Preview Mode Links will not work in preview mode
Welcome to the AI in Education podcast With Dan Bowen and Ray Fleming. It's a weekly chat about Artificial Intelligence in Education for educators and education leaders. Also available through Apple Podcasts and Spotify. "This podcast is co-hosted by an employee of Microsoft Australia & New Zealand, but all the views and opinions expressed on this podcast are their own.”
  1. All Episodes
  2. AI - Hype or Hope? Live!

AI - Hype or Hope? Live!

Mar 25, 2020

Recorded live during the ISQ Educational Data Symposium in Brisbane, Queensland, this podcast takes a new approach. We talk about data and whether we need standardised testing, and then dive into answering questions from the audience about everything from predicting student progress, to how AI can help in the administration of a school. It's our six month anniversary podcast, and a really deep conversation that takes us back to the Hype versus Hope discussion!

And right at the end, Dan makes it clear that we don't know all the answers, and the podcast has been a journey of discovery for us too

 

________________________________________

TRANSCRIPT For this episode of The AI in Education Podcast
Series: 2
Episode: 12

This transcript was auto-generated. If you spot any important errors, do feel free to email the podcast hosts for corrections.

 

 

 

Hi, it's Dan here. Welcome to our latest episode. We recorded this live at the Independent Schools Queensland Educational Data Symposium. We did have a live audience, but the travel restrictions meant that we all took part remotely. So, Ray and I were stuck in our homes in Sydney and the questions came in from all over Queensland. We really hope you enjoy.
Dan,
yes,
we need to have a serious conversation because we've been doing for 6 months now. In fact, this is our 25th episode of the AI in Education podcast. We've been rambling on about AI in education for 6 months and we said that it's time for us to take stock of where we are and to get some real input into the conversation rather than just us wibbling on. So, the question is, Dan, we said this podcast episode is going to be titled hype or hope. What's your feeling at this point given the number of people we've talked to around things?
So, I think I think it's I'm sitting on the fence here. I think it's a little bit mix of both. I think there's a lot of technology. What I would say is with the use of data and the abundance of data now in education and the way this is coming very very quickly at schools, I think I don't think it's as gimmicky and as hype orientated as um you know some other things in the past around educational kind of transformation. I think um this is moving forward very quickly. I think the the hype is real in lots of cases but there are fringe cases where we've seen such as um facial recognition and things where we may be looking at um uh some of those outlander or outliers. Sorry, not outlander that's a TV show out out ers in the in the area. So I think yeah there is there is some hype around this but I think around big data specifically and machine learning and the actual use of AI to predict things like results against ATR and things that that's actually very achievable and lots of schools are doing that now.
Right. Okay. And and I agree with you because you know I find it quite interesting we're in the middle of a of an interesting digital revolution aren't we? You know let's be honest and it's all happening in a week. It's we're we're not waiting for this digital online revolution. You know, it's it's coming at us ready or not.
But I think one of the questions that I sit thinking about is well, have we done have we have we taken enough time to prepare and are there things that we could have been doing up until now? Because I think we've been having a lot of conversations about AI, but the implementation of it is one of those things that's suddenly having to have a happen at breakneck speed. So the example I would use is outside of schools is the universities. They're suddenly having to think about how do they cope with this challenge of supporting students that initially it was the ones that couldn't get into the country because they were stuck in China and they had hundreds of thousands of calls to deal with from students and so some of the universities very rapidly deployed AI to provide virtual agents conversational agents that students could talk to and it's really interesting that we'd been having the conversations but not implementing the technology because of all the organizational ways of doing things which tends to move very slowly and then suddenly my goodness you know, how do we get this done in five days? How do we do this fast? And so, you know, it's really exciting that the theories we've been talking about in AI come to value, but it's how how do we do this at a broader scale, I guess. And that's going to be the difference between hope and hype.
