Nov 13, 2019
This week Dan and Ray talk about the use of our data, with artificial intelligence on top, to support our wellbeing. From talking about our own time management and how data and AI is adding itself into that bit, to discussing the use of facial recognition for student registration, we discuss the humanistic elements of the story. For example, the important role of the teacher and their interaction with students, or our preferred working styles. And somehow we even end up mentioning massage in Swedish schools.
We discuss the importance of staying focused on impacting the desired outcome - like keeping students engaged in school or university to graduation - rather than getting lost by just focusing on the technology and AI thinking.
TRANSCRIPT FOR The AI in Education Podcast
Series: 1
Episode: 8
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 explores the ethical and practical applications of Artificial Intelligence (AI) in education and professional wellbeing, contrasting its potential for positive intervention with concerns about mission creep. The hosts discuss a concerning report about an AI tool predicting student mental health issues, where a think tank warned this technology could be misused to stream pupils and limit their educational potential. They then pivot to the positive use of AI in professional settings, detailing how a tool like Microsoft’s MyAnalytics provides personal data—such as focus hours and quiet days—to drive individual behavioral change and improve overall wellbeing. The conversation concludes by emphasising that the goal is not merely to build predictive algorithms, but to focus on business outcomes, such as reducing student dropout or improving engagement, by combining AI insights with essential human interaction and support.
Hey Dan, it's time for another one of our AI and education
podcast.
How are you?
I'm good, thank you. Say the loveliest day today. Did you know
that?
No,
it is
actually today.
I think it is today, but not not for anybody listening.
The day we're recording.
Yes.
But Dan, they're going to know this was recording. in 1972.
Okay, look, last time we talked about AI for evil.
Yes.
And I think we might have sent ourselves down into an evil scenario
in which I discovered that you were Dr. Evil. Um, what caught my
attention after talking about that last week was I saw a story on
Sky News out of the UK and and it said, you know, that literally
the story said one of the England's biggest academy chains is
testing people's mental health using an AI tool. which can predict
self harm, drug abuse, and eating disorders. Now,
wow,
that caught my eye. But what really caught my eye was the next
sentence. It said, "A leading technology think tank has called the
move concerning, saying Mission creep could mean the test is used
to stream pupils and limit their educational potential." And that's
where I thought it was really interesting because we're starting to
dive then into what is the use of some of the AI services but also
what is the consequence of some of those things
and we and we kind of did touch on that in the last episode in the
A for evil episode because there were lots of gray areas around
marketing and and where the use of those technologies were so it's
quite interesting on that
but that bit which is about using AI using data to be able to build
models of how somebody's feeling and how somebody's operating you
know I I know that I've told you before I've told this podcast
before I'm a Fitbit addict I you know wear a Fitbit and I look at
how Am I going to hit my target? Now, that is a physical activity
thing. But I know for me, the physical activity thing is me is
related to how I feel. You know, if I'm have a lazy week, I don't
feel so good as if I'm having an active week. So,
but but then on on that as well, because I'm a fit addict, too,
sort of. And um and but it does draw you in. So, the AI does give
you that data, but also draws you in. So, I see you do 100,000
steps a week or whatever, and I'm like, "Oh, I'm only on 70,000."
And I go, "Right, I need to get out and do that." So, this
challenge is in this. The AI drives you to to catch Ray as well.
Follow him on Fitbit, everybody. But um yeah,
just tell the listeners, close your eyes ears, Dan, but this week I
think I'm beating you. Last week I was behind you. I actually look
at those things because for me it's a self motivation thing.
Yeah.
You know, and so, you know, it's really interesting. We've got all
these things that measure that give us data back and it's about how
we use it.
Um I know that we have the my analytics thing that you look at.
Oh yeah. Yeah. That my mine came in actually. That's an
testimony.
So just tell me first of all what is it Dan?
So my my analytics is is inside the Microsoft 365 suite and
essentially uses AI to look at your habits across your email across
your one drive and documents and what what you're doing online when
you when have you been online um and it's personal to you. This
isn't shared with your manager or anything. It's just to give you
your own well-being checklist. So each each week I get an email and
while I'm talking bring yours up right we can have a game of top
trumps for this. Basically bring gives you an email each week.
tells me how how I've performed in terms of my own well-being. I
can set my own parameters. So, I can say when I work. So, we
flexibly work at Microsoft. So, I can say that my working hours
between say 8 and 4 or 9 and 5 or whatever it might be around my my
working hours. So, then it can then say, well, how long have I been
working and emailing outside those hours?
