Peer coaching before writing may reduce AI shortcuts
A guest post describes using ungrading and peer coaching to deepen student thinking and make ChatGPT shortcuts less tempting.
TO
← Back to index
Tools
AI4Education Education topic
Public readings arranged by original publication date. People and institutions use explicit mappings; article sources remain in their details.
Latest updates
Discernment and Disclosure for AI Agents in Education
Marc Watkins argues faculty should question when to use AI agents and disclose their use to students.
Beyond chatbots: AI agents as software, not characters
Furze argues chatbot framing hides corporate responsibility and misdirects education's assessment debates.
Aila trial: less lesson-planning time, but mixed views and declining use
An English primary-school randomized trial found roughly 24% less planning time with Aila, about 49 minutes weekly, without inferior assessed resources. Use declined during the ten-week trial, and teacher views were mixed.
Six Important Findings from the AI in Action Learning Tour
Meyer highlights observations about teacher roles, curriculum fit, actionable suggestions and multilingual access, alongside disengaging uses. He asks how people actually use AI before treating capabilities as educational outcomes.
Notice, Judge, Redesign: Working Through the AI Productivity Dip
Jisc proposes noticing unproductive interactions, judging what AI can reliably handle, and redesigning workflows. Institutions should allocate learning time; a useful outcome may be recognizing when not to use AI.
What Should We Teach Children When AI Can Do So Much? (translated title)
Liu reports on a program at Kaohsiung Girls’ Senior High School in Taiwan emphasizing problem selection, collaboration, AI-supported making and explaining concepts to other learners.
The Overhang
Mollick argues that expertise, breadth, taste and agency help people use AI capabilities. His experiments invite a question: how do learners acquire the knowledge and judgment needed to direct and evaluate increasingly capable tools?
Reading Across Fields and Working with AI: A Course Retrospective — Shishufeng EP29
The second retrospective describes AI-built demonstrations and source navigation, while exposing the limits of keyword retrieval.
Don’t Throw Ed Tech Babies Out With the Bathwater
Teacher workload, student thinking, product evidence and family communication are treated separately.
Finding Sources in the AI Era — Shishufeng EP27
Hanyang and Liushou discuss moving from an unfamiliar topic to a syllabus, search terms and original sources.
Reading In An Age Of Thinking Machines
A documented design experiment with open uncertainties.
What does assurance of learning actually look like in practice?
UniSQ’s team describes a prototype connecting program outcomes, tasks and assessment security.
How An AI Detector Made Me Trust People Less
A personal trial raises questions about labels and trust.
The black box of AI assessment
A vendor’s validity review recognizes that stable scores can still reward the wrong learning.
China’s AI+Education Action Plan: curriculum, teacher development and governance
The five-department action plan connects AI literacy, university curricula, teacher development, applications and safeguards, with goals for 2030. These are policy directions, not completed implementation or demonstrated learning effects.
This is an Excellent Paper: The Effects of Prompt Injection on Grading
A Wharton team tests how embedded instructions affect AI grading, finding important differences across models and conditions.
Teaching Friends to Use AI — Shishufeng EP9
The hosts find that tacit knowledge about files, versions and structure often explains beginners’ difficulties.
Scalable and Personalized Oral Assessments Using Voice AI
Two cohorts show operational feasibility, while agreement among model graders does not establish validity.
Six ways to use artificial intelligence tools in alternative grading
Talbert separates AI assistance with design and practice from delegating grading itself.
Listening to My Students at Scale: Exit Tickets, NotebookLM, and the Tightest Feedback Loop I’ve Ever Built
Ipeirotis describes turning exit-ticket responses into targeted follow-up teaching.
A brief update: Why I’m still using AI detection after all, alongside many other strategies
A bounded pro-detection stance with acknowledged costs.
AI feedback customized for student writers: the updated PAIRR prompts
Prompt revision driven by feedback overload.
Playing Pretend: Expert Personas Don’t Improve Factual Accuracy
This report tests whether assigning expert identities reliably improves answers to difficult factual questions.
AI is uncannily good at judging writing
A large vendor report concerns judging agreement, not student learning.
Pitfalls in Evaluating Generative AI — Hung-yi Lee, Lecture 4
This public lesson extends AI literacy to evaluation: metrics, human preferences, model judges and benchmark contamination.
Getting the most out of AI feedback
Student-facing strategies, including a failure of flattery.
Bringing our feedback philosophy to life with AI
A design report explains why unreliable generated questions were replaced with human-written sets.
Prompt Engineering is Complicated and Contingent
Repeated tests illustrate why a successful answer does not establish a reliable prompting technique.
AI as Design Accelerator with Ryan Wetzel
Scaffolding collaboration through roles and playtesting.
ChatGPT in lesson preparation: the 2024 EEF/NFER trial
A 259-teacher trial supports planning-time savings without establishing higher student learning.
Trying out Grammarly Authorship Beta
An early account of process evidence and privacy tradeoffs.
Andy Matuschak — The reason most learning tools fail
Matuschak distinguishes forgetting from never understanding, and discusses questioning, memory and fading scaffolds. The interview offers a lens for AI tools: do they help learners notice confusion and act, or mainly ease consumption?