AI tutoring can safely and effectively support students: An exploratory RCT in UK classrooms

LearnLM Team, Google & Eedi · 2025

Read AI tutoring can safely and effectively support students: An exploratory RCT in UK classrooms by LearnLM Team, Google & Eedi (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-2512-23633.txt.

Description

A large team from the LearnLM project and the tutoring platform Eedi report a randomized trial of AI tutoring in British classrooms. One-to-one tutoring works well but is too expensive to offer every student. In five UK secondary schools, one hundred sixty-five students received chat-based mathematics tutoring in which a pedagogy-tuned model, LearnLM, drafted replies that expert human tutors supervised and could edit before sending.

The supervision was light: tutors approved about seventy-six percent of the draft messages with zero or minimal edits. Students supported this way learned at least as well as peers chatting with human tutors alone, and they were somewhat more likely to solve novel problems on later topics. Interviewed tutors praised the model's Socratic questioning and some reported learning new teaching moves from it. The authors present the study as exploratory evidence that supervised AI tutoring can scale safely.

Catalog record

CREATORLearnLM Team, Google & Eedi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE93 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-23633

Browse the full library · About the Tower of Babel