Oct 2, 2019

In this week's episode, continuing our journey on exploring Artificial Intelligence in Education, we talk about the common AI services. We talk about chatbots (or 'conversational agents'), predictive analytics, and making sense of student and campus data . 

We also discuss how Artificial Intelligence services are getting easier to use every day, and becoming part of every day life - like expecting to talk to our phone and it can understand us. We also attempt some long words - like pedagogy and data democratisation - and how it relates to artifical intelligence. And we disagree with some of the thinking of Silicon Valley gurus on the use of AI in education, and whether there's a risk of turning students into data objects.

TRANSCRIPT FOR The AI in Education Podcast
Series: 1
Episode: 2

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

This podcast excerpt features Dan, a Microsoft specialist in K-12 education, and Ray Fleming, the higher education lead for Microsoft in Australia, discussing the application and impact of Artificial Intelligence (AI) in the education sector. They initially explore the potential of chatbots for educational scenarios, such as creating personalized reminder bots for students, before shifting to the profound impact of data democratization and the need for schools to better integrate their traditionally siloed information. A key theme emerges with the idea of "data as the new oil," suggesting that education is at a tipping point where leveraging vast amounts of student data is essential for achieving true personalisation of learning and meeting rising consumer expectations from students and parents. Finally, they touch upon the use of AI in smart buildings to optimise campus efficiency and student well-being, highlighting that AI offers cost-effective solutions to long-standing educational challenges like retention and timely intervention.

Okay. Hi Dan. Good to see you again.
Good to So we introduced ourselves last week for our uh podcast around AI, but maybe just introduce yourself again.
Yeah. So I'm Dan. I work for Microsoft. I'm I specialize in schools, so K to 12. And I used to be a school teacher for about 10 years. also a school inspector in the UK.
And I'm Ray Fleming. I'm the higher education lead for uh Microsoft in Australia. So I'm interested in this stuff because AI really fascinates me because I've had a background in education technology unlike you, not a background in education itself,
but still very valid.
Oh, thank you, Dan.
So let's start this one then, Ray, by thinking about bots.
Okay, because last week we talked a lot about AI and the basics of AI. We talked a little about the ethics and things like that, but we didn't really dive into the education scenario. So, so let me ask you Dan.
Okay.
If you were building a a chatbot using a bit of AI to build a bot for for some education scenario, right?
What would you love Dan?
Oh, that's that's a good question, Ray. Um, probably something I was thinking of my own kids. Probably a reminder bot for them. Something to not an alarm, something that would pester them a little bit more. So, something that'll actually give them a big nudge, you know, on and on in the morning. in about the clothes and what they've got coming up and what kind of clothes they need for sport and things like that. A reminder bot would be awesome. Somebody they can communicate with.
I tell you what, I actually I I need that in my life. I've needed that in my life ever since I was at school actually. And there's two specific scenarios I need it for. The first one is can I have an end of term bot that reminds your children to take the banana out the bottom of their rucks sack that you don't find until the start of term. Definitely that one. And the other one I need is I need a bot that tells you when the start of term is. Because when I was a kid, which is a long long time ago, Dan, as you know, I used to go to school with my parents. So, they would drop me off on the way to work. And one year they took me to school a day early.
Um, so they literally dropped me off and drove away and then I walked into the playground and I was the only child there. I was a day early. So, I need something for the last day of term and the first day of term. If you can build me a bot for that, you'll be I will be so grateful.
It's interesting though because when we look at the education element, like you said, you know, I think some of the the the two big uh drivers for AI and these technologies in education is the fact that the tools are becoming democratized now. So when you said could I build that bot for you? You know I know I did a demo in editech it only takes an hour
if that to build a bot you know to really get going with something quite quickly. So I think when we're looking at why why all of these um AI tools and technologies are kind of becoming more interesting in to the education sector a lot of it is to do with that dem democratization of the tools.
I think we've learned from week one, we shouldn't go for words with too many syllables.
True.
And see the democratization of this one.