Yeah. And and I think it is positive. Listening to some of the conversations in the room earlier on just when we were testing the microphone, I think there's some amazing ideas the schools are actually taking on board. And just actually in this live podcast event now, the folks in the room and online listening in have done some great input during the day and sent us some great questions. So, we can we can tackle some of those in a minute. But actually even having this event and having somebody like Kate who's kind of running this at at an entire district level uh up in Queensland and actually bringing these conversations to the front and even towards the end there she shared two amazing books for people to go and read uh based on Australian educational um uh data use. So, so we can put those in the show notes as well. But I think even just having this event is is fantastic and it's not it's not happening in in many states in Australia at the minute. But I think there is hope because there are dascese, large school systems looking at data now and thinking about how we can use that.
Okay. So Dan, how are we going to handle the questions because got a couple of interesting questions that um well you were given enough time to prepare for. So
shall I shall I read the question to you Ray and then let's
Oh, damn. I was going to ask you because you're the one that's prepared, Dan.
Oh, okay. Well, let well that's even better. Let me fire at you. And I've got my little points on there. But essentially, question one was from St. Andrew's Anglican College. Uh, and the question was, with a new Queensland senior ATAR system occurring for the first time this year, how can AI assist schools with confidential ATR predictions so that schools and students know if they're on track to achieve their tertiary goals as there are no past big data sets for some of the uh other exams to make such predictions. What big data sets would be needed for any AI predictions to occur and if that's possible at all or even if it would be reliable. Um well one thing I would say at the outset while you're having a think about that one Ry one thing I'd say just on on a tactical note here you know we just going through this pandemic at the minute and um And I think currently what's going to happen I was just reading lots of comments online around I know that's dangerous in current times but um with some of my colleagues in the UK and they're actually stopping all the testing uh this year. So the high stakes testing's being stopped uh nobody knows what they're going to do about there but I'm assuming based on past experience as being an ex teacher they rely on teacher assessments. Now that's an interesting one because when we do have the data set and and lots of data collected on students. And we do have a lot that are already in in the systems and teacher judgment, then you know, you can you can create a model to predict or or even at least give a a good indicator of where that student would have performed. So, so I'm when I was a teacher, I could probably tell at this time of year when I was teaching in the UK what really the kids are going to be getting at the end of their um exams because I've been teaching them for two years. I knew roughly where they were. I knew and I'd been, you know, changing their learning as I was going and I had lots of that recorded and I knew which students in my class would be achieving the top grades, the middle grades, lower grades and I was pretty accurate with that. So teacher assessment was still
So Dan, you're like a a walking AI machine is what you're claiming.
Of course I am.
So So I know you've got a couple of other points. I'm going to jump in and talk for a little bit and then I'm going to get to see you on camera cuz I want to see you talking. So, you know, it's interesting because as as I thought about that question, there were a number of things that jumped out. You know, you made the observation about the year 10 exams in the UK uh being stopped this year. Well, it's really interesting. We stopped what six years ago in Australia and nobody noticed because actually it's only the year 12 exams that matter. But it's really funny how in some countries they get fixated on some of those exam systems. So, that that was an interesting observation. But now into the AI bit. Um look, let me talk about two examples. The first is I think about the work that's been done by the Fisher Family Trust in the UK, which was all about predicting exam outcomes for students. So, as you went into year 7 in your high school, they would look at your achievement in uh your primary school and predict an outcome um of your grades when you got to year 11 and year 13 in the UK. I don't know where the extra year comes from, but it's great. Um but they would predict the outcomes based on the test that you'd done and then what they would do is go forward um with those students and effectively they were using it as a as a measuring stick to say to the students here's what the system forecasts you're going to be able to do see if you can beat it now this happened a long time before AI and ML it was some deep statistical science by uh the Fisher family trust which is an education trust in the UK but it became so reliable that almost every high school used it now the data that they were using was the test outcomes from students at the end of year six. Um but then in order to predict where they were going into the future, they were using whatever data they had available. So it was some some things around the socioeconomic status of the students. It was things around um some of their family circumstances, things like that. And then they refined that as that child went through year seven and they gave them a few more tests. They had more results to come from. And so that is really really fascinating because it implies to me that we may not have the perfect education data but there are other data sources that we could be using that acts as a proxy for some of the things.
