Download transcript PDF

 

00:04

Uh this session today is all about the AI assisted language learner which you maybe you think um does not apply to you but I I would argue that for many of us it is our students are using AI more and more and we'll be talking about that in a little bit. So how can we help them use it in the best way for um self assessment in particular? So what we're going to be covering is first of alla why focus on self assessment why it matters to our learners then how can we help our students make the most of AI technology

00:44

to enhance self assessment and along the way we'll be looking at principles prompts and strategies uh to help do that. So, we'll be looking at lots of examples, practical examples. And I've used Chat GPT for that, but you can use Claude or Gemini or Copilot, uh, whichever you prefer, or you could do mix and match. I quite enjoy um using different ones to see how they they react differently. And I'll also show you some examples from McMillan's AI enhanced learning platform because it illustrates some of the

01:21

points quite nicely as well. So, uh, we will get going in just a second, but just to say please do use the chat to to interact to share any experiences you've had with your students using AI. Um, because I think we can all learn a lot from each other in the chat as much as what is on the slides. So, let's look at self- assessment first of all. Why are we focusing on that in this talk? So, why self assessment matters? first of all is that if it is done well, it can help our students really progress. So if we do self assessment

02:02

well as a learner, we become more aware of our strengths, we become more aware of our weaknesses and then we have a better direction of where we should focus our attention. So that could be very helpful when we're preparing for exams. But even if we're not preparing for exams, just if we want to keep improving, which I'm sure um we all do, then self assessment can be a really great tool to do that. It also helps our learners become more autonomous or more independent. And that's something that I think as

02:37

teachers we're always trying to do because it is important. We can't always be there and we don't want our students to be completely reliant on us uh for feedback, for guidance. Of course, we play a really crucial role in that. But 100% of the time uh is not possible for anyone even if you're a one-to-one tutor. So learner autonomy is really key. And this links to the final point as well that we want our students to become lifelong learners. We know ma now more than ever and with AI especially disrupting so many parts of our lives

03:13

that we want our students to be able to adapt to learn new things, we're going to need to do that in the future. So learning how to selfassess and guide yourself along a path can be really really powerful for becoming a lifelong learner. But there are some challenges to self assessment. So I don't know if you have experiences of these challenges. Um I certainly have found this when I've done my own self assessment as well as using self- assessment with students. But we often have the issue that it's difficult

03:52

to assess accurately to assess yourself accurately. So you might have students who are really um overestimating their uh performance in in a piece of work or if you're like me, you underestimate it and all you see is the errors everywhere. Um and you can't get a clear picture of where you are and where you need to go next. So that's one issue I think that is really uh really important to address. When we're learning, it's also quite difficult to spot certain types of errors or things that we should improve.

04:31

So, for example, if your students do a bit of writing and they're they're self assessing and they see, oh, I forgot to put the third person s on on a verb like he plays. I wrote he play. That can be quite easy to spot because probably it was just a slip. I didn't uh it's not because I didn't know that. I just accidentally wrote too fast and and it it slipped my mind. But other errors, we have to have quite a deep knowledge sometimes to be able to spot them. And that can be again a challenge if we don't know something

05:10

enough. So we might not realize what what would be the next best thing to work on. And finally, it can be hard to focus on what is really important. So there's a great temptation when you do uh self assessment or certainly when I was doing self assessment for the first time to just look in detail through every single thing that you've written and try and correct everything and you end up with a whole page of of scribbles and and things. Now that can be very helpful if you're trying to edit something and

05:47

you're trying to polish it to become perfect. But if you want to use self- assessment to help you improve, you need to narrow down your focus. You need to really think about what is the most important thing for me to look at in my self- assessment, not look at absolutely everything. Um, but it takes quite a lot of um support for our students to be able to do that. It's not a natural thing, I think, to be able to self assess well. At least let me know if you've got any students who are just instantly brilliant at self- assessment,

06:15

but I have not uh come across any yet. So the question is can AI help us do self- assessment better and get more out of it and in doing so improve our learning. So, in order to talk about this, I just want to get a bit of context from you. And you'll have a poll coming up on your screen in a second that is going to ask you, are your students using AI for learning English as far as you know, yes, no, which is now choice two and I don't know. And the second question is, if so, what effect do you think this is having

