Educators Technology

Peer feedback is a powerful pedagogical strategy we can use to help students improve their writing.

But its value goes beyond improving a draft. It is also a practical way to help students develop judgment, metacognition, and the ability to evaluate suggestions critically.

These skills are especially important now that AI is becoming a persistent element in learning contexts.

In writing classes, students can already turn to AI for feedback, revision suggestions, sentence-level improvements, organization, tone, and even argument development.

In order to help students make the best of AI as feedback provider and revision helper we need to focus on how we can structure that use so it supports learning and scaffolds students thinking.

This is where the PAIRR framework becomes especially useful.

In their paper, Sperber et al. (2025) propose PAIRR, which stands for Peer and AI Review + Reflection. It is a human-centered formative assessment model that combines three important elements:

Peer feedback
AI feedback
Structured reflection

The sequence is not random though.

Students first write a full draft. Then they receive peer feedback. After that, they receive AI feedback. Finally, they reflect on both sources of feedback and use that reflection to guide revision.

I like this model because AI figures only after students have already produced a draft and after they have already engaged with human feedback. By the time students encounter AI feedback, they have something to compare it against. They are not simply accepting AI’s suggestions; they are evaluating them.
That reflection step is crucial.

Without reflection, AI feedback can easily become another form of passive consumption.

IF you are looking for a structured and research-backed way to bring AI into formative assessment, this framework is definitely worth exploring.

Reference:
Sperber, L., MacArthur, M., Minnillo, S., Stillman, N., & Whithaus, C. (2025). Peer and AI Review + Reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition, 76, 102921.

ChatGPT is neither a miracle solution nor a disaster for education.

This, for me, is the main takeaway from Mai, Da, and Hanh’s systematic review of 51 empirical studies on the use of ChatGPT in teaching and learning.

The authors used a SWOT analysis to examine ChatGPT’s strengths, weaknesses, opportunities, and threats across three stages of teaching and learning: before learning, during learning, and after learning.

The picture that emerges is balanced and practical.

ChatGPT can support teachers with course materials, lesson ideas, rubrics, explanations, feedback, and personalized learning support. It can help students brainstorm, revise, ask questions, and receive immediate guidance.

But the review also reminds us of the risks: inaccurate information, bias, fabricated citations, plagiarism, academic dishonesty, overreliance, and possible weakening of critical thinking.

Link in the first comment!

Reference:
Mai, D. T. T., Da, C. V., & Hanh, N. V. (2024). The use of ChatGPT in teaching and learning: A systematic review through SWOT analysis approach. Frontiers in Education, 9, 1328769.

One of my favourite papers to cite when I talk about the cognitive impact of mindless AI use is Shaw and Nave’s “Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender.”

The concept they outlined is known as cognitive surrender.

Shaw and Nave describe cognitive surrender as the tendency to adopt AI-generated answers with minimal scrutiny. In simple terms, it happens when we stop doing the mental work ourselves and allow AI to take over judgement, reasoning, and decision-making.

This is different from using AI as a thinking partner. This is whereAI can support reasoning, generate alternatives, challenge assumptions, and help us see new possibilities. But when used mindlessly, it can short-circuit our own thinking. We accept the output because it sounds fluent, confident, and plausible.

That is where the risk lies.

The problem is not AI use per se. It is surrendering our judgement to AI.

Bookmark the paper and read it when you have time.

Link in the first comment

Reference:
Shaw, S. D., & Nave, G. (2026). Thinking—fast, slow, and artificial: How AI is reshaping human reasoning and the rise of cognitive surrender.

Now that AI disrupts the traditional relationship between evidence and student performance, validity becomes one of the most important lenses through which we can approach AI-mediated assessments.

For a long time, assessment evidence was easier to locate. A student submitted an essay, completed a quiz, wrote a reflection, solved a problem, or delivered a presentation.

We could still question the quality of the assessment, of course, but the relationship between the student, the task, and the evidence was relatively stable.

AI complicates this relationship.

When students use generative AI to brainstorm, draft ...etc., the evidence we receive no longer speaks in a simple way. A submitted product may still tell us something, but what exactly does it tell us?

Does it show the student’s understanding? Their ability to prompt? Their judgment in revising AI output? Their capacity to evaluate evidence? Their dependence on the tool? Their strategic use of support?