So what are those data sources?
So let me move to an Australian example and let me move to an Australian Queensland example. As you would know Dan I have been fascinated by nap plan. Um fascinated in the fascinated in the sense that
you know uh it's it's interesting but It shouldn't be the heart of everything that we do, but unfortunately it becomes the metric. There are only six national metrics for education and nap planan results is one of them. And so what is beautiful about Queensland is that the education results from nap planan are published by the department of education for Queensland online. And so you can download a spreadsheet for the results for all of the public schools in Queensland and you can start to work with it. And what I became interested in is knowing that AI is getting easier and easier and therefore within my capabilities or heading towards my capabilities. Let's be honest, we were in a situation where I could then start comparing data sets and what I was trying to do was build an AI model that could predict the net plan scores for a school. I don't know if we've ever talked about this, Dan, but do you know where I ended up in terms of accuracy?
Um, go on.
We did it 90 odd percent, wasn't it?
You forgot it's 85%. Now, look, when people are talking to you about AI accuracy, remember that 50% is tossing a coin. You can get 50% accuracy with a coin. So, don't go spend any money on technology, but 85% is pretty good, especially when you consider that I didn't have any education data available. So, I had the nap plan results and I knew where the school was and I knew how many students that were there, but I didn't have any insight into what was going on inside the school and the few data points that I had made no difference to the nap plan results. So, what I used was ABS statistics data and what I tried to do was work out how little data I would need to maintain that 85% accuracy. And the answer was three pieces of data. I needed to know how many students were in the school. And the reason I think for that was the variation. If there were only 100 students, it might mean you'd only have 10 taking the test and therefore the data would be all over the place. The second piece of data I needed which was linked to the postcode was the parents education level. So what proportion of parents in the suburb are educated to degree level or cert level level and then the third piece of information I needed was actually income household income and out of that I could maintain a model that got 85% accuracy now I started pouring lots more data into it because with autoML it becomes really easy to change the data and run another test
you can explain to the listeners again and the people in the conference today what autoML is
autoML so
uh so autoML is auto machine learning so normally you'd get a propeller head data scientist to build you a system or to work with a complex system that can take data and predict the result by working out what influence the different bits of data have on the end result. So let's think of an example. They use it all the time in trying to predict whether if they show you this advert you'll buy or click if you buy this product that you'll also do this. It's about taking data from the past to be able to predict future behavior. And what I found with this with with AutoML basically you just give it the data and it says okay I'll try all the different configurations of everything and tell you what's important and so you know it becomes a mouse click to be able to predict numbers and so the really interesting thing was we may think we haven't got much education data but we've got a heap of other data so if you think about the data that's in your school information systems actually some of the information there like where the student lives that maybe the street that they live on you know a whole host of things. You can actually inform yourself with that data and it doesn't cost the earth in technology or time terms to be able to run a model and see if it works.
It's really interesting to think about that because I think you did answer that really well Ry with examples and everything which is fantastic. But yeah, I suppose a subset of that question was is it actually possible to actually do that and you were saying actually with 85% accuracy and you just hacking it together it was. And then you know in terms of our reliability 85% is pretty reliable but the data sources you used were really models that are data sources that were freely available in Queensland or nationally. Correct.
That's correct. And it's organizational not level not student level. I guess the question is what data do you have at student level but think outside of the box about what that data is because it may not be learning or what you think of as conventionally learning related data. So if I can if I can give you an example from higher education to be able to predict which students are going to drop out the most important things are things that are nothing to do with what happens in the institution it's about the students circumstances before they arrive are they the first student in the family to go to university are what is their uh score when they're coming into university those things turn out to be more indicative than many of the others so if you think about that in the education in the school context it's like well what what data have we got Um, and probably the other part of it is how much data do you need? Can you do it at an individual school level or do you have to do it at a district level? Do you have to do it at a dascese level? Do you have to be able to share with 10 other similar profile private schools the models so that you can build it?