You're allowed to use education, but longer words than that, you're not. So, so, so tell me, give me some examples. So, why now? Why in education now?
Well, education's always been data rich. And I think we come to a tipping point now where people are starting to organize that data. Schools and school systems are starting to realize the importance of um bringing that data together. There's a big push around personalization especially in Australia about every child matters and actually the fact that we want to improve every student's outcomes in whatever setting whether that's about their well-being or whether it's about their um nap plan results or HSC results so it's really important that the uh we start to bring those things together and it just seems to be a perfect storm at the minute where people are starting to think about their data the state traditionally in schools we've had school information systems which a big bohemoth of something which will hold, you know, postcards and and all all the address information and the all the the kind of race information and gender information of the students. And then you've got a school learning platform which usually holds the learning data of the students. So, how they doing in English and literacy and numeracy and all the kind of subjects and never the twain could kind of meet and we've also got these other platforms around there that are that are holding data and it's starting to become uh and school systems are starting to think more about where their master data services are being held and what data is core for the education and the personalization of that learning. So, so there's a bit of a tipping point there.
So, so that's just about the not just but it's that volume of data thing because as you know from my background you know student management systems dear to my heart of course from the UK but it was always that recognition that there's a huge amount of data in there and when I talk to students today around the way that they interface with other organizations outside of educ There's a huge amount of data on them that's being used to apply to personalize things. And then I reflect on that way that's being used within education. We probably have much more data on individuals, but we're not using it in the same way that intelligence isn't being applied. And and probably the big thing is we're not seeing the hidden patterns in the data.
Yes.
Because the hidden patterns in the data are out there in the in the world of the websites that we use and things like that. You know, when I go to Amazon, They show me the things that they remember me looking at before. Yes. And it's a continuation of a journey.
Yeah.
Whereas those hidden patterns, what you're saying is they're they're locked inside the different systems.
Yeah. Yeah. Absolutely. And I think we get into a point where that where we start to unlock those and some of the technologies, you know, without sounding too geeky about it, but where a lot of the platforms now are are allowing us to get access to our data through APIs to be able to actually interface with our application and pull data out of people's systems.
Sorry, Dan, I got to stop you because you said without getting too geeky and then you said APIs. What are you talking about?
Well, it's the way we can interface with a lot of the um tools and technologies. So, for example, um uh companies developing um uh an application for a school can actually interface and pull data out of say Microsoft 365 um using a Microsoft um graph API to kind of pull out things like calendar in. And it might sound a little bit boring, but then for example correlating the calendar in and and at a third party being able to do one of our little bots like a reminder bot and be able to put that information back into you know school calendars and
so essentially applications have been able to talk to each other much more effectively between each other so allows that data to come out
so I'm just going to use my layman's
on the off the top of my head an example of an API so um I realized I I've been using one recently with an example I've been talking with universities around which is the English language testing certificate
right
so what currently happens is I So, so students have to have IELTS of six or seven to get into a university in Australia, overseas students. And so what the students do is they take a picture of it and send it to the university and somebody looks at it and types the stuff into the system.
Um, but with the vision services, you can actually get a computer to do that reading. So an API is the equivalent of sending it an email saying, "Here's the certificate." And it's sending you the data back and going, "Well, here's the reading result, here's the comprehension result, you know, here's the numbers. The API is the computer to computer equivalent of sending an email and getting the answer back.
Getting the information back. Yeah. And we when we're able to transfer that data between different applications, it makes it much more effective because like I said, schools have usually been very siloed in the tools and technologies they're using. So the way technology is moving, the way the data is, the way we bring that data together, and the fact that we can use tools to interrogate that data simply, you know, even things like Excel, you know, being able to do some amazing sentiment analysis. Now, you did a sentiment analysis.
Um, yeah. And so,
element in Excel, right?
Yeah, you're absolutely right. I mean, if I think about the complexity of this AI stuff, there's no way that two people like you and I would be sitting on a sofa talking about artificial intelligence 5 years ago. Yeah.