So, this is about giving you data back for you to be able to make
change. Yes. A bit like the Fitbit. I know we're connected on
Fitbit. That's why It doesn't give me much doesn't doesn't give me
any advice. Just stores my data and just kind of says you've done
this amount of steps. And sometimes there's elements of advice in
there. It'll kind of be very generic. You know, you should sleep
more, you should do this more, but it doesn't really correlate
things like you'd have in a productivity suit like Office 365. So,
where it correlates your emails and your work and your
collaboration and things like that.
Okay. So, I I recognize that, you know, in my working week, I spend
my life in email and my Outlook calendar. It's either meetings or
emails. So, I suppose it's that same thing in the same way that if
it was education, you'd be looking at a child's use of a virtual
learning environment or a student's use of a campus for example.
So, my analytics, I've got it up, Dan. What do I need to look
at?
Okay. So, have a look at your focus hours then, Ray. So, what
percentage of focus have you done last week?
So, this is top trump. So, I'm looking for a big number or a small
number.
The bigger number is better. I suppose you're focusing in,
right?
Okay. So, appar According to my week, 72% of my week is spent in
focus time. Oh,
I'm only 51% focused in a week. Is that good or bad? Half of my
week I'm focused.
That doesn't surprise me because I'm a bit more introverted than
you are. So, okay. You know, I get more of my energy from inside.
So, focus time is something I value. And I guess I work in a
smaller team than you do.
So, there's less people around that. So, I'm kind of regularly out
and about and and and catching up with different K12.
Okay. So, that's one number. The other one I see is well-being. So,
what does that what does that mean?
So, wellbeing, let me just try to look for mine. Um, where's my
well-being number? I can't find it.
Well, well, I've got one thing that says I had nine quiet days.
Oh, yeah. Quiet days. Oh, yeah. Yeah. So, what are days? Yeah. So,
quiet days are when you've actually done nothing with your uh
emails or your collaboration. Actually, literally
Oh, so you've had a real life rather than IRL. Yeah. So, how many
how many quiet days have you had Nine
over what period?
I think it's over a month. But the interesting thing is I thought
when I looked at that just a second ago, I thought that's great.
Nine. That means, you know, that's that's every weekend I wasn't
checking in on my email and things like that. And then I realized
seven of those days I was on holiday. So, thank goodness I had
quiet time.
I did.
So, that kind of means I'm slightly addicted to my email and
things.
Yeah. I had zero quiet days, unfortunately. Zero quiet days. That's
that's horrendous. That means I've been checking my email.
You need an intervention, Dan.
I do. Yeah, absolutely. So, what about collaborations? That's the
other big number I saw.
So, collaboration, it tells you how many active collaborators you
got. I had 187 active collaborators.
Oh, I had I had 60.
60.
And a quarter of my time is spent in collaboration, which I'm guess
is in in meetings and, you know, working visibly in my diary with
other people.
Yeah. I 35% of my uh week was in meetings apparently. So, 16 hours
a week it told me this week. And the interesting one, 55% of those
were recurring meetings. So, it makes you think you know some of
these meetings are recurring and they're they're often you know
every week every month
so sometimes we creatures of habit we are creatures habit don't we
and we put things in and it does let make you think do we need
recurring meetings do we always have to have a stand up on
a Wednesday afternoon do we always have to have
specific meetings you know forecasting or whatever it might be or
or teachers and schools have meetings every Thursday night you know
some schools are starting to break that mold a little bit and try
to do different things similarly with larger teaching uh folks but
in an enterprise way I think it really does allow me to to have a
look especially the time out of hours is is key for me that's the
that's the measure I look at when I get the email on a Friday to
say well how much have I because it it's an impact on my family
time really it starts to get quite personal so it is AI looking at
what I'm doing because I think we are wedded to our emails
currently you know we use the things like Microsoft Teams and stuff
to to kind of move from email to chat which makes things a little
bit more or less persistent in terms of your email but you know at
the end of the day I think it's really important to to to have that
full well-being element to us in in bringing to work I know about
yourself
well so I'm I'm just thinking about how that how that data gets
hooked together because in a way it's a bit of analysis but where
the AI bit becomes useful is I know that that data then gets
aggregated and anonymized
so at an organizational level You can look at a departmental level
saying is that department collaborating well with other people is
that department there using getting the quiet time for example. So
being able to use it at an organizational level to be able to lead
to organizational improvement. And I guess that's a big thing for
me is it's great having the data but what do you do? What
difference do you make with it? And how do you use AI to plan the
right interventions? So
well you just mentioned that cuz because I'm just looking at my my
my my analytics on my my analytics or um but um it it basically
says here start to create a focus plan. So it actually gives you
some ideas of what to do. So it'll actually use AI to set up some
quiet time for you. So there's a little tool that you know it'll
say do you know it's told me I've had zero quiet days. Let's book
some time in now. And it'll do it for me and book that into my
calendar um by by clicking one with the button. And and then it
also says one that's just appeared now. Create a focus plan. So it
says Uh, my analytics can keep you um can help you make time for
distractionfree deep work every day because research shows that it
can take 23 minutes to refocus after checking just one email or
instant message. So, I think it's important to reflect and improve
on what we're doing just rather than just telling you nothing. And
this isn't a sales pitch for my analytics here, but it's talking
about how AI is used really productively.