Correct. Yeah. But but the other interesting thing in this question as well which is you know I'm obviously in K12 space um and you're in higher ed there. The other subtlety in this particular question was um the fact that they talked about tertiary. So what are the indicators? So if I'm a student, is it an indicator that um how well am I going to do in the future past K12? So if I if it's my if the ATAR is my key thing in Queensland, what about things that you know what data would I need to put into there to then think about you know what am I going to do in terms of a degree? How am I going to perform generally in life. I know this is kind of quite out there, but it was part of that question. I thought it was quite an interesting one. What about from a higher ed point of view?
Yeah, it's interesting. So, yeah, in Queensland, you work on the OP model um which slightly different to but slight slightly different enough. That's one of my learnings as I move to Australia.
New Queensland point of question is there's a new Queensland senior ATR system coming in which is the new
Yeah. So, I guess it depends on where you want predicts it. The problem that we have is that many of the data points we have in education are the old world data points. How did you do in your nap plan test which to be honest doesn't mean anything to an employer. How did you do in your skills test which may not mean as much to an employer. Um I I don't know if we've got the measures across organizations to be able to predict the things like you know how effective is this person as a communicator? How effective are they in being focused on outcomes which is what we care about as as employers. So
yeah I think I think it's it's the wild frontier and there's probably the world where people are trying to build systems with the data.
Yeah, brilliant. Thanks R. So question two was what areas and applications are having or will have the greatest positive impact on and this is from Dr. Peter Britain principal and CEO of such girls grammar school. What applications will have the greatest positive impacts on learning, teaching and then school operations. If I take the first one just for a second, um learning the biggest things in learning for me will definitely be about personalized learning. I think there's a lot of hidden elements which are appearing now in software and in tools and technologies that we are um seeing which is driving proper personalized learning. Going back to some of the other podcasts we looked at about David Kellerman in University New South Wales, you know, that's where our end goal is going to be and he's delivering that now. Um, I think it's that invisible element of learning that'll kind of support and help. Um, obviously I do worry a little bit in terms of learning around uh, as one of my colleagues Travis calls it cognitive amputation. You know, if we are taking a lot away from us um, you know, for example, Google maps for me, you know, I use that all the time to get around or ways. Um, and then eventually I start to need it need it more than actually use it as a tool. I actually need it now to go to my shop. So um you know there's oh go to the kids cricket so it becomes a bit of a cognitive amputation. So I think it's going to augment learning and it's going to personalize learning and that's what I'm going to I'm going to see lots of that invisible learning that's what I'm thinking it's going to appear more. How about yourself Ray?
Yeah it's interesting that personalizing learning thing because my goodness the uh world has been talking about personalizing ing learning for too long. But often they seem to rely on the teacher to be able to do that. Um I I'm not sure if if delivering an individual class to 30 student sorry delivering an individual lesson to 30 students in a class is genuinely personalizing learning and genuinely achievable. So the examples we've seen are where you use exactly as you talked about AI to augment things. So that could be about making sure the students have the right task in front of them um and that that's chosen for you the bit the bit that I've always imagined Dan is that one day I'll walk in to a classroom as a teacher and I'll say what should I teach today and that the system has been able to look at the work that students are doing in real time and be able to go well you got a choice you know there's a a concept that this group of students don't understand so you could teach this concept to this group of students based on what I saw them doing in their homework or their assessments or assignment last night or it could be actually you know I feel that gosh I wish I had the robot voice at this point I feel that everybody's doing well but communication we've not spent much time looking at that in the last few weeks and maybe an activity related to communication and this particular topic and that's where that bit about helping rather than replacing
and then that comes into the teaching part which is part B so you've kind of covered that a little bit how it's going to help teachers you know as a tool I think think teachers are going to be able to pull data out when we're looking at AI around give them more predictions. A bit like I think in episode one we talked about judges in in the US being able to you know there was a scandalous headline about AI sending prisoners or highlighting areas where prisoners had to go go to prison and whatnot and sentencing AI and things like that. Whereas actually using some of that data when you looked into the headlines the judges are using some of that data in lots of cases rather than prof file in but actually to help them base judgments and and similarly from a teaching point of view I think teachers will have more value judgments given to them um which they can then utilize to kind of predict things in into the future. So I think for teaching it's going to be about inclusive learning the tools say we got in Microsoft sack like immersive reader and the maths tools. So we've seen a lot of tools already in there and I think to be perfectly honest and it's a bit of a out there question but I reckon it's going to kill off standardized testing. I think AI will be the final sword to support teachers because we'll have enough data out there. Why do we need to test at the end of the year if we testing all the way through six 12 years of education? And I think that's what AI will be able to be more and more accurate and give us a much more reflective report card on uh Rey when he finishes school.