Because, you know, we don't have all of the skills that a data scientist has got. Um, but even more so now, yeah, it's the AI that's built into the products. It's the fact you can talk to your computer and it puts it into word for you. The fact that you can talk to your phone or as I found in Excel is you can do sentiment analysis. So if I the reason I kind of got on to this one is my goodness so many surveys answering so many questions.
Yeah.
And it's always difficult to deal with when somebody puts a comment in. And so I found that in Excel there's an addin that does sentiment analysis. So basically you take your student survey with 10,000 answers in it typed in individually by students and you can just get Excel to do a sentiment analysis and give every single answer a score from 0 to one about negative.
So that reads in the comment and it gives a gives a score essentially.
And so it's that way of taking a very complex issue
and boiling it down to something that me as a everyday user can use to do some analysis.
But but does that that also there has this inherent difficulties as well if you're coming up with a strategy if you're a business decision maker in a school for example or a dascese or university then I suppose the tipping point we're at as well with it's great to democratize the end users to go and find their own data and do their own analysis out but but what can we do centrally to make sure that a they managed without micromanaging it allow that power of the democratized end user go and interrogate the data but also have control of that data said last time around the analytics and things
yeah And that's me that was one of the reasons why I was really happy to sit down with you every couple weeks to try and do this because part of what we've um got to do as a society is better understand the potential,
the upside um and the opportunities of what AI could mean because we're going to be in situations where somebody like me can do an analysis of something but I need some basic skills to be able to use that data. So, You know, for me, I've gone on to do AI courses. I've played with my own data in order to understand things. But I think we need to be in that same boat because, you know, if I look back over my career, I have had a more successful career because I started off as a developer, but I've not done technical jobs for most of my career. I've done jobs involving technology, but I've been a marketing person, an advertising person, a PR person. I've done a whole load of different roles, but
I've done them better because I understood digital technologies and I think we're now in the situation ahead where everybody is going to be more and more impacted by the use of technology. So if you think about a principal in a school or a
leader that's dealing with data or you think about almost any role, those roles are going to become digital.
And so the skills are going to need to be there both to make the technology decisions about using technology to support something. You know the the I've got these 2,000 the answers, what do I do with them? But also the understanding how the technology works a little bit. And that's why it's important we talk about AI because we can't just apply it. We need to understand a bit behind it.
And it's interesting you mentioned that as well because the other thing is the expectations as well of just being able to do this, the expectations that the parents would have um or and individual students would have because they're paying in a lot of cases, you know, thousands of dollars to go to do a university. course, she'd expect to get that personalization of service cuz you're used to seeing it from brands like the iconic or or or your online shopping tools and technology. So, I think there's that expectation now that people want that information and pair and myself as a parent,
you know, I I still get the reports every term, but I do start to feel that we want more from the school. The school needs to be giving us more information than that termly report which says my son is is doing really well or struggling cuz you want to have that time the intervention to fix any issues that are going on.
Yeah, you're absolutely right that that change in expectations that's being driven by a change in a consumer experience is bleeding into the expectations of what education is going to deliver. You know, I had that experience where the school waited until the last report to tell me about a weakness with one of my children at a point I couldn't do anything about it. It's like, well, why didn't I know this earlier on?
And I suppose that My expectation was set in a more challenging way because of the experiences that I get. You know, I looked at it I like many Australian parents, I paid for my children to go to private school. And so I saw myself as a consumer of services and had consumer expectations on it. Um whereas, you know, I'm sure my parents didn't have that of me. In fact, I know my parents didn't have that of me.
They looked forward to the report they would get at the end of the school year that told them how I'd done in school. And unfortunately, those reports arrived way too too late for them to intervene
and and from a university point of view, I was just thinking as you're talking there about some of the key things that people looking from say from that sector where where they're really looking at kind of retention of students because every dollar counts, doesn't it? So, so a lot of that is built on actually personalization and knowing the students really well so you know if they are going to drop out because that has a massive effect on them over a 4-year university course I suppose.
