Really enjoyed this, thank you. It makes me think about some advice I was given about formative feedback: focus on the HOW to do better, not the WHAT to do. As the excellent 'systematic' anecdote said, 'if I knew how I'd have done it the first time'. My worry with automated feedback is that the practitioner may not have enough of their own cognitive skin in the game.
Reading automated feedback is not the same as unpicking the text for why something didn't work. If the practitioner hasn't had to think about it, they are unlikely to fully understand it. If memory, after all, is the residue of thought, then that applies to staff as well as children.
My ongoing worry with AI in schools is that we are pursuing efficiency at the expense of efficacy.
Very off the cuff and I may well have missed the point entirely but very much enjoyed it so thank you for sharing.
Yes, we think about this a lot. The challenge is to make sure that the time and effort teachers spend on assessment is manageable and as high-value as possible. We don't want to eliminate it - we want to reduce it and focus it on the most valuable aspects.
My first question, having tried this out this year, is about the "control": while re-drafting with the AI feedback has produced an improvement, what would be the equivalent outcome for re-drafting without specific feedback - either more generic 'whole class' pointers, or even no feedback at all. I have a suspicion that a clear focus on fewer points of generic feedback might be more effective for those who most need it.
Yes, definitely. Fair point re the control. Also the experiment I really want to do is to re-assess the students on a cold task and see if improvement is sustained. But it's hard to do with this cohort as they go off to secondary. We'd need to run a Y5 redraft and then measure improvement from there to Y6. And we have approx 10 years of non-AI 5 to 6 average improvement, so could use that as control. But we are super busy at the moment and I don't know how much appetite there is for a Y5 redraft.
Thanks for the post . Loved the Dylan William anecdote, especially yhe student replying, “I don’t know. If I knew how to be more systematic, I would have been so the first time.”
This is such a powerful argument against so much marking that still happens in schools particularly in primary. If I'd known how to spell it correctly I'd have spelled it correctly. Most kids don't make errors on purpose, they make errors because they haven't learned the correct versions. As to reading feedback I really don't think most primary learners do that. They'll look for a mark, a grade or a smiley face but in my experience they won't often read, understand or process written feedback.
I agree. I think that in the current system, most pupils (and parents) do see a written comment as a sign that their teacher cares about them and has given them time and attention. So the written comment may well have motivational value, which is important. But that is totally different from understanding and acting on what has been written!
That is exactly the problem with the system we have created. It is less to do with does it help a student improve and more to do with perfunctory obligation to placate the vested ignorant (or willfully ignorant where Ofsted and the higher escelons of a school are concerned).
I wonder if today that all staff in every school just said no to written feedback would outcomes change at all?
My gut instinct is we would muddle along just fine (even senior leaders might be okay with the time saved by eliminating the obsessive need to do a work scrutiny). It would especially be so if we could now focus our attentions on improving the efficacy of verbal feedback in the immediacy of a learning episode.
I fear that many SLTs, now ,would consider your proposal way too radical especially in the current climate with the irresistible rise of multi academy trusts where policy is often centralised and many primary schools no longer have a headteacher but rather a head of school which is absolutely not the same role.
No truer words spoken. Unfortunately where bravery is needed to eliminate the countless hours of wasted bureaucracy we are led by lambs, rather than lions.
As an SLT member myself, I am so sick of the 'perpetual need to have evidence incase we cannot see it in classrooms', it makes me want to vomit.
Once again, it comes down to executives in charge needing to have a tangible measurement to justify their often huge salaries. If it can be measured the executive class in education (and I include Ofsted in that bracket) want it regardless of whether the actual data they are auditing means diddly squat.
I think you're right. I'd go further and say that much of the marking and comments are aimed at the parents rather than the pupils. I remember visiting a school as an L.A. adviser and puzzling at the lengthy comments on FS2 pupils' written work. The head was open that this was for the parents' benefit, a high percentage of whom would come into the class every morning to look at their child's books. It was evidence that the children's work was regularly assessed and commented on by the teacher.
Having been part of this, I did have some thoughts about why some of the children appeared to go ‘backwards’ after the feedback.
What I noticed when administering this was that the higher attaining students tended to want to ‘write a new story’ following the feedback, whereas those who struggle more with writing adapted their original writing in line with the feedback they received. This inevitably improved the lower attaining writer’s work but potentially stifled the more creative writers as they tried to include the feedback prompts more than they tried to maintain their authorial voice. This was what I thought I saw anyway.
I wonder if the 36% deemed to move backwards were more likely to have started a new narrative/had a higher scaled score from the first draft?