Here's my thought about the teaching pieces. Definitely there's a culture question as well, isn't there? Is that we've got Think about the way that we adopt the right culture to supporting this that if all we are doing is forcing teachers is you know with the best will in the world giving them a tool and saying you must use this then it's going to be seen as oppressive. So how do we help them to adopt it? I'm a big believer in selling people on ideas rather than telling people especially stuff like this can that can be quite disruptive. It's about thinking and let's let's be honest it People aren't great at doing this, but thinking about how we influence other people to adopt things. It's more than just technology. It's psychology as well. How do we make sure it's a genuinely genuinely useful assistant rather than becoming yet another thing that has to be done? Now, I'm going to go on to my favorite hobby horse, though, because the part of the question was school operations,
you know. So, I I have a health fund. I'm sure you do. And uh yesterday, I had to claim for a a a massage for one of my kids. and I just had the receipt and I went into the app and before I'd have to fill out all the item numbers and everything. Now I take a photo of the receipt and hit yes. That's it. They use AI to read the receipt and then to put that into the system. I don't have to do anything. And I think there's a massive opportunity for that in education. The amount of paperwork we see everywhere. It's been really interesting. Dan, I don't know if everybody's aware. Everybody in Microsoft has been told work from home. You know, 50,000 people in America, couple of thousand people in Australia. And so what that meant was that I walked home with my laptop back as I do every night. Nothing else, nothing else for the next two months because I've got everything digitally. And I know that's not possible in most education institutions. So, you know, how do we use AI to help us do things? You know, we we can use it to help process information, but we can also help to use it to help us find information
and and they did they did talk earlier on in the in towards the back end of the last conversation with Kate and the team there, you know, was some great stuff around culture change and that is tricky, but we are getting there with culture change with the use of data and and teachers starting to think about the fact that it's the data that's pointing to Ray's progress in the classroom doesn't necessarily reflect on uh it does reflect a little on Rey obviously as a teacher, but So Dan as a teacher for example, but but Rey obviously is is in charge of his own learning and the data that Rey is kind of attaining is not necessarily directly correlating. You know, I don't know what I'm trying to say here, but it's the way that the teachers kind of are not responsible for the progress of the data. They are responsible for the actual teaching to drive that progress, but actually ultimately it's the student data that allows them to be able to adjust the teaching to be able to fix that. Uh so the culture change is there and come in.
So let me ask you a question Dan and we'll um
is this question three Ray?
It it's not question three Dan it's a question that's coming in on the live chat as well which is sometimes a lot of this technology change causes stress and attrition for staff and I've certainly seen that as you're changing things. So how do you cope with that bit? You know what what do you do as you're driving some of these AI projects that are going to be amazing and change the world to help people to cope with them.
Yeah, that's that's a million-dollar question again, isn't it? But
that's why I asked you instead of letting you ask me.
