Towards an AI-Augmented Textbook  ## Metadata - Author: [[LearnLM Team, Google, Amy Wang, Anna Iurichenko, Anisha Choudhury, Alicia Martín, Amir Globerson, Av...]] - Full Title: Towards an AI-Augmented Textbook - Category: #articles - Document Note: --- tags: - #ai - #education - #personalizedlearning - #edtech - #assessment - #learningdesign - #openstax - #research Author: LearnLM Team (Google) Source: Whitepaper/PDF Reference: https://services.google.com/fh/files/misc/ai_augmented_textbook.pdf date created: 2025-09-16 00:00 --- # Title: Towards An AI‑Augmented Textbook ## Table of Contents: - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#-notes) - [📖 References](#-references) - [🗣️ Paraphrase](#-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-dei-inclusion-belonging-assessment) - [✅ Logical validity](#-logical-validity) - [🔍 DARVO Analysis](#-darvo-analysis) - [🔍 SEO Evaluation](#-seo-evaluation) - [🎯 Conclusion](#-conclusion) ### References: - OpenStax/LibreTexts chapters cited in figures and study - Core learning science literature on personalization, dual coding, SRL, formative assessment ## 📝 Summary: The paper proposes “Learn Your Way,” a two‑stage generative AI system that augments textbooks with (1) text personalization (grade level, interests) and (2) multiple content transformations (narrated slides, audio‑graphic lessons, mind maps, “Immersive Text” with visuals, timelines, memory aids, embedded questions, and quizzes). Expert rubric evaluations rate components highly overall; a randomized controlled study with high‑school students shows statistically significant gains in immediate and delayed assessments versus a standard digital reader, suggesting AI‑augmented, multi‑view, and personalized materials can improve learning efficacy while maintaining content integrity. ## 📖 Main Content: - Problem: Textbooks are one‑size‑fits‑all and hard to adapt at scale. - Approach: Use GenAI (Gemini 2.5 Pro) to re‑level and personalize content, then render multiple synchronized representations, plus formative assessment. - Personalization: Rewrites for targeted grade level (FKG) and interest domains (e.g., sports, music), with selective substitution and highlighting to preserve source fidelity. - Transformations: - Narrated slides: concise, engaging, class‑like deck with optional narration. - Audio‑graphic lesson: dialogue between teacher/student personas plus evolving concept graph. - Mind maps: hierarchical, expandable with illustrative text and images. - Immersive Text: the full, coherent text interleaved with visuals, timelines, memory aids, embedded questions, and section quizzes. - Visuals: Fine‑tuned image model used because general image generators struggle with simple, pedagogical diagrams. - Assessment: Embedded MCQs per segment and section‑level quizzes with “Glows/Grows” feedback. - Expert evaluation: High scores across coverage, accuracy, cognitive load, active learning, SRL, motivation, adaptability, clarity. Visual illustrations lowest among components but still positive; narration meaningfully boosts slide engagement. - Efficacy study: 60 teens randomized to Learn Your Way vs. Adobe Reader; same chapter (adolescent brain development), same expert‑written assessments. Learn Your Way yields higher immediate and 3‑day retention scores (Mann‑Whitney U, p=0.03 both). - Limitations/future: Need ablation to isolate contributions, more topics/chapters, richer adaptivity to performance/signals, stronger teacher controls, more interactivity. ### 🗒️ Notes - Motivation - Textbooks lack scalability for personalization; GenAI can bridge the gap. - Two‑stage pipeline - Stage 1: Personalize to grade level (FKG) and a declared interest; selective rewriting to preserve core facts; highlight changes. - Stage 2: Transform into multiple synchronized modalities (slides+narration, audio‑graphic, mind maps, immersive text). - Immersive Text add‑ons - Visuals (via fine‑tuned model), timelines, memory aids, embedded questions, quizzes with targeted feedback. - Learning science grounding - Multiple representations (dual coding). - Learner agency and adaptability. - SRL support via formative assessment and visible learning