These are validity questions.

Over the last few weeks, I have shared several papers by scholars working at the intersection of AI, assessment, and higher education, including work by Dawson, Bearman, Crobin, and others.

These scholars help us think about assessment in a time when authorship, evidence, feedback, and academic integrity are being reshaped by AI.

Today, I wanted to go back to one of the seminal works in assessment theory: Samuel Messick’s 1995 paper on validity.

Messick’s central point is powerful and still highly relevant today: validity is not a property of the test itself. Validity concerns the meaning of scores and the decisions we make based on them.

In simple terms, an assessment is not valid simply because it looks rigorous, covers the right topic, or produces a score. It becomes valid only when we have strong evidence and sound reasoning to support the interpretations and uses attached to that score.

So, if a student submits an AI-supported essay, we cannot assume the essay alone provides clear evidence of learning. We need to ask better questions:

What construct are we trying to assess?

What kind of evidence does this task actually produce?

What role did AI play in producing that evidence?

Does the task capture the student’s thinking process, or only the final polished product?

Validity is ongoing inquiry into meaning, evidence, use, and consequences.

This is why validity is becoming even more important in AI-mediated assessment. AI does not simply create a cheating problem. It creates a measurement problem, a design problem, and an interpretation problem.

That is where Messick’s work remains so valuable.

Reference:

Messick, S. (1995). Validity of psychological assessment: Validation of inferences from persons’ responses and performances as scientific inquiry into score meaning. American Psychologist, 50(9), 741–749.

New AI Literacy Framework for Primary and Secondary Education!

This new resource comes right in time for me as I prepare to teach a course at MSVU on Curriculum and Technology Integration for preservice teachers.

Shout out to Pat Yongpradit for the share!

One point I keep coming back to is this: using AI with younger learners, especially students in elementary schools, is not the same as using it with older students.

I would even say it is a totally different story.

Children are still developing cognitively, socially, and emotionally. Their reasoning, self-regulation, sense of agency, and ability to judge information are still in formation. Our AI integration has to reflect that reality.

For elementary students, AI use should be highly guided, carefully monitored, and grounded in clear pedagogical purpose. It should definitely not be about giving children open access to tools and hoping for the best.

The limited capacity I would say for using AI with kids is to help them understand what AI is, what it can and cannot do, how to question its outputs, and how to use it in ways that protect their thinking, creativity, privacy, and well-being (Common Sense Media has some wonderful resources and activities for this).

The current framework (which resembles previous frameworks they shared namely AI Competency for Teachers/Students) organizes AI literacy around four important domains:

Engage with AI

Create with AI

Manage AI

Shape AI

The framework also foregrounds human agency, critical thinking, ethical judgement, creativity, responsibility, and the need to help students understand that AI is not human, even when it sounds human.

Link in the first comment!

Reference:

OECD / European Union. (2026). Empowering learners for the age of AI: An AI literacy framework for primary and secondary education. OECD Publishing.

Assessment in a digital world cannot be reduced to putting old tasks on new platforms.

That is one of the strongest ideas I took from Bearman, Nieminen, and Ajjawi’s framework on designing assessment in a digital world.

Too often, digital assessment means convenience: LMS submissions, online quizzes, plagiarism detection, automated grading, or online proctoring. These tools may improve efficiency, but they do not automatically make assessment more meaningful.

The deeper question is this:

What kind of learning, judgement, identity, and capability should assessment develop in a digitally mediated society?

The authors offer a helpful way to think about this through three purposes:

1. Digital tools for better assessment
Here, technology supports assessment design, feedback, efficiency, or authenticity. But the key is intentionality. Why are we using the tool? What does it improve? What possible harms might it introduce?

2. Digital literacies
Students need opportunities to develop and demonstrate digital skills. But this should not be assumed. Asking students to create a video, submit an e-portfolio, or use a digital platform does not mean they have been taught the digital literacies needed to do the task well.

3. Human capabilities for a digital world
This is perhaps the most important part of the framework. As digital technologies and AI take over more routine tasks, assessment should help students develop what remains deeply human: judgement, creativity, ethical reasoning, collaboration, identity formation, and the ability to define quality.

This resonates strongly with our current AI moment.