But changes changes out there in in many ways in schools. They're going through this last couple of weeks is one of the biggest changes that everybody's had to do. Everybody is preparing for remote working. Like in the UK, schools are closed for the foreseeable future. I'm sure the Latin over here at some point coming up as well. So, it's about us managing that change effectively and obviously that's the senior leaders in the room and the people listening to the podcast and that's why thinking about you know once once teachers understand about how this is going to impact on the actual students progress then change is a lot easier you know it's a lot easier when you use Microsoft teams like we are today and we changing backgrounds and doing all this kind of stuff and AI is is once you use that in in the classroom and and start to do remote learning then you get used to it and you realize that it's not really rocket rocket science. So change management is hard. People have written books about it. There's millions and millions of dollars worth of consultancy fees being spent on change management. But really in a school it's very much around showing people the the value of what they're doing and then being able to move forward with that. Um we support schools generally with pedagogy from a Microsoft point of view as do other companies. But you know I think I think culture change when we look look at back as 's podcast and I'd encourage everybody listening to this one. Zani Vanwick was the um data scientist at C Catholic Dascese of Maitelland Newcastle. She had some great strategies on how she went into this and it was all about outcomes. It was all about we want to improve boys attainment in a particular year group. We want to improve well-being of students across the board for example. So once we know those goals then everybody's bought into those goals. Once we're aware of those goals and the transformation of objectives then people uh are more likely to adopt that technology. How about yourself Ray?
Well, let me ask you the question Dan. So um so in the February briefing from ISQs there was a a piece about research on artificial intelligence in education and there was a a very strong statement and I'm going to read the statement to you because I want you to take it apart and talk about it. Schools need to provide the tools, the time and the training first and then as they develop the strategies they can plan for the use of AI. So that can unlock opportunities to automate lots of tasks like marking, like freeing up teacher time, to focus on teaching and learning, all those things that we talk about all the time. And teachers could incorporate intelligent tutoring systems, you know, and bring those into lessons so they can start breaking off small groups into of targeted students and providing personalized learning for those students for yeah, every part of the lesson. Just as we were talking about earlier, the research said AI provides opportunities for teachers to gain deeper into insights into what is working well in teaching and identify different ways students approach problems in their working. Now, the good news is Dan, we didn't write that. We're not responsible for what we said there, but the question is, have you seen this working well in Australian schools? And what does it take to make AI a success schoolwide?
Yeah. Well, I I I've seen several different examples of this. You know, I I concentrate at the minute on Catholic school systems across Australia. Um, the way we've worked at Microsoft this year And I've seen a lot happening in Catholic Education Office in Brisbane. Ken and the team there doing some fantastic work. We'll get him onto a podcast soon. I've been working with him in the wings about that. Um they're doing great work bringing together some of the data, the actual academic data and doing predictions into the future and kind of highlighting risks and things like that in in learning. Sydney Catholic schools are creating some amazing dashboards. Now when we spoke at Zani earlier on, we talked about um and Ziv was from Catholic Dascese in Mton Newcastle. She's probably got the best holistic approach um to this in terms of the strategy that I've seen in any institution whether it be school, corporate or um university um myself. You know, Ray, you've probably seen a lot more in your university context, but but I've seen a lot of this great work and a lot of good dashboards and um strategies being set out. Obviously, you're right on the comments right about tools and time and and I think often when we come and talk about education time always comes up and that's very true um however you know you do need to make the time to have these impacts at this large scale so yes putting the tools in place these tools though however what I would spin that uh on I said you know you did the Azure machine learning project you know there are people who employed as data scientists in school systems I'd kind of say that the de democratization of these tools and the ease of use of these tools and the fact that anybody can go out and do machine learning using tools like Excel and Azure ML um without needing a degree in data science uh you know means that it does help. So I think yes the tools and the time absolutely and we we in education we're educators ourselves and I always value education over anything else but I would say that there are lots of people cess doing a lot of work in New South Wales department for example with machine learning um and even kids are using it predict breast cancer and things and survival rates on the Titanic using tools like Excel. So there are lots of departments doing this. There's lots of individual schools doing projects as well. I remember my my favorite example uh it was about two years ago I had a call from a school in uh Melbourne weekend and they said Dan we just worked out the correlation between our uh Napan results HSC results and English teaching and it correlated casual teaching uh frequency against naplan performance uh HSSE performance and again recently I was working with Matt Robinson down in Victoria and he did some fantastic work on the analysis of casual staff and how many students got affected by casual teachers across a large high school. So we've seen some great examples and people using different ways. The strategy is different in different schools. But when you think about unis ray where they you know hundreds of thousands of students what kind of examples have we got from that area?