Thank you for this Daisy. I hope all senior leaders read this in the UK. Not just because of the pit falls of written feedback from AI but the whole concept of written feedback. As an SLT member of staff I despair when I am asked by the head teacher to do a work scrutiny. The whole concept of written feedback iny opinion is ineffective and almost pointless. Dylan Wilian as always puts it best. I cannot help feeling the whole approach to written feedback is purely to justify senior leaders jobs and entertain Ofsted when they arrive to judge.
20 years and I still haven't seen any written feedback that led to significant improvements in students scores.
Writing is one of the things I struggle most with, as a teacher. I've tried to have my students use LLMs to review their own writing, but the whole thing didn't prove to be as straightforward as I thought. I told them to use AI to produce a three-column table with their text, a review and a rewriting of their own text with the fewest changes.
As a last step, I wanted my students to copy the revised text into their notebooks by hand. One could argue that this kind of work was sort of passive or mechanical, but I wanted my students to see the difference between their original text and the revised version.
It seemed to me they hardly noticed the difference.
In another occasion I skipped the AI-rewriting, to make sure my students would make a reasoned effort by rewriting themselves, but this wasn't really successful either.
There are some practical hurdles to overcome before I can actually say AI revision is useless. For example, in some cases my students had submitted pieces of writing that didn't need, for some reason, much revision.
My enthusiasm has been eroded, though, and I am far less enthusiastic now.
Many years ago we had a Writing Workshop approach to writing, where students drafted first - sometimes more than once - then ‘conferenced’ with the teacher before producing a final version. That way, the teacher knew the individual student’s capacity, and could pick up on any systematically misunderstood features in general. Surely, reading student work is the way we actually come to know our students.. and even a few words means more than a mechanical ‘reply’.
Really enjoyed this, thank you. It makes me think about some advice I was given about formative feedback: focus on the HOW to do better, not the WHAT to do. As the excellent 'systematic' anecdote said, 'if I knew how I'd have done it the first time'. My worry with automated feedback is that the practitioner may not have enough of their own cognitive skin in the game.
Reading automated feedback is not the same as unpicking the text for why something didn't work. If the practitioner hasn't had to think about it, they are unlikely to fully understand it. If memory, after all, is the residue of thought, then that applies to staff as well as children.
My ongoing worry with AI in schools is that we are pursuing efficiency at the expense of efficacy.
Very off the cuff and I may well have missed the point entirely but very much enjoyed it so thank you for sharing.
Yes, we think about this a lot. The challenge is to make sure that the time and effort teachers spend on assessment is manageable and as high-value as possible. We don't want to eliminate it - we want to reduce it and focus it on the most valuable aspects.
"My ongoing worry with AI in schools is that we are pursuing efficiency at the expense of efficacy." Love this!
Thank you.
I would hope you would expect me to respond :)
My first question, having tried this out this year, is about the "control": while re-drafting with the AI feedback has produced an improvement, what would be the equivalent outcome for re-drafting without specific feedback - either more generic 'whole class' pointers, or even no feedback at all. I have a suspicion that a clear focus on fewer points of generic feedback might be more effective for those who most need it.
Yes, definitely. Fair point re the control. Also the experiment I really want to do is to re-assess the students on a cold task and see if improvement is sustained. But it's hard to do with this cohort as they go off to secondary. We'd need to run a Y5 redraft and then measure improvement from there to Y6. And we have approx 10 years of non-AI 5 to 6 average improvement, so could use that as control. But we are super busy at the moment and I don't know how much appetite there is for a Y5 redraft.
Yes I can imagine it'd be difficult to persuade people to add another assessment - particular just to prove my point 😂
It's a shame we're not spending more money nationally in projects like this to address teacher workload rather than vaguely-worded charters!
Having said that, I spoke to someone else today who said please add redrafts to every assessment! Will have to see...
Vaguely-worded charter tyranny, tell me about it.
Thanks for the post . Loved the Dylan William anecdote, especially yhe student replying, “I don’t know. If I knew how to be more systematic, I would have been so the first time.”
This is such a powerful argument against so much marking that still happens in schools particularly in primary. If I'd known how to spell it correctly I'd have spelled it correctly. Most kids don't make errors on purpose, they make errors because they haven't learned the correct versions. As to reading feedback I really don't think most primary learners do that. They'll look for a mark, a grade or a smiley face but in my experience they won't often read, understand or process written feedback.
I agree. I think that in the current system, most pupils (and parents) do see a written comment as a sign that their teacher cares about them and has given them time and attention. So the written comment may well have motivational value, which is important. But that is totally different from understanding and acting on what has been written!
