking How Theory Is Made And Taught
Pietro Coretto. (09/2026). arXiv.
Machine formalization identified 32 mathematical conditions across five statistical cases, linking each condition to specific proof steps that informal textbook citations usually compress.
Beyond Helpfulness: A Teaching-Over-Solving Diagnostic For Measuring Educational Impact In LLM Tutors
Junyi Yao, Zihao Zheng, Baichuan Li. (08/2026). arXiv.
Problem-solving and pedagogical scores were only moderately correlated across eight AI models (Pearson correlation of 0.421, Spearman rank correlation of 0.690).
Gen AI In Proof-Based Math Courses: A Pilot Study
Hannah Klawa, Shraddha Rajpal, Cigole Thomas. (08/2026). arXiv.
Most students found Microsoft Copilot helpful for finding information and understanding concepts, but 8 of 19 reported not using it and 7 viewed using AI as cheating.
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%.
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.
One Click Away: AI Tutoring with Khanmigo in a Two-Year School Experiment
Philip Oreopoulos, Nina Low. (08/2026). NBER.
Assignment to Khanmigo raised math achievement by 1.3 national percentile ranks per term, or about 0.06 to 0.08 standard deviations over a school year.
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.
Making AI Tutoring Productive: Evidence from a Mastery-Based Math Practice Experiment
Philip Oreopoulos, Michael Liut, Alp Sungu, Nina Low. (08/2026). NBER.
AI tutoring students progressed more slowly and attempted fewer questions, but they answered more accurately once reaching an attempt.
What Remains Human In Mathematics In The Age Of AI
Amir Moradifam. (07/2026). arXiv.
As routine execution and computation become automated, distinctively human mathematical skills such as problem framing, conceptual insight, and creative judgment become more visible and valuable.
Generative AI Can Harm Teaching
Alp Sungu, Benjamin Lira, Angela Lee Duckworth. (07/2026). SSRN.
Providing teachers with an AI support tool reduced student intrinsic motivation by 0.11 standard deviations, with students rating courses as less enjoyable, interesting, and important.
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.
From Answer Generators To Reasoning Facilitators: Designing AI Tutors For Mathematical Reasoning In High-Stakes Environments
Harry Feng, Yuan Tian, Erica Zhao. (07/2026). arXiv.
Students actively resisted unskippable Socratic dialogue, using final answers as diagnostic checkpoints to decide whether and where to inspect step-by-step reasoning.
The Generative AI Learning Penalty: Evidence from Chinese Secondary Education
David Stromberg, Victor Lei, Yanhui Wu. (07/2026). SSRN.
AI adoption raised homework scores by 18% and cut completion time by 30%, but lowered monthly exam scores by 20% within six months.
LLM-Generated Summaries For Teachers: A Randomized Field Experiment In A Digital Learning Platform
Wen-Chiang Lim, Eamon Worden, Adam Sales, Neil Heffernan. (06/2026). Proceedings of the 13th ACM Conference on Learning @ Scale (L@S ’26).
Providing AI-generated summaries significantly increased student assignment start and completion rates by about 3 to 4 percentage points.
Awareness Of Technological Isomorphism: Integrating AI Into Elementary Mathematics Teaching On Data And Prediction, A Case Study Of The Compound Line Graph
Li Li, Yu Cao. (06/2026). arXiv.
Dual-track comparative analysis prompted students to spontaneously express that AI identified data trends using logic comparable to their own manual estimation.
Context-Aware Prediction Of Student Quiz Performance With Multimodal Textbook Features
Samin Khan. (05/2026). arXiv.
Combining textbook text and image features with prior exercise scores reduced student-grouped quiz prediction error by 9.1% relative to prior performance alone.
Facet: Multi-Agent AI Supporting Teachers In Scaling Differentiated Learning For Diverse Students
Jana Gonnermann-Muller, Jennifer Haase, Nicolas Leins, Moritz Igel, Konstantin Fackeldey, Sebastian Pokutta. (05/2026). arXiv.
Surveyed teachers gave favorable ratings to didactic structure and clarity across differentiated worksheets, scoring highest for profiles targeting low-motivation learners.
"Would You Want an AI Tutor?" Understanding Stakeholder Perceptions of LLM-based Systems in the Classroom
Caterina Fuligni, Daniel Dominguez Figaredo, Armanda Lewis, Julia Stoyanovich. (05/2026). arXiv.
A review of 38 studies revealed that existing research heavily focuses on university students (82%) and STEM fields (50%), entirely omitting parents and administrators.
A Scoping Review Of Large Language Model-Based Pedagogical Agents
Shan Li, Juan Zheng. (05/2026). arXiv.
Across 52 reviewed studies, large language model-based agents support student- and teacher-facing roles across K-12, higher education, and informal learning in subjects like science, programming, and language learning.
Addressing The Reality Gap: A Three-Tension Framework For Agentic AI Adoption
Jason Fournier, Kacper Lodzikowski. (05/2026). arXiv.
By the 2024–2025 school year, 60% of surveyed US K–12 teachers reported using AI, yet only 35% of district leaders reported providing student training.