07:02

on their learning? Is it helping them? Is it harming their learning or is it doing both? So, I'll just give people a couple of uh seconds to choose some options there. And I'll just look in the chat as well in the meantime. So, there's a bit of praise for AI. Um yeah, Christina talks about unfortunately some of them don't know how to use it and uh sorry I've just lost it as more people have written. Uh so AI has become a tool to help them write emails at work. Yes, I had someone in a previous session say I now get perfect emails

07:42

from my students uh because they've put it through uh chatgpt or another chatbot. Mostly helpful. Yes. don't think it's harmful as long as we teach them exactly how to use it. Uh, and there's this cheating um issue as well. Uh, copying and pasting. Okay, brilliant. If we can close the poll and just get a sense of roughly what the the figures look like. Okay, so it looks like 70% of you are saying that your students you know that your students are using AI. um 10% say no and 20% are not sure. Um

08:23

and of course that will probably depend on the age of your students. So I imagine if you're teaching pre- primary or primary it's unlikely that they'll be certainly independently using AI. Um but at adult and secondary we see a lot of usage now. So um if so is it helping actually that yeah okay the second highest answer is it's helping their learning with 24% 9% say it's harming and the majority of you are saying that it does both it helps and harms learning um depending I guess on on how they're using it.

09:03

So let's look at how we can use AI to really help learning rather than hinder it. Because what you're saying in this poll is very similar to research you see more broadly. I'll just quickly outline what we're seeing more generally. So students are using AI a lot as I say from secondary upwards. And there's a real mixed picture. So some are really getting a lot out of it, but for some they might be using it as a shortcut to offload their thinking. So they are not using their brains really. They're not

09:43

engaging in critical thinking, just um typing something into a chatbot. Um they could also use it just as a magic answer generator or they think of it as a magic answer generator. Um, so this is perhaps where the um the cheating also comes in where you just say I need to write a report on this and then please write it. Um, and then it comes up with this beautiful report for you. And in general, they're using it in ways that are not supporting their learning. But interestingly, a lot of students in in research uh that that has

10:23

been done believe that they are learning something from AI. We'll look at an example in a minute, but the reality is that they aren't. And we'll talk about how we can get over that in in a moment. So, let's look at an example. Imagine that you have set your students this writing task from Language Hub where they have to write a report on trends in communication and they have to write a certain number of words, they have to use formal language and so on. So just imagine that you have asked them to write a draft and they really have

11:00

they really have written their draft perhaps in the class already and you've told them now to self assess to see what they could improve or what um what kind of things they could work on. So many of our students might do this write into chat GPT please correct and then write their draft. Um so chat GPT of course will say yes here we go here is a corrected and polished version of your text. So I'll zoom in a little bit to show you the kind of things that it did. So in my original I said in the modern world the

11:43

communication is changing very fast. Technology make big changes in how people talk with each others. This report explained two important trends that I notice in communication today. So there are a few little errors um in there bit of uses of tenses that could be a little bit better. And what AI did was just rewrite and replace the the errors with um the correct version. So I'll just show you I then asked it to highlight um what what it changed and it showed me here that it changed the word communication it took away the article.

12:22

So we don't need the for communication. And instead of saying fast, it changed it to rapidly because the the kind of tone and the the formality is better when we use rapidly for a report. Um and then you can see there's some present perfect it's it's put in and suggested where I just use present simple and those kind of things. So as a student you might think, great, I love this. I can see where my mistakes are. Um, but m maybe just put in the chat, do you think students have learned much from this? Or maybe if I rephrase the

12:59

question, when they next write a report, do you think that they will take on the um the all the corrections that they've had in in this uh report and bearing in mind the report is about this long and they have about as many um corrections there. Uh so yeah, I don't think they will learn like this. Um, no, no, no. Unfortunately, yes, I like that. Unfortunately, um, yes. So, they won't notice their mistakes. So, we talked about the one of the reasons for self assessment being so powerful is to raise self-awareness and

