Feedback should improve the teacher, not the lesson

A challenge that sometimes presents itself when giving feedback to students is that their work is already of a pretty high standard, and it feels like we’re really nitpicking with our improvement points. The reality is that what they’ve produced is probably already near the top of the mark bands, and one or two small tweaks might not necessarily make much difference. It’s one of the reasons why I don’t like giving marks on pieces of work (other than under meaningful test conditions) because it moves the focus away from how to get better. One argument I’ve found useful in this context is to point out that my suggestions for improvement might not really make much impact on this piece of work, but they could be important on future pieces of work. This, I think, exemplifies the well known Dylan Wiliam adage that feedback should improve the learner, not the work. In this post, I want to consider how that principle extends to the feedback we give teachers too. I’m looking at this from a coaching perspective, although I think the principle should apply in other contexts too.

I’m observing a lesson in which a language teacher, Dave, is introducing some new vocabulary to the class. 

Dave projects a page from the students booklet onto the board which has a list of 10 different words in the target language. Dave points to one of the words, demonstrates how to pronounce it, gives the translation, repeats the pronunciation, and then asks the class to recite it in unison. Dave moves onto the next one and repeats the process. 

At this point, I have a thought. There are a lot of words on the board, and while the teacher is trying to direct their attention, it’s quite likely that some of them will also be looking at the other words at the same time. Having read Adam Boxer’s post on the hypothesis model of observation, I make a prediction that some of the students will struggle to later match up the pronunciation and/or translation of the word with the written version because there is too much information to process in one place (cognitive overload). So I quickly shuffle over to a student and ask them to say a word they’ve just heard and point at it on the page. To my slight surprise, they do this correctly. I go and check with a few different students and get the same result. Oh. 

There are (and have been), of course, times when my hypothesis will be correct, and that cognitive overload has interfered with students’ learning. And it’s a reasonably safe bet that this kind of overload is still happening in many classrooms because getting teachers to change their habits (or even their slides) can be pretty tough, even when there are very obvious reasons for doing so. In which case there’s a simple action step to follow which can remedy this. In this instance, however, it turns out that my prediction was wrong. 

What are the implications of this?  Does this mean that the mental model underlying my hypothesis was wrong too? Does it mean I don’t need to worry about cognitive overload with this teacher or these students? Does it mean that Dave is aware of issues with cognitive load but has carefully considered it in this context and (correctly) judged that it’s unlikely to be a problem? This is where the diagnostic part of the coaching conversation is so important because I need to determine the extent to which Dave is or isn’t aware of the potential problems with cognitive load, and therefore the extent to which this might need to be addressed.

In this instance I might decide I need to look elsewhere for an action step, and consider other elements of this particular activity, or other phases of the lesson, because telling Dave to consider changing the way he’s presenting vocab to the class doesn’t feel like it would have made any significant impact on the learning in this lesson and he might legitimately reject this as an area for improvement. Even if I have reason to think it might still be important, it might be that I feel this isn’t the most important thing to work on straight away; perhaps it doesn’t represent the highest leverage action step*.

As a coach, choosing the right action step for a teacher to act on is key. As I’ve written about here, this is not a simple matter. A coach needs to determine an action step that is achievable for their coachee, that will have a significant impact on student learning, that the coachee is motivated to work towards, and that does represent too big a departure from the coachee’s existing mental model of teaching. No tall order then!

In this context, then, I think there’s room to extend the hypothesis model. Rather than just making specific predictions about the outcome of one specific phase of a lesson, we might make some slightly more general predictions about how learning might pan out in the future. Perhaps cognitive overload wasn’t a problem in this specific moment, but, left unaddressed, could it cause problems in the future? Let’s consider what happens in the next lesson I observe with the same class:

Dave is now getting the students to use some new vocabulary in a sentence. He explains and models the sentence starter he wants them to use. He then realises he wants students to be able to say this in response to a specific question (so that he can get students practising basic conversation) and so teaches them how to ask that too. Then the class is instructed to have a go practising the question and response in pairs.

Now I have a new hypothesis (albeit a tentative one): the addition of the new information (the question) might interfere with students learning the original sentence. This time, though, I don’t even need to go and ask students individually to gather evidence because as soon as they start their conversations it’s clear that many of them are unsure on the precise phrasing they’re meant to be using, and are muddling up elements of the question and answer. So perhaps some feedback around cognitive load would have been useful after all. It might not have made a discernible difference to the specific activity in the original lesson (or the same activity in a future lesson), but it might have helped to avoid problems further down the road. 

So to add to my list of things to consider when determining an action step, I need to consider the extent to which the problem I think I’m seeing might manifest itself in a number of other situations. This also gets around another problem with instructional coaching in some contexts: that an action step might be so content-specific that it’s not going to solve a problem that the teacher will encounter again until they teach the same lesson a year later. 

There’s also a tension to acknowledge here between the generic and the specific – something which is also often raised as a challenge for instructional coaching. Our feedback needs to be precise enough that it respects subject-specificity and doesn’t seek to implement strategies which might well work in one subject but not another. At the same time, it needs to be generalisable enough that it’s going to have a lasting impact on the teacher’s practice. 

This is also why the dialogic nature of a coaching conversation is so important. The coach, with careful questioning, firstly needs to make a diagnosis of the alignment between their own and their coachee’s mental models of teaching. They also need to ascertain that any agreed action steps are fully understood by the coachee. This means not just in terms of their immediate utility and the precise enactment (how to do it), but also in terms of the underlying principles and insights that they are derived from. In this way, our feedback is more likely to improve the teacher, not just the lesson.

Summary

• Generating hypotheses is a useful way to identify potential action steps for a teacher
• Hypotheses need to be tested which means gathering evidence from the lesson
• Evidence which appears to refute a hypothesis might only be valid for the specific context in which it is gathered; the theory or insight underlying the hypothesis may still hold true in other contexts
• Action steps need to be selected on the basis of their long-term impact; this might not necessarily be the thing that would have had the greatest impact on that specific lesson
• Action steps need to be co-constructed with a coachee to ensure they understand the underlying principles of the strategy being suggested
• High leverage action steps will be those that allow transfer to multiple situations, to increase the adaptive expertise of the teacher

*There’s a separate discussion to be had here around how much feedback we might want to give our coachees. Certainly, in some cases, multiple pieces of feedback and action steps may be appropriate, but this won’t be helpful in a lot of cases if we want teachers to be able to focus on embedding something new properly.