Uh look it's easier in unis because they've got a big data set to start with because you know a uni's got 50,000 students. They might have 50 data interactions with them during the week. I'll tell you what is really interesting is that I don't think universities are using the data in the same way that schools are. Schools are thinking more about things from a datacentric approach whereas universities because they have so much data they don't necessarily know how to bring it all together. So, it tends to be focused in particular areas like student attrition or student support with the exception of the brightly burning stars like Dr. David Calaman. You know, we interviewed him on the podcast a while ago. He's doing amazing things at UNSW, but it's not institutionwide. And I think this question was around institutionwide. Hey, now look, Sandy uh made a point in the in the chat about, you know, agree that we don't need standardized assessments as set points in time. which as I think we all agree on and technology frees us up but we do actually need the quality of the information that we get from the standardized tests. So not all data is actually meaningful. So for example teacher devised tests well they might be meaningful within a particular class but they don't have the rigor and the assessment of information that we get from for example nap plan when you can compare across schools and I guess my point to that is yeah you're absolutely right I I wear a Fitbit and it tells me every day how many steps I've done. do not believe that it is accurate. Um, but I can look at the trend and I use the trend to inform my um, thinking. You know, whether I did a,000 steps or 10,000 steps, that's kind of the trend. Um, and if I have a great day and I go for a run and I get 20,000 steps, I know I've had a good day. I look at that over time at a a bulk level. I'm sure somebody else somewhere is using that data at a um very wide level and then maybe that's the opportunity in education. But I think it is always about direction of travel. Often with our students, it's about what does that glide path look like as they go from hopefully from here to here. What does that glide path look like and how can we increase that glide path rather than an a single absolute point in time because you know my daughter went to do her nap plan with a cold and therefore she probably got 10 points less. Or she went and I'd given her a big bag of red frogs on the way up to school and She was so hyped up that she got 10 more points. You know, this stuff is subjective, but not when we get to large populations. Okay, Dan. Uh, we need to finish. I'm going to finish with uh an awkward question. I've got one for you. So, Dan,
yes.
When we're thinking about AI, is it hope or height? Can AI replace teachers? Go.
Um, no. AI can't replace teachers. However, they can make teachers more effective by removing a lot of the administrative burden and allowing them to be able to make more effective decisions that allow teachers to modify their teaching and be more personalized in their uh use of their teaching tools and pedagogy in the classroom. So
have another go. Your first answer should be a sentence. Your your second sentence shouldn't exist.
Yeah.
Tell us what you think, Dan.
Yeah. No, no, it can't replace teachers. Of course it can't replace teachers, but it can be supportive to teaching and learning. How about yourself, Ray? What do you think about that?
Look, we had to get the bleet machine out on the last podcast when I said that I thought it was codswallop that teachers could be replaced by computers. You know, I know that uh Dr. Richard Swin from Wellington College in the UK was, you know, he's he's he's very PR savvy. He said uh teachers can be replaced by robots by 2027. I think that just fundamentally misunderstands two things. One, the role of the teacher and the second is it fundamentally misunderstands the relationship and how learning happens. So I absolutely agree a technology is about empowering other people it's not about replacing them.
So my question for you then awkward question of the day.
Oh thank you
with the advances in machine learning um and AI do we really need standardized testing?
We need to know how
sorry your question.