If AI can generate products that look polished, then assessment has to move closer to the processes of thinking: how students judge, revise, question, critique, justify, and make meaning.

The future of assessment is not simply more digital.

It is more intentional, more human, and more connected to the kinds of capabilities students need beyond the classroom.

Here is a very good framework to help integrate AI in your instruction using a combination of SAMR and Bloom’s Taxonomy.

Both SAMR and Bloom's Taxonomy are teaching/learning frameworks that have been with us for ages.

They are still very relevant and can actually provide pedagogical scaffolding to teachers AI integration.

SAMR helps us ask: Is AI simply replacing an existing tool, improving the task, redesigning the task, or making something entirely new possible?

Bloom’s Taxonomy helps us ask: What level of thinking are students actually doing: remembering, understanding, applying, analyzing, evaluating, or creating?

Together, they offer a practical way to design AI-supported learning activities that preserve student thinking.

This is especially important now because AI has changed the relationship between task completion and evidence of learning. A polished final product is no longer enough. We need learning activities that make thinking visible and these frameworks can help us design these learning activities.

Reference:
Keys-Harris, A. (2026). A practical SAMR + AI framework for instructional design. EDUCAUSE Review.

This is an old paper, but its insights are deeply relevant to us in the age of AI.

In “Teaching and Testing to Develop Fluid Abilities,” David Lohman argues that schools should not only help students acquire knowledge, but also help them transfer, adapt, and use that knowledge in unfamiliar situations.

I think this message is just as important and relevant today.

AI can retrieve information, summarize content, generate examples, and produce polished answers in seconds. But what students still need to develop is the ability to think flexibly: to solve unfamiliar problems, connect ideas, reorganize knowledge, evaluate responses, and apply learning in new contexts.

Lohman’s point is powerful: if we only test students on disconnected facts, we teach them to value disconnected facts. But if we ask them to organize ideas, explain relationships, and apply knowledge in novel situations, we help them develop the kind of thinking that remains essential in an AI-rich world.

Link in the first comment!

Reference:
Lohman, D. F. (1993). Teaching and testing to develop fluid abilities. Educational Researcher, 22(7), 12–23.

One of my favourite frameworks for integrating AI in teaching is Understanding by Design.

This seminal work by Grant Wiggins and Jay McTighe reminds us of something crucial: good teaching starts with the end in mind. Before choosing tools, activities, or platforms, we need to ask: What do we want students to understand? What evidence will show that they understand it? What learning experiences will help them get there?

This is exactly the kind of thinking we need with AI.

AI should not be added to lessons simply because it is new or exciting. It should serve clear learning goals. It should help students think deeper, apply ideas, create, question, revise, and transfer their learning to new situations.

UbD gives us a powerful reminder: technology comes after purpose. AI integration works best when it is aligned with outcomes, assessment, and meaningful learning experiences.

Link in the first comment!

Reference:
Wiggins, G., & McTighe, J. (2005). Understanding by design (Expanded 2nd ed.). ASCD.

AI detection tools are not the ideal response to student use of generative AI.

A recent study by Sun, Liao, and Ma tested 13 AI detection tools on more than 280,000 samples of student academic work, including coursework, theses, and engineering code.

The findings are sobering.

Detection tools worked better on long academic writing, but their performance dropped sharply on short assignments and code. They also struggled with STEM writing because technical writing is often structured, precise, and formulaic, which can make strong human work look “AI-like.”

Even more concerning, lightly edited AI-generated text often escaped detection. In some cases, a hybrid editing strategy allowed up to 88% of AI-generated content to bypass detection.

For teachers, the takeaway is clear: AI detectors may offer a signal, but they should never be used alone to accuse students of misconduct.

The better response is assessment redesign.

Ask students to show their process. Include drafts, reflections, oral checks, version histories, rationale statements, and conversations about how ideas developed. Teach students how to use AI ethically, critically, and transparently.

Academic integrity in the age of AI will not come from catching students after the fact. It will come from designing learning in ways that make thinking visible.

link in the first comment!

Reference:

Sun, Y., Liao, Y., & Ma, X. (2026). Trusting AI to detect AI? A systematic evaluation of the reliability and robustness of current AIGC detection tools for student academic work. Computers & Education, 249, 105616. https://doi.org/10.1016/j.compedu.2026.105616