cues. - Expert rubric outcomes - Overall >0.90; narration improves engagement vs. slides alone. - Visual illustration is hardest to get right; fine‑tune helps. - RCT design highlights - Same content for all; time‑bounded study; immediate and delayed assessments; nonparametric stats due to non‑normality. - Positive user experience ratings across measures. - Practical implications - Personalization can be done at scale without sacrificing fidelity. - Multi‑modal pathways plus embedded assessment improves outcomes. ## 📖 References: - Ainsworth (1999) on multiple representations - Clark & Paivio (1991) on dual coding - Zimmerman & Schunk (2011) on SRL - Hattie (2008) on visible learning - Black & Wiliam (2009); Earl (2013) on formative assessment - Hattan et al. (2024) on prior knowledge - LearnLM “Evaluating Gemini in an Arena for Learning” (2025) - Kincaid et al. (1975) on readability/FGK ## 🗣️ Paraphrase: - Original idea: “We present an approach for transforming and augmenting textbooks using generative AI, adding layers of multiple representations and personalization while maintaining content integrity and quality.” - Paraphrase: We describe a GenAI‑driven method to refactor textbooks into personalized, multi‑modal experiences that preserve the original material’s accuracy and completeness. - Original idea: “The students who used Learn Your Way received higher scores than those who used the Digital Reader, in both the immediate (p=0.03) and retention (p=0.03) assessments.” - Paraphrase: Participants studying with Learn Your Way outperformed the digital reader cohort on tests taken right after learning and three days later, with both differences statistically significant (p=0.03). ## 📰 Newsletter: This week in WordPress learning: Want your docs to teach themselves? Google’s LearnLM team unveiled “Learn Your Way,” an AI‑powered layer that turns static textbooks into personalized, multi‑modal lessons—think narrated slides, audio‑graphic dialogues, mind maps, and an “Immersive Text” that peppers in visuals, timelines, and just‑in‑time quizzes. In a randomized study, students scored higher immediately and three days later compared to a standard PDF reader. Food for thought as we craft contributor guides and developer handbooks. Plugin corner: keep an eye on TTS/voiceover tools for docs, mind‑map generators for curricula, and quiz blocks that deliver instant “Glows & Grows” feedback. Community highlight: educators experimenting with grade‑level re‑leveling and interest‑based examples are seeing better engagement in onboarding materials. Links: - Learn Your Way whitepaper from Google’s LearnLM team: https://services.google.com/fh/files/misc/ai_augmented_textbook.pdf - OpenStax (source chapters used in examples): https://openstax.org/ - LibreTexts (study chapter source): https://libretexts.org/ ## ✅ Action Items: Prompt for aligning any article with your Developer Advocate/OSPO trajectory: - Objective: Explain how this article informs open‑source advocacy, contributor engagement, and education within WordPress and an OSPO context. - Instructions: - Summarize core claims and evidence. - Map to open‑source education: How can multi‑modal, personalized docs improve new‑contributor onboarding (PHP/JS/HTML/CSS/Git/GitHub)? - Contributor engagement: Propose interest‑based examples (e.g., WooCommerce, block themes, REST API) and grade‑level re‑leveling for docs. - Sustainability: Describe how formative assessments in docs (quizzes, check‑your‑understanding) can shorten time‑to‑first‑PR and reduce maintainer load. - Transparency/inclusivity: Outline DEIB, accessibility, and multilingual pathways; define measurable rubric‑based evaluations for docs quality. - OSPO readiness: Identify governance, content integrity, and curriculum alignment practices; recommend telem - URL: https://services.google.com/fh/files/misc/ai_augmented_textbook.pdf Full text of this document omitted because this document is a PDF [Download original file →](https://readwise.io/reader/document_raw_content/364018299), Towards an AI-Augmented Textbook, LearnLM Team, Google, Amy Wang, Anna Iurichenko, Anisha Choudhury, Alicia Martín, Amir Globerson, Av..., 01k5abf4t9hrxvbdqa9876rqy4.