The Illusion of the Perfect Page
Long before artificial intelligence entered our classrooms, we had a weakness for pretty work. Young learners with beautiful handwriting could look reassuringly successful before we had properly considered what they had written. Meanwhile, the child whose page looked as though a spider had dipped its legs in ink and walked across it had some convincing to do. We knew, of course, that tidy handwriting and good thinking were different things. Knowing that did not always stop one from influencing our judgement of the other (Malpique, Pino-Pasternak and Valcan, 2017). Research on handwriting and assessment has examined precisely this problem: presentation can influence how we judge the quality of an answer.
AI gives us another reason to examine that habit. The sentences flow, the slides look professional and the argument arrives neatly organised. Sometimes the learner has made considerable progress. Sometimes the product conceals how little has changed. We need to be curious enough to find out, particularly when the work is pleasing to look at.
The Value of the Teacher-Student Relationship
I found myself thinking about the relationship between assessment and learning when we first introduced AI into our assessment and feedback practices. The students did not like it. I spent time with them asking why, explaining that AI could help provide more feedback than I could manage alone, especially as a school leader who also taught a partial teaching load. I explained, too, that I often dictated my comments and used AI to turn my spoken words into written feedback. My thinking could still be there, even when I had not typed every sentence.
What emerged from those conversations was how much the students valued their relationship with their teachers. They had spent time and effort on their work, and they wanted to know that I had actually read it, as well as reviewed and assessed it. They wanted to please me. They needed to know that I had noticed their effort and understood what they were trying to do. The comments mattered because of the relationship behind them.
That gave me pause. I had been explaining the volume of feedback available while they were asking about my attention. Their response reminded me that assessment sits within a relationship. We should want students to develop judgement beyond pleasing us, but the wish to be seen by a teacher is a meaningful starting point. It deserves care.
The Temptation of Shortcuts
Students can want our attention and still want their homework finished as quickly as possible. We need to be honest about that. Children are using AI. Some watch their parents use it, then wonder why they should not do the same. Some want to explore an idea; others want to get the assignments done and go and play on their game console. The urge to finish quickly is familiar to anyone who remembers homework completed on the bus to school or copied from a friend. This time, though, it looks pretty. With examples of their writing and enough direction, AI can even produce something that sounds like them.
We cannot keep putting our heads in the sand, or expect punishment to resolve the understandable attraction of an easier route to completion. Clear expectations about honesty and authorship still matter, and so does asking what our tasks invite students to do. If we set work, send it home and assess only the finished product, we leave much of the learning out of sight. Bolting AI onto that arrangement, whether to generate answers or evaluate them, does little to help us understand what the student has gained. We need to reimagine the workflow.
Reimagining Assessment Policies
That is a leadership decision and it needs to appear in our policies. An AI policy can tell students which tools they may use, what they may share and when they should acknowledge assistance. Those are necessary questions, but they leave something unresolved if the assessment policy still assumes that a finished piece of homework gives us a clear account of what a student can do. Read together, the two policies should help teachers decide which thinking students need to do independently, where AI might help them practise or improve, and what evidence of learning the task should produce. The answer will differ between an early draft, a formative task and an assessment intended to show independent achievement.
We also need policies that describe what happens when the work raises a question. A student should understand the expectations before starting, including how to explain the help they used. A teacher who is unsure about authorship needs a fair way to follow it up through conversation and evidence of learning, rather than treating a suspicion or an automated score as a verdict. Equally, if a school chooses to use AI in feedback, students should know what role it plays and when their teacher will read and respond to their work. These choices shape trust. They cannot be settled by adding a paragraph about AI to a policy written for a different classroom in a different time.
Integrating AI Thoughtfully into the Learning Process
My own response has been to build opportunities to notice and discuss thinking into the task itself. In AI for Thinking, I bring together metacognitive strategies for before students begin using AI, during the task and when they review what they have learned. These moments help me see where understanding has deepened and where a student remains concerned primarily with getting the product finished.
Before using AI, students need an opportunity to establish their own starting point. What do they already understand? What do they think the task requires? Where are they uncertain? A brief explanation, a rough plan or a sketch can make that thinking available for discussion. This is the purpose behind “think first, then AI”. It gives students something of their own to bring into the exchange and something against which to consider what AI offers.
During the task, a brief review session creates space to notice what is happening. A student might identify a suggestion they accepted and explain why, or return to an earlier idea that now seems less convincing. The useful question is what prompted that change. Did they encounter better evidence, recognise a gap in their reasoning, or simply prefer how the new version sounded? Talking about these decisions helps students recognise the difference.
The Danger of Fluency Without Understanding
When students look back over their work, I have found metacognitive AI tutors useful in helping them make sense of the learning behind it. The tutors help students find words to explain how their understanding has developed. Prompts can ask them to revisit an initial assumption, explain a choice or identify something they still cannot do confidently. This has helped reveal learning that the finished product alone would not show, as well as places where an attractive outcome has run ahead of understanding.
There is a danger here too. A beautifully worded reflection can become another pretty product. Saying that your understanding has deepened does not, by itself, demonstrate that it has. I would want to follow the explanation back into the work, ask a further question or see the student use the idea in a different context. A sketch or a spoken explanation may reveal more than another written paragraph. Fluency should not become our only evidence of thought.
This is where assessment as learning becomes useful. When students compare their first attempt with a later decision, consider the evidence and work out what to do next, assessment can help develop their judgement. The Education Endowment Foundation’s guidance similarly emphasises explicitly teaching students to plan, monitor and evaluate their learning (Education Endowment Foundation, 2021). An AI tutor can support that practice; teachers remain responsible for interpreting what it reveals.
What Leaders Can Do
To successfully reimagine workflows and integrate AI meaningfully, school leaders can take the following practical steps:
Evaluate Existing Tasks: Begin with one existing task at a departmental meeting, placing both the AI and assessment policies on the table.
Map the Learning Journey: Identify exactly where the student establishes an initial idea, where AI could usefully support or challenge it, and when a teacher hears the student’s reasoning.
Assess Independent Application: Determine how educators will see whether students can use their newfound understanding in a new situation.
Align Policies: Ask whether the current policies give teachers and students a clear, unified answer to these workflow questions.
Protect Human Connection: Agree which thinking needs to happen independently and where assistance serves the learning. Build those decisions into the sequence of the task, ensuring that any time saved through AI feedback is used to protect conversation and revision time with the teacher.
Final Takeaway
We should give as much attention to how work comes into being as to its finished appearance. My students wanted to know that I had read what they had made; I wanted to understand what making it had helped them learn. Reimagining the workflow means making room for both, from the first uncertain idea to the conversation after the work is handed in. Our policies should make that possible in classrooms across the school. The prettier page can still be something to celebrate. We should be able to explain what there is to celebrate in the learning behind it.
References
Education Endowment Foundation (2021) *Metacognition and Self-Regulated Learning Guidance Report*. London: Education Endowment Foundation.
Malpique, A., Pino-Pasternak, D. and Valcan, D. (2017) ‘Handwriting automaticity and writing instruction in early years education: An exploratory study’, *Reading and Writing*, 30(8), pp.1655-1678.
Stephanie Holt, Teaching and Learning at Riverside International School, Prague
LYIS is proud to partner with Beyond Classrooms
Leave a Reply