13:36

to notice um to notice where your strengths and weaknesses are. But if you have just a whole page of highlighted changes, it you have to work quite hard to to do something with that. But many of our students will stop at this point and think, okay, I've I've got these corrections. I have learned something from doing this. So, this is what we want to avoid. This is not um helping them learn massively. And I just want to put a quote up because I think it sums up what we're trying to do. Um this is a um a summary

14:16

of the research that has been done so far on AI and chat bots in in learning in general um by Ethan Mollik who I recommend following on LinkedIn um or wherever you um are on social media because he has some really interesting things to say about AI and learning. So he sums up the research by saying if we just let our students use AI chatbots, this can undermine education because AI gives the illusion of learning while actually just providing answers. So like in the screen before they might have the illusion that they

14:53

are learning something but the reality is that they they are not. But the hope um the hopeful part of this quote is when AIs are properly prompted to act like tutors especially with instructor support so with teacher support this seems to be able to boost learning a lot through customized instruction. So this again represents what you were saying about it can both help and harm their learning. It depends how we use it. So, let's look at some strategies that might help get us closer to helping students boost their learning

15:32

through AI and self- assessment. So, we'll be looking at some principles and prompts and strategies now in uh that you can try out in your classroom. Yeah, just before we uh go to that, I completely agree most of the vocabulary corrected will be out of their mind. Thank you, Luis. Um yeah, they can see the mistakes but they're not aware of them. I really like that. Yeah, absolutely. Okay, so our principle number one is to use this idea of thinking about AI as a tutor. So that might seem quite obvious,

16:09

but the reality is that a lot of us tend to think of AI as something transactional. And I'm borrowing a um an analogy from Eric Martinsson here who talks about students treating AI a little bit like a vending machine. So you put a prompt in and you get an answer out and then you go away and you take that take that with you and that's the end of it. and he said, "If we have that kind of mindset, we're not going to get the most out of our students. Um, and they're not going to get the most out of their learning. Instead, we

16:48

should think about AI being conversational, being like a tutor who makes you think, who is a dialogue partner who asks you questions that are going to make you think and learn. So if we have that in mind, whatever prompts we we put into chat GPT or any other chatbot are likely to be more effective. So I was looking for a prompt uh that would illustrate this quite nicely and I came across this study in which researchers were prompting AI um or creating AI chat bots to help students learn. And they did this for

17:31

self assessment as well as some other tasks. And I thought it was really interesting to show this prompt. And maybe you can see how it's asking um AI to be like a tutor. So it says in case you can't see it, you are a teacher who facilitates self assessment for learning. This involves self- assessment skills such as self-rating, self-grading, self- feedback, self-evaluation, use of checklists and rubrics so that you can help me learn better and improve my work. When I ask for answers or examples, guide me in the

18:09

form of hints, clues, or suggestions one step at a time. Try not to give me a direct answer. Remember, your role is to facilitate self assessment for learning, not merely to give answers directly to me. So, that sounded quite nice to me. I can see um that it's asking it to facilitate self assessment for learning. It's not giving me direct answers. So, it's it's giving me hints and guidance, but it is making me think while it does that. So, I thought I would try it out with the same report that I created for

18:47

um the previous prompt that we looked at. So, I put this into chat GPT and it replied and said, "Great. I'm here to help you learn how to learn," which is a good start. So, it said it's going to guide me guide me step by step. And it starts off by asking me what I am working on and what I would like help with today. So then I very simply said, I am learning English. I'm writing a report. Here is my draft. And then I pasted the um the text right into the chat bots. I could have written more here. And maybe when you're working with

19:31

your students, you might ask them to experiment a bit with what you write in in there. I might have given more information, but I wanted to see what it did with just really basic um a basic couple of bits of information. So, I pasted that in and it came back and gave me a bit of praise, which I enjoyed. Nice job drafting your report. um and gives me a little bit of information about um my strengths and then tells me to break things down to look at the introduction and gives me a bit of a rubric to do