So, I think where what we're getting to is that actually the problems we're talking about aren't new problems. you know, students dropping out of university, it's always been around 15 to 18% of students that drop out. It's always been about I always think about it, one of the ways I think about it is in financial terms about $3 billion worth of students leave every year from university early.
Um, that problem isn't new. The problem of personalizing learning isn't new.
The opportunity is new is to be able to use artificial intelligence to help with some of that. Yeah. Yeah, absolutely. And I think with those dem democratized tools um and the way that people are thinking about this and the way technology is moving, you know, it's also providing cost-effective solutions as well, which which we can never kind of dismiss because obviously lots of people when they're thinking about strategic decisions inside school systems and education generally, you know, cost is always a factor and you want to get that return of investment. So I think, you know, I know from my example I've used recently where um uh some customers that I'm working with, they've put bots in, um to solve help desk problems, and that's that's had a tangible benefit from a $40 a month outlay sitting on some services um uh in the back end being able to kind of do simple things like password changes, you know, where do I go to download the Microsoft suite, where do I go to download the Adobe software, simple questions like that, which the help desk is inundated with, they managed to save one of those persons time on the front line and they could do more second line support and really more in-depth um
so development.
There's a couple of types of costsaving there.
One is
the cost of the technology to be able to do that is
becoming lower and lower all the time.
Um and the second one is saving time which is a cost in itself. So
we know how overworked the teaching profession is.
So if you can save them time there's a personal cost there
to to being able ble to mark things or provide good feedback or whatever it might be. So, you know, both of those those sides and and I look at it at a very simplistic level and see for example a researcher in a university in social sciences for example. Um I was talking to one last week and they couldn't meet me on the Monday, the Tuesday because they were transcribing their interviews
and my immediate reaction to that is why are you transcribing an interview? Why don't you just get a service to do it that can do it in 10 minutes. Yeah,
but it might cost you $5 to get it done,
but it's going to save you two days of work. So, yeah, that kind of cost side that we've reduced the cost of the technology to be able to do it.
But I think there's a
we we had to there's an interesting example there because we had to use when I was working in a in a uni in Australia, we had we had several people employed to just transcribe video because video content in the learning platforms were becoming more and more prevalent and for accessibility. We had to do that transcription service and it was all manual. It was listening to headset, typing that in. So, it's really interesting.
And you were probably doing it, I imagine, for a specific group of students with specific needs.
No, generically, right?
Every video needed to be transcribed and that the AI wasn't there at that point. Every single video,
how fast is this stuff moving? Cuz, you know, that seems like almost eons ago and it was probably only 5 years ago.
Yes. Exactly. That's exactly right. So, so when we look in So, we know that, you know, education, you know, has got had the silo data. There's a almost like a perfect storm there at the minute in in all areas of education around the data, the availability of, you know, putting that data together, the democratization of the tools, the cost, the expectations that people are having. You know, this is almost like a perfect storm at the minute with with AI and the use of this technology around personalization. So, if we look at the three or three or four areas such as like bots or or data being that new oil almost paradigm and then obviously some of the other things that people are doing around smart buildings to save the environment and things. If we kind of drill down on some of those um you know we've kind of talked on bots a little bit already you know and that personalized element too. What are your thoughts on bots?
Okay well let's just quickly whiz across all three topics. We got to run out of time because I'm sure we can go into detail in each of them. So let's say that but but the the example around bots is
how do you personalize information down to user especially when you think about students these days that talking to a person is the last thing they want to do.
Yeah.
And and while that's you know in many ways sad and you know I absolutely think disagree with the Silicon Valley view that you know all we need to do is replace a teacher with technology.
Absolutely not. But there are ways in which students can get support or staff can get support from something like a a chatbot that allows them to personalize something down to them. So So if I think about the example of what's happening at UNSW with D uh professor David Kellerman in the the faculty of engineering um he's got a bot service he's built for his students where they can get help with assignments. So they get to a particular question in their assignment and they're not quite sure how to tackle the problem. What's the theory behind it?
