AI + Education Weekly Update
Week of December 8th, 2025
To catch up on another busy week in AI + Education news, check out our weekly updates for our review and analysis of some of the key highlights worth staying informed about.
Week of December 8th, 2025
Traditional Note-taking Outperforms LLM Use
Cambridge research found that secondary school students learned significantly less using LLMs alone versus traditional note-taking, despite preferring AI—revealing a gap between perceived helpfulness and actual learning outcomes.
405 students, ages 14-15, were divided into 3 groups and tested: LLM only, Notes only, and LLM + Notes
Note-taking alone scored the highest across comprehension, retention, and recall; the LLM + note-taking group outperformed the LLM-only group.
Students preferred LLMs and rated them as more helpful, despite lower performance and approximately 25% copied AI responses instead of creating original notes
OECD: AI Capabilities and the School Curriculum
OECD's 2025 report examines how AI mastery of human tasks may require fundamental curriculum redesign—questioning whether traditional educational assumptions remain valid.
The research suggests that as AI handles composition, curricula should pivot to higher-order processes—planning, critical evaluation, and judging idea quality.
Depending on how rapidly AI advances, curriculum implications range from modest adjustments to fundamental educational redesign.
Education should shift from asking 'what should students know?' to 'what uniquely human capabilities remain essential?', while systematically monitoring AI to reconsider essential knowledge.
Mckinsey: Skill Partnerships in AI
McKinsey's new report shows that AI is changing how we work—shifting the focus from automation risk to human-AI partnership.
AI is now the fastest growing skill requirement—demand is up 7X in 2 years
70% of human skills remain relevant but applied differently: Less routine work, more high-level cognitive functions (framing problems, interpreting AI outputs, orchestrating hybrid teams)
Digital skills face the highest transformation, while human-centric skills (assisting, caring) remain stable
Unlocking $2.9T in annual productivity value by 2030 requires redesigning entire workflows around human-AI collaboration, not just automating tasks
Google Launches Workspace AI Agents
Google launched Workspace Studio, enabling any employee to build AI agents that automate work without coding—democratizing automation across organizations.
Build agents in minutes using natural language prompts; no coding or specialized syntax required
Agents work across Gmail, Drive, Chat, and connect to enterprise platforms like Salesforce, Asana, and Jira; shareable across teams like Drive files
Agents work within company policies, processes, and generate content in your tone; Gemini 3 powers reasoning to handle simple tasks to complex multi-step workflows
Chatbots Sway Political Opinions
A UK study found AI chatbots can significantly change political opinions through conversation—but making AI more persuasive systematically made it less accurate.
Flooding conversations with factual claims increased persuasion 27% over storytelling or moral appeals
Specialized persuasion training boosted effectiveness 51%—making smaller models as persuasive as frontier AI
More persuasive models made more inaccurate claims; GPT-4.5 was less accurate than earlier versions despite being more convincing