20:08

that. So, does the introduction clearly state the topic? Does it avoid repetition? Is the grammar accurate? And are the articles used correctly? So differently from the first prompt where it just corrected all those things, it's now asking me to to consider these myself. And then it takes me through step by step. So first of all, it picks out that first um part of the introduction and asks me, is there a phrase or a word that sounds a bit off or unnecessary for a formal report? And it gives me a clue to think about that

20:44

sentence. The communication is changing very fast. um and then asks me to think about which word might not be right. So I have a guess and I say is it fast and it replies and says good thinking fast is a bit informal for a report but instead of then correcting it it asks me can you think of a more formal synonym for fast that would better fit and again it gives me a bit of a clue. So it starts with R and is often used in academic writing to describe quick changes. So I have a bit of a guess and say, "Oh, is it

21:21

rapid?" Add. Yeah, exactly. Rapid is a much better fit here. Um, so then it gives me the corrected sentence. So it's not perfect. you can maybe think of ways that it you would want to change what it did, but it's a lot closer and it's getting our students to think a lot more deeply about what they're doing than our please correct prompt. So, there's a lot more value to this. Um, if you're interested in trying this prompt out yourself, you can uh scan the QR code or go to the link and that's got a slightly longer version of

21:57

the chat that I just put in there, including the prompt. So you can copy and paste that prompt um if you like. I'll also give you a link at the end to the research paper um where I found the prompt um because it also has some other prompts in there as well that might be interesting. So that's what um AI as a tutor might look like in a chatbot such as Chat GPT or Claude. And I just wanted to also show you what it might look like in a learning system. So this is Pivot English which is an AI enhanced learning

22:35

um environment from McMillan. And you have AI's tutor really or AI playing the role of a tutor um built into the system. So it highlights things that you might want to look at. It gives you hints and then adapts depending on your strengths and weaknesses. And I think what is nice about learning environments compared to chat bots is that at the moment unless you are incredibly well um or you're quite an expert in in prompting and you might have a high level kind of version of of chat bots. It's quite hard for you to see progress

23:19

over time because you're you're just focusing on one piece of work at a time and then it it kind of forgets. The next time you come, you're a new person and it it can't build that up over time. Whereas in the system, you can see where over say a month or a week or even a term where your strengths and weaknesses are. Um, and not only can you then build your self-awareness of your strengths and weaknesses, but then as I said, it's an adaptive system. So then it can start uh giving you more practice on those

23:55

areas that you're not sure of. So we're seeing now AI being used um a lot more powerfully um than we could have imagined I guess a couple of years ago. Okay. So principle number one, AI as a tutor. Principle number two is to use effective learning techniques in your prompts. So we already had quite a nice example actually in our prompt from before. If you look at the bit that's underlined, we've asked it to guide me in the form of hints, clues or suggestions one step at a time. So as teachers, we know that that is much more

24:34

effective for learning than just telling our students the answer. So when we elicit answers from students in class, this is the same kind of thing that we're asking the chatbot to do. So sometimes we just need to think about what we would do as a teacher and and use that in a prompt. But if we're getting our students to write the prompts, they need to know as well what these effective learning strategies are. And I personally think that it would be great if we um we could help our students understand more about how their brains

25:07

work and how then we have to use certain strategies that will be more effective than others because if we don't then we can't be too surprised if um they're using AI in a way that doesn't enhance their learning because they're not um yet aware of how how it um affects learning. So, one example if you want to go a step further um is by learning or teaching your students about learning strategies. Now, it's probably a little bit small here, but there's a nice website called the learning scientists and they give you uh

25:47

six strategies for learning that are based on psychology research on the brain and memory, how we learn. And there are two strategies in particular that have got the most research behind them and are probably the easiest to to implement quickly. Um and they are spaced practice. You might have heard of these already. And retrieval practice. So space practice is this idea that we need to not just learn something once. So for example, we've we've written our report and we learn something new. We learn a new bit of