You know they can take a photo and send that to the bot and the bot will go and look for references over in the lecture recordings and in the text that have been ass find and then come back and go, well, go and look here and here. It's going to provide you with some input. Um, you know, that's a really good way of being able to say, well, we can connect to this massive data because, of course, every lecture recording is data.
And did he did he was he the end user that actually developed that himself? Did he work with people?
Yeah. So, he was um as a as a lecturer, he has great support from from the university team around being able to help to develop that and get some of the technology pieces in place to be able to do it. I mean, it came from his vision.
Yeah.
Um, But but also it came from a very student centric view of the world because
he recognized that if you're a if you're a student at 11 o'clock at night and your deadline is 9:00 the next morning,
you still need help. You know, you might have started late, you might be in, you know, not fully equipped, but you still need help to be able to get there. So he's coming from a vision of how does he help every student to pass.
Yeah.
Personalization, right?
Yeah. Yeah. And so, you know, what he wants to do is just increase keep increasing the pass rate
and so identify the specific help and it's not all reactive. A lot of the work he's doing is about proactive help for students but also reactively to be able to connect them to more effective
and and and I suppose the bots and our personality thinking we're back to like the the theory around it all I suppose and when we looked in in episode one about you know what what our favorite bots were you know I suppose lots of the bots now are kind of bringing personality in there you know you can add witty elements to some of the bots but obviously you might not need a witty bot if you doing some medical diagnosis, you know, it's kind of a jovial bot saying, "Yeah, well, I think you've got some terminal disease, you actually want to start to kind of and and it was interesting because it was there was something I was listening to the the UK bringing the NHS um or the NHS archive in through a bot at the minute through Amazon Alexa services in in the UK." And um and that was quite interesting because the the the government in the UK worried about people just googling off their uh their ailments and things and they wanted to use the source of data, the source of truth, the source of their knowledge base. So, um, you know, I suppose that personality depending on the bot you want, you know, because people get very stressed out, don't they? People get stressed out when it's searching for university courses. You know, normally when you're asking somebody for help, you're in you're in a quite a stressful situation sometimes.
Yeah. And and often it's just knowing where to go. You know, some university websites have got 20 30. I've seen one with 100 FAQs.
Well, you almost need somebody to tell you which FAQ do I read to the answer to the question. Yeah, exactly.
So, yeah, bringing that all together.
Yep.
But you mentioned earlier the kind of data is the new oil thing which is that
economist headline for a few years ago um which was kind of saying that the business value in the future is going to be created through data and I think there was some work done earlier this year that said data is now worth more than oil. Um
but how do you relate that back to education? Because we've got tons of data
or almost too much data and and and I suppose that going back to my original point, there's so many third parties that operate in schools in silos that they sit on fantastic amounts of data. You know, every school has got access, for example, to a math athletics platform which has got all their maths learning over years. My kids have been using that for years and years and years. And you just look and think about, well,
wow, what data does that hold to support my kids through their maths education, you know, um in in high school, you know, what what can what what insights are in there which we've never even looked at. Um or or these other platforms you know the reading platforms you get lots of the online apps or lots of the online um uh you know apps the kids get on their iPads to follow through the literacy and things.
So I'm going to stretch the analogy a bit more around the data is the new oil thing idea which is there are the kind of data fields
of rich data that we already know are there the student information system the information that's in learning management system, you know, the school reports, although often we don't think of them as data because there's a lot of text, but text is just as important as the ones and zeros of data.
Um, then there's the oil fields, the data fields that we know are there, but we don't know how to use them. So, mathletics would be a great example. You know, we know that what is it, 6 million students take part in the Mathletics World Games.
So, you've got data on six million students and their progression through learning.
Um, you imagine if you could use that in order to deliver improvements and personalization and doing forecasting about what's going to happen next. And then I think you've got huge undiscovered assets in terms of data. You know, I think about um children I see being being um babysat by iPads watching Peppa Pig.
Yeah.
And I think I wonder what the data about children doing things like that is going to tell us. about their progress through school.
And so all of the all of that information, if you could join that together, well, you know, yeah, data is the new oil, but there's this whole education economy that isn't about the financial side of it. It's about how do you use that data to deliver a rich, deep, complex, and supportive learning experience.
