AI Tutor Paper Claims Large Effect Sizes at Dartmouth
*A preprint reports standardized effect sizes of 0.71 to 1.30 for an AI tutor used in a Dartmouth course, according to a Hacker News discussion.*
The claim
A paper posted to an academic workshop site states that an AI tutor produced effect sizes ranging from 0.71 to 1.30 standard deviations in a Dartmouth course. The finding appears in a PDF linked from the 2026 International Textbooks workshop site.
Limited public detail
The Hacker News front-page thread on the paper collected 112 points and 75 comments within hours. No abstract or additional results beyond the headline figure are reproduced in the discussion link. Readers are directed to the PDF itself for methods and data.
Reactions on Hacker News
Commenters noted the reported range sits well above typical education-intervention benchmarks but questioned sample size, controls, and whether the result holds outside one course. No on-the-record rebuttals or confirmations from Dartmouth appear in the thread.
Why it matters
Large claimed effect sizes in real university settings draw immediate attention from educators and tool builders. Until the full paper is examined and replicated, the numbers remain a single data point rather than settled practice.
---
Sources:
No comments yet