Okay, so I get it. Now, let's go back to where we came into the
conversation at the beginning of this podcast, which was the
reporting about mental health and well-being of students and using
AI to do that because we've jumped into a particular scenario but
actually there's a parallel isn't there.
Yeah. Yeah. Absolutely. Because I think when when we're looking at
students mental health there's a big push at the minute to to
really support students holistically and in schools I think the
people listening to this podcast like business decision makers and
IT managers would be looking at utilizing the tools to actually
make a more healthy work force, you know, cuz email and some of the
technologies can be used ineffectively. You know, I don't think
we're the best at ever showing people how to use some of
technologies the best.
So, the media have latched on to the oh gosh, horror could be used
for this kind of thing. So, where's it being used in a positive
way?
Well, obviously in enterprise way when you look into my analytics
and pushing those out, but there's lots of school systems in in
Australia, for example, that I've been working with, you know, in
in the Catholic sector where they've got a big push around the
ethos and the kind of way that they work with the communities and
things like that. We've got some amazing projects which schools are
school systems are working on around well-being and kind of student
happiness in Catholic education in Western Australia. They're doing
a big project around that which ties together Office 365 products
and then also some other dascese like I'm working with one diocese
at the minute looking at applications using power apps to kind of
really drive uh students to get support. So students who might be
struggling, suicidal, maybe have bullying issues, then almost like
an application the students can go to to actually get that support
and and also bring some data together to actually identify those
students as well. So it's not just about identifying it's about
well what can they do immediately you know because sometimes those
students have those applications at hand all the time. So there's a
lot happening in K12 around well-being and and and focus.
I I remember a conversation with one of the state education
departments where their question was well can we use AI to help us
to predict incidences of bullying. The challenge for them was they
were collecting no information about what was happening already. So
they were, you know, not centralizing information. They weren't
really collecting the precursors. So it became very difficult to
say, well, you know, you can predict something in the future if you
don't have the data of the past because the data of the past helps
you predict the future. So that led them into a cycle of well, we
need to start collecting the data. We need to crawl before we can
walk, before we can run. It sounds like you working with some
customers that have got data and can you know
move forward.
Absolutely. And the other interesting thing as well is about
getting that data realizing what data you want to collect as well
because you know I remember from my inspection days when I was in
the UK you'd go into schools and they you have like a register of
incidents that happen like an incident register which you had to
check under safeguarding legislation at the time. You know it was
it was always interesting and we always got it drummed into us that
if you go to a school and they had one incident of bullying or go
to school has got a thousand incidents of bullying it doesn't mean
that the one has got a thousand incidents of bullying is worse at
bullying prevention than the one that's got one. It may be the
reporting mechanisms and the data collection mechanisms are out. So
I think it's a really good point because you do need to start
thinking about what data you are collecting. We mentioned this in
one of the other podcasts about data being the new oil because that
data is the key to have timely interventions around us and I think
it's correlation as well. So I was speaking to actually speaking to
a school last week in Melbourne and one of the uh FAB deputy heads
there was talking to me about project that they've done around
PowerBI and they've brought together some data and they around they
he started off to work out uh the not viability but the a project
that they started about casual staff and they would try to work out
how many casual staff they had in on that one particular day. Were
they being used properly? How did that correlate with the the
utilization and could they make it more effective and save some
money? But then they actually that kind of ended up the reports
that came out for that started to highlight some students had
casual teachers for 70% of the term because and they because of the
data was so disparate they wouldn't have been able to correlate
that because Rey would have a casual maths teacher one day and then
he might have a casual English teacher the next day and Dan might
be in your same class but be in different sets and you'd never you
never correlate that data but they they worked out that there was
um several students who actually had 70% or more casual teachers
and that has a massive imp act on their their performance. So um
and well-being, you know, catching things really quickly and and I
think you mentioned earlier on about facial recognition for say uh
when we talking offline about facial recognition for say
registrations. What are your thoughts on that around
uh wellbeing?
Yeah, I think it's interesting because you can automate some of the
things but often if we go back to our mission
Yeah.
the Microsoft mission is about how do we help people to achieve
I thought you talking about our mission in the podcast. Yeah. To
get three listeners.