Yeah. And it's interesting. I think what as you're talking there is what jumped into my mind was also about, you know, if we use a state, I think the if we go to the big kind of um uh issue in education with standardized testing. You know, I'm I really think that if we've got enough data and we're using our data more often, then you don't need standardized testing. Same with a car. You know, if you've got a car that's connected into a computer system that can tell the Toyota garage that you've got a a leak starting in in in the engine, then you don't need to put your car in yearly for a service and you know, all all these kind of things because you it's just time. So if we can be just in time without data around students, then all of the standardized testing, nap plan stuff, because ultimately the goal of that is to work out where the kids are, how they improve. But if you get better data, then you don't need to do that. But then you get the kind of um the the worry that we turn in kids into data objects themselves rather than concerned about their well-being. So there's an entire conversation around that.
Okay. So we can have a deeper conversation about that next time round, but let's find something less content. as well because the other area where I know people are talking about it is smart buildings and buildings don't have feelings. So maybe we should talk about that.
Let's do that. Yeah. So so you know interestingly you know we've seen some quite interesting campus designs for our own uh uh campus in Microsoft in Redmond there at the minute but um we've seen quite a lot of universities start thinking about this on for lots of different reasons and um and even that's starting to trickle down into the K12 space for smart buildings. What's your experience been so far around smart buildings? I once heard a a great case study of um a school system in Denmark that was using smart sensors in classrooms. Yeah. To be able to work out what was happening around learning because what they had was that they had something that they identified as a seasonal afternoon drop off. So what was happening is that certain subjects taught late in the afternoon
um the students weren't learning as effectively and progressing as far as they were with the same subjects taught in the morning. And they initially thought it was just you know the imagining the And when they were able to get out of the classroom, what they actually found was it was a winter problem which was that the buildings were so well sealed that the carbon dio
dioxide levels in the afternoon were so high that pupils couldn't concentrate. And so, you know, it's a really smart example, but a very micro example of how you might use data in a smart building scenario in order to be able to help improve
green plants in things like that. Some of the research that Stephen Hel did years ago as well about light and classrooms. We do the same in our own campus. We have sensors that we use to look at meeting space usage and and we use it for all kinds of things. You know, to understand what facilities people need, but also to do things like if there's a meeting booked in a room and nobody's been in that room for 10 minutes, you can pretty much assume that meeting has been cancelled or moved or something. So, you can cancel the room and release it for other people. We imagine if you use that in a timetabling kind of way around schools and universities or lecture spaces, classrooms, meeting spaces, all that kind of thing.
Yeah.
Um and then, you know, whenever I'm on campus, I'm always looking for paper flowing around and I'm also looking for cues because that's where you can get to smart campus ideas around, well, how can you optimize the use of campus? Anytime I see a queue of people, I think, well, what's the way to get rid of that queue? So, you know, if I see student services, I think about chat bots.
But if I see people queuing up for coffee and so some of the universities where everyone comes out of the train station and stops at the first coffee shop. It's like, well, if you've got a queue of 300 people at that coffee shop, but you've got three waiting at another one,
what do you do around the nudging on campus to move people? You know, how do you improve the experience for all of those students by nudging some to go and have a coffee there? It might be discounts or whatever it might be. And that's where you couldn't really do that with people. You can start using your smart sensors and things like that to design a better student experience.
Yeah. And I was in ities are such large campuses as well and some of the schools you've got where you've got multi- campuses you know to to monitor traffic around those places but also security you know for for um students in one particular site if there's a fire knowing who's on site and using those technologies to bit maybe um do registrations or whatever it may be
okay so I think what we've done is we've talked about AI and what it is
yeah
we've now dived into a bunch of really detailed education scenarios. Maybe next week what we should do is join those two things together to go from the theory of changing a coffee queue or whatever or a bot or whatever to what would you actually do and what are the considerations for doing things like that.
Fantastic. Let's do it.
Okay. See you next week, Dan. Thanks.
See you next week.