Beyond chatbots: AI agents as software, not characters
Furze argues chatbot framing hides corporate responsibility and misdirects education's assessment debates.
TH
← Back to index
Theory
AI4Education Education topic
Public readings arranged by original publication date. People and institutions use explicit mappings; article sources remain in their details.
Latest updates
Things to bear in mind when talking about AI
Selwyn calls for precise, contextual discussion of AI and warns against treating a power user’s experience as universal expertise. Constructive critique should ask what alternative forms of AI and educational practice people want.
Must Children Keep Acquiring More Abilities in the AI Era? (translated title)
Educator Zhang Shuling questions defining children’s worth through an ever-expanding list of abilities AI cannot yet perform. Useful capacities matter, but do not exhaust the purpose of education.
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?
Varying Retrieval Practice to Promote Transfer of Learning
Sumeracki explains Butler et al.’s (2017) experiments contrasting repeated and varied application during retrieval practice. Her account invites asking whether an AI activity makes students retrieve, explain and apply concepts across contexts.
Two validity conversations in a time of AI
Dawson separates validating AI assessors from validating assessment when students have AI access.
The algorithm in the room: AI and the future of education
This UNESCO collection brings together 26 think pieces on agency, young people’s development, pedagogy, expression, public governance and educational futures. It broadens discussion beyond tool capabilities.
What if transferable skills don’t exist?
A deliberately strong argument locates reasoning and problem-solving in domain knowledge.
How Should AI Be Used in Schools?
Examples of planning and writing ask which cognitive tasks can appropriately be delegated.
Moral Instruction Will Not Stop Cheating
Reframing misconduct around incentives and learning.
Is This a Good Use of the Technology? Is It a Good Use of Your Brain?
Furze proposes two contextual questions: is this a good use of the technology, and of the learner’s own thinking? He examines accessibility, outsourcing and assessment, questioning whether fixed categories capture individual needs.
Our Duty of Care in the Age of AI
Course purposes and faculty working conditions belong together.
Students should only use AI for things they already know how to do well.
The proposed restriction concerns independent general-purpose AI use, not all instructional software.
How Alpha School Teaches Writing
The critique asks whether interest-driven writing loses opportunities to consolidate curriculum knowledge.
Three ways alternative grading addresses the EDUCAUSE Horizon report
Talbert argues that changing tasks without changing grading leaves a core incentive problem untouched.
Let’s emphasize human effort and intent in educational materials
Disclosure as a question of educational relationships.
What if We Accidentally Outsource the Real Work to AI?
Automation can alter the practice behind an artifact.
What Is Understanding? — Shishufeng EP23
Chongqing uses Hinton’s account of understanding to question what learners lose when AI explanations become too smooth.
Just an Exam — Shishufeng EP22
Hanyang and Chongqing use university choices to question the idea that one decision fixes an entire life path.
Why personalising education based on student interests doesn’t work
Performance-based adaptation is distinguished from rewriting every topic around a learner’s interests.
Choosing to Stay Human
Mollick considers how convenient AI defaults can bypass the effort through which people develop valued abilities.
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.
On AI glasses and wearable AI in assessment
The authors question assessment designs that depend on detecting and excluding wearable assistance.
What are we going to do about marking now?
Tifflin asks markers to make conceptual grasp, evaluative judgment and transfer visible.
Understanding the Evidence Base on AI in K-12 Education
A Stanford review identified 20 higher-quality causal studies within over 800 K12 AI papers available by October 2025. Assisted performance and unaided learning diverge; long-term, equity and wellbeing evidence remains limited.
The University Within the Limits of Automated Knowledge
Yuk Hui asks what universities lose when knowledge production and economic utility become their defining purposes.
Education in the AI Era Should Foster Independent Thought and Lived Growth
Li Zhengtao argues that educational purposes should shape technology, protecting students’ own thinking and lived learning.
Is Relying on AI Cognitive “Offloading” or “Outsourcing”?
Outlines illustrate a distinction between external memory support and delegating the thinking itself.
Alison Gopnik on Childhood Learning, AI as a Cultural Technology, and Rethinking Nature vs. Nurture
Gopnik connects exploration, apprenticeship and AI literacy in an interview that preserves disagreement about machine reasoning.
Assessment after Artificial Intelligence: The Research We Should Be Doing
An assessment research agenda asks for precise practices, historical grounding and connected inquiry.
Can LLMs be personal tutors?
Question quality, accuracy, alternatives and persistence matter beyond fluent explanations.
Valid assessment is a plausible argument not an absolute
Steel explains why assessment security alone does not establish a valid learning inference.
The wicked problem of AI and assessment
Interviews with 20 educators reveal competing assessment goals and unavoidable trade-offs.
Education isn’t working
The essay distinguishes productive work from the practice needed to become an expert.
Against “Brain Damage”
Mollick challenges sensational interpretations of AI studies and examines learning, idea generation and writing practices.
Generative AI without guardrails can harm learning: Evidence from high school mathematics
A mathematics RCT directly separates assisted performance from subsequent unaided assessment.
Talk is cheap: why structural assessment changes are needed for a time of GenAI
The authors distinguish compliance-dependent instructions from changes to assessment structure.
Do Students Who Rely on ChatGPT “Benefit”?
Wexler challenges interpreting better submitted prose as improved writing ability.
Exploring two frameworks from The Open University: Critical AI Literacy Skills and Responsible by Design
OU’s designers connect AI literacy with access, bias, environmental and labour questions.
What to do about artificial intelligence now: Examining the real impact on learning
Lodge calls for evidence about learning mechanisms beyond uptake and perceived usefulness.
Reading in the Age of Social Media (and AI)
Reading habits are also an educator’s problem.
Academic Integrity in an LLM World
What should a course credential actually certify?
The Writing Effect
Writing’s retrieval and elaboration benefits must be balanced against novice cognitive load.
AI competency framework for teachers
UNESCO outlines 15 teacher competencies across five dimensions and three progression levels, supporting teacher development and assessment design while foregrounding ethics, human agency and pedagogy.
Historical foundation: Validity matters more than cheating
The authors place cheating within the broader question of defensible judgments about capability.
Historical reference: AI-Generated Summaries Are a Problem—for Authors and Society as a Whole
An author’s inspection of generated book companions reveals losses beyond factual accuracy.
No One is Talking About AI’s Impact on Reading
Earlier classic: reading assistance versus reading replacement.
On AI and ‘Meaningful’ Feedback
Earlier classic: ask whether feedback is actually used.
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?