That is exactly the problem with the system we have created. It is less to do with does it help a student improve and more to do with perfunctory obligation to placate the vested ignorant (or willfully ignorant where Ofsted and the higher escelons of a school are concerned).
I wonder if today that all staff in every school just said no to written feedback would outcomes change at all?
My gut instinct is we would muddle along just fine (even senior leaders might be okay with the time saved by eliminating the obsessive need to do a work scrutiny). It would especially be so if we could now focus our attentions on improving the efficacy of verbal feedback in the immediacy of a learning episode.
I fear that many SLTs, now ,would consider your proposal way too radical especially in the current climate with the irresistible rise of multi academy trusts where policy is often centralised and many primary schools no longer have a headteacher but rather a head of school which is absolutely not the same role.
No truer words spoken. Unfortunately where bravery is needed to eliminate the countless hours of wasted bureaucracy we are led by lambs, rather than lions.
As an SLT member myself, I am so sick of the 'perpetual need to have evidence incase we cannot see it in classrooms', it makes me want to vomit.
Once again, it comes down to executives in charge needing to have a tangible measurement to justify their often huge salaries. If it can be measured the executive class in education (and I include Ofsted in that bracket) want it regardless of whether the actual data they are auditing means diddly squat.
I think you're right. I'd go further and say that much of the marking and comments are aimed at the parents rather than the pupils. I remember visiting a school as an L.A. adviser and puzzling at the lengthy comments on FS2 pupils' written work. The head was open that this was for the parents' benefit, a high percentage of whom would come into the class every morning to look at their child's books. It was evidence that the children's work was regularly assessed and commented on by the teacher.
Having been part of this, I did have some thoughts about why some of the children appeared to go ‘backwards’ after the feedback.
What I noticed when administering this was that the higher attaining students tended to want to ‘write a new story’ following the feedback, whereas those who struggle more with writing adapted their original writing in line with the feedback they received. This inevitably improved the lower attaining writer’s work but potentially stifled the more creative writers as they tried to include the feedback prompts more than they tried to maintain their authorial voice. This was what I thought I saw anyway.
I wonder if the 36% deemed to move backwards were more likely to have started a new narrative/had a higher scaled score from the first draft?
Thank you for this Daisy. I hope all senior leaders read this in the UK. Not just because of the pit falls of written feedback from AI but the whole concept of written feedback. As an SLT member of staff I despair when I am asked by the head teacher to do a work scrutiny. The whole concept of written feedback iny opinion is ineffective and almost pointless. Dylan Wilian as always puts it best. I cannot help feeling the whole approach to written feedback is purely to justify senior leaders jobs and entertain Ofsted when they arrive to judge.
20 years and I still haven't seen any written feedback that led to significant improvements in students scores.
Real time verbal feedback however...
Really great post. I love being made to think, not just do the marking and really like tbu, really opened my eyes.
Writing is one of the things I struggle most with, as a teacher. I've tried to have my students use LLMs to review their own writing, but the whole thing didn't prove to be as straightforward as I thought. I told them to use AI to produce a three-column table with their text, a review and a rewriting of their own text with the fewest changes.
As a last step, I wanted my students to copy the revised text into their notebooks by hand. One could argue that this kind of work was sort of passive or mechanical, but I wanted my students to see the difference between their original text and the revised version.
It seemed to me they hardly noticed the difference.
In another occasion I skipped the AI-rewriting, to make sure my students would make a reasoned effort by rewriting themselves, but this wasn't really successful either.
There are some practical hurdles to overcome before I can actually say AI revision is useless. For example, in some cases my students had submitted pieces of writing that didn't need, for some reason, much revision.
My enthusiasm has been eroded, though, and I am far less enthusiastic now.
Many years ago we had a Writing Workshop approach to writing, where students drafted first - sometimes more than once - then ‘conferenced’ with the teacher before producing a final version. That way, the teacher knew the individual student’s capacity, and could pick up on any systematically misunderstood features in general. Surely, reading student work is the way we actually come to know our students.. and even a few words means more than a mechanical ‘reply’.
The longer term studies will be really telling. What happened next for the 36%. For me, you are asking all the right questions.
This is such an important distinction.
AI feedback is not automatically useful because it is personalized, detailed, or well-written.
The real test is whether it changes the learner’s next action.
A comment can be accurate and still be unusable if the student does not know what to do differently.
That is why feedback needs to move from description to guidance:
What is weak? Why is it weak? What should the student try next? What does better look like? How can they practise the missing skill?
For AI in education, this feels like the real frontier.
Not generating more feedback faster, but designing feedback loops that students can actually understand, act on, and internalize.