26:21

vocabulary or a new bit of grammar. But if we never pick it up again, then it's unlikely that we'll be able to use it next time. But if we maybe look at it again the next day and then maybe wait a week or and then a month and we keep returning to it, then it's going to more solidly embed in our into our brains. And retrieval practice is basically saying we learn a lot better and a lot more deeply when we have to remember things. So if we test ourselves on the new knowledge that we've we've learned

26:59

um there's again what feels good for for us when we're learning is not necessarily reflecting how good a strategy is. So I remember when I was teaching Alevel in the UK, um I was teaching psychology actually in the UK and um my students when they had an exam they used to get their notes out and they used to highlight in different colors or some of them would copy out their notes. They would copy out exactly word for word their notes from the class and they said to me that they felt this was a really they they were learning

27:36

loads. It's like oh yeah I remember it. while while they're reading like, "Yeah, remember this. Remember this." But actually, there's loads of research now that says that feels great, but you're not actually learning. If you make it really hard and like test yourself all the time, that is when you're learning because actually by testing yourself, you are strengthening the memory for that thing. So, that was a bit of a ramble into uh learning strategies, but think about how maybe you can extend these kind of prompts and

28:07

experiment with um with these learning strategies when you're doing self assessment. So, you've maybe you could say at the end of the self assessment, give me some ideas or help me create some retrieval practice questions. So some questions that I can test myself on next week to make sure that I I learn these new things. All right. Um so just again as an example in a different context. So in pivot English you have a memory training area which uses both of those techniques, the spaced practice and the

28:46

retrieval practice uh techniques. Um, so this is really nice now that we're seeing these strategies that have been that had loads of research behind them are now really coming into learning products. So great. All right. So that is our second principle about using effective learning strategies. The last principle is to say that we should experiment and reflect. There are no easy answers. There's no perfect prompt. Um it this is all new. It's all changing all the time and no one I say is or the vast majority

29:29

of us are not experts. I do not consider my myself an expert at all in in AI. So what we need to do I think is encourage our students and encourage ourselves also to experiment and reflect and that is how we'll start to build up with the skill. So you can either just use that with the prompts. For example, the prompt that um I showed you earlier, the one that was the longer one that was um better for learning, you could experiment with with trying out different things, changing little bits and pieces to to improve the

30:04

prompt. If you want to go even another step, you could ask your students to build their own prompt using something like this. So this is a build a bots website from uh Stamford University and they have broken down how to build up an effective prompt into blocks basically. So it's nice and scaffolded. You can focus on just a few of them or all of them. Um so you can adapt it to your students and it just gives them lots of ideas to to experiment with and then you can try out the the bot as well. So, for example,

30:44

you need to choose who you want your bots to be. Do you want it to act like a teacher or a singer or anyone you like? Um, what kind of things do you want it to to ask you to do? Um, and that can actually be a really good speaking and writing activity, collaborative activity, as well as being good for helping students create good prompts for self- assessment. In fact, they even have um things that you can choose um for examples and they have spaced repetition in there for example. So you could practice

31:22

that. All right. But yeah, you can use that website. It's you don't need to log in or anything. It's it's just completely free. But whatever way um you do it, I would really encourage you to experiment and get your students to experiment. Talk together about what you find. reflect and then refine your prompts. So that is principle number three and I only have three. So that means that we are coming to the end of this session and I just want to wrap up by uh talking about what we have discussed today. So the first thing is

31:57

that self- assessment can boost student learning when it is done well and AI can either short circuit this so it can probably completely take away the need for students to self assess at all which we do not want or it can enhance our students self assessment and that's obviously what we're aiming for and if we use these three principles so AI as a tutor getting them to think using effective learning techniques in our prompts and experimenting and reflecting and refining. I could add to that, then

32:33

we will be getting some way to get students to to use self- assessment better and improve their learning in doing so. So, I will leave you with a a challenge to think of one way you might be able to introduce AI assisted self- assessment if appropriate in your classes next week. And apart from that, thank you very much for listening. These are my references um which I encourage you to check out. The first one is the research paper that has the prompt in and others. The second one is a a nice summary of self-directed learning and

33:09

self-regulated learning or what the research says about using AI for those things. And the last one is the learning scientist website um that they have really nice simple materials to help students understand how they learn.