Yeah I look we need to know how students are progressing. Um because if we don't know how they're progressing, we we aren't able to provide the right support along the journey. The benefit of standard it's let me answer this entirely from an artificial intelligence datacentric point of view. The benefit of standardized testing is that we get a data set that is large enough to be able to do something with. Uh the downside of standardized testing is that um it just happens to be a result in time. Wouldn't it be great if you know we can get a student to take the nap plan test when they're ready? Wouldn't it be great if the NAP plan test was actually personalized to a student? Wouldn't it be great if a standardized test was something that you could take four times? You know, like kids do with games. They keep going until they do the boss level. And so they eventually get 100% in their test in a game. Well, why on earth don't we allow that for students in education exams? Why not? Why not? I I mean, apart from the huge cost of marking it all by hand, I don't understand why we don't have that. So, yeah. Uh We need some form of data that we can comp compare across schools. I absolutely agree with Sami and standardized tests give us a way to do that. That doesn't mean I support standardized tests. What about you, Dan?
I I think it it will do well. Let's be bold about this. You know, it does need to be changed. Everybody globally standardized testing. Maybe this might be an opportunity. I can't see this current this pandemic that's currently happening. I can't see anybody fairly doing standardized testing after schools break. wake up early depending on how long this thing happens. It's definitely been pulled in the UK. Will it be pulled in this country this year? Possibly. Will we see the end to it? I hope so. But uh you know there is a point in time where you do need to benchmark people um and make sure that you can get applications to university sorted. So I think that the the system doesn't help itself. If I want to get to university, how do I prove that I'm academically able? Are we going to trust teachers more and say, "Hey, Ray is actually bloody good at maths. You need to put him in your university course. How can we make a more cohesive education system, but I think AI, yes, will um I think it will put standardized tests into the boot because it it's just crazy the way that we do it now. And if you've got so much data on kids, you know, a teacher knows exactly how students going to perform. So, I think it will put the boot to it. Being brave and bold, let's do
and Rosanna Hanara in the in the group chat said, you know, one of the games changes for them was realizing that the teacher isn't the conductor in their own little world, you know, in their own classroom without external things. But if you hand control back of learning back to the students, you allow them to see that they own their learning journey, they can make choices as to what and how they want to learn with learning goals, learning styles, learning space, all that kind of stuff. And and you know, that's how we learn as adults. Don't try and tell me what courses I need to do. Let me define the journey that I need to do and find within that. You know, me, Dan, a big learner. I take lots of courses, but I hate it when I'm mandated to go and take a course rather than being able to take the course that I know is right for my next stage of journeying. So, look, if this great period of experimented innovation with online learning discovers that standardized testing isn't needed and the sky doesn't fall, then that's the world we should be preparing for where students are more in control of the journey and the role of the teacher becomes more critical to supporting students through that. But that's a different world to the world we're playing in. So, it's going to be really exciting what we see at the end of this.
It sure is. It's been great today. Ray, I wonder just if there's any other questions from the floor. Um, I know, you know, this is myself and Ray. You know, we've been doing this AI and education podcast to try to highlight and and bring some of these questions to to the four for school leadership and for business decision makers. You know, we we don't know the answer to everything, but um uh you know, the people in the room are the people going to implement it and know often um uh a lot more than we do. So that's why we've been going around and interviewing people to find out what people are thinking at the minute. So hopefully you find this kind of useful today. Um really appreciate all the interactions on chat and things. So and I appreciate um today you know Kate running this session via um teams and doing this remotely because you know it would be quite easy to uh cancel this in the current climate but I think it's a really a great thing you're doing in Queensland. We'd love to support you from Microsoft point of view in in whatever way we can. So, thanks for having us today for the podcast.
Yeah. And Dan, this has been fearless experimentation.
Yes.
Fearless experimentation. We didn't know if this going to work for the podcast in a room live with people even more so online. So, well done for making this happen. And I guess we've just demonstrated alto together that a step into the unknown. What's the worst thing that could happen? Well, the worst thing is you can get Dan and Ray blathering on at you for a bit. So, thanks for putting Goodbye everybody.