Sorry. completely thrown. You know,
the the mission to help people to achieve more.
Yeah.
And that empowerment piece is really interesting because you you
could I've seen the stories that say, "Oh, let's just replace
taking the register with facial recognition." So when the child
walks into the classroom or the student walks into the lecture
hall, great, we know who they are, job done. But the really
important thing, especially in primary schools, but also in
secondary schools, is that interaction with the teacher first thing
in the morning, sets the framework for the rest of the day.
So don't just replace the technology and say that job is done.
Sorry. Don't replace the job with technology and say it's done.
Yes. Actually go how do we help the onetoone interaction between
the teacher and the students? How do we make that pastoral time
more effective? Great. We take away that bit where everyone has to
sit down and say yes to their name. But how do you replace the
interaction? Because if a student does doesn't have an interaction
first thing in the morning, it affects the rest of their day.
You know, and there is a lot of push around that in especially in
younger years. Well, and it'd be interesting to tease that out for
higher ed as well because I'm going to be really brutally honest
here like just based on like very and this is completely you know
there's no science behind this and I worked in a university context
as well in in Australia when I look at say some of the settings
where my kids are in kindergarten it's really personalized it's
really friendly and it's really kind of touchyfey and they hugging
each other and you know and even in you know in some school systems
around doing like Sweden and things they do massage at the
beginning of the day they do lots of things that are kind of really
good for emotional well-being and things like that. Um, and really
personalized. It does feel as if in in kindergarten, year one, year
two, you get a personalized relationship with your teacher. You
know, they got one teacher there usually and it's very connected
and the older the kids get, the more disconnected they get from
that to to a case where you end up in a lecture hall where there's
no um personalization in a lot of cases.
Yeah. And you have 500 people sitting in a lecture hall and you're
you could feel that or just a number.
So I guess that then comes back the conversation that I think we
need to have. This isn't just about a technology thing. This isn't
just about a oh we could do this thing to build this thing to do
the take the register or predict who's going to drop out or
whatever. It's the what do you actually do about that. It's about
the outcome you're trying to achieve. Not the technological input
or output you're trying to achieve but the outcome. So if I think
about student dropout one in five students don't graduate year 12.
One in five students drop out of university in the first year.
Building an algorithm predict to predict that isn't rocket science
anymore.
But the job isn't to build an algorithm. The job is to keep more
students and get them to graduate year 12 or keep more students to
get them through a university course rather than dropping out in
year one because we know their life chances are going to be better
under both of those circumstances. And I think sometimes we get a
little bit lost in a technology problem solving mindset.
Absolutely. Yeah.
And let's build a perfectist algorithm
that predicts who's going to drop out or who needs an intervention
or how somebody is going to do in a particular test rather than the
ultimate goal which is how do we keep them engaged in school in
university so that they graduate.
That's the real goal and it's about combining the beauty and the
art and the science of AI with with the humanistic approach to how
do we support those people?
Yeah.
If I think about student dropout in universities, you know, a
university product or service, the most difficult year is year
one.
So, how do you use the data in order to be able to help you improve
year one for that group of students that are really struggling? And
one one example, I read some research from an Australian
institution. I'll find it and put it in the show notes. There was
some research that said how you do on the first test. The first
assignment in a university course is the thing that makes you
understand or believe whether you're the kind of person that goes
to university. So if you're the kind of student that's going to
struggle
and we can predict that very early on with a bit of smart bit of AI
and then you do badly in your first assignment, then it emphasizes
the personal self-t talk which is university isn't for me and so it
leads you to drop out and that's assignment number one.
Yes.
Actually has an impact throughout year one.
Yeah.
And so let's predict those things. But then how do you make that
better? How do you pick out a certain group and go do you know what
this group of students need extra support for the first couple of
assignments which means more clarity, more level of understanding
the expectations because my daughter still comes home in year two
in university saying that she doesn't understand what is expected
at the end of an assignment.
And and I think that's really good because I think at the end of
the day It was so good that she doesn't understand it. But it's
good to to have that insight because I think if you look at where
Ken Robinson was going with the fact that in an industrial age, you
know, often when I speak to schools, they always do the when we
talk with digital transformation, they often will talk about we
can't move a timet, it's impossible. Some principles do try to push
boundaries and try to do some fun stuff um and try to innovate um
fun stuff in in quotes and innovate in the next one, but they try
to innovate and and think about how they could um manipulate things
to make it different. You know, I was in one school in in the UK
under the builder schools to the future program. I'm sure you
remember that, right? They basically sacked all their PE staff
because they had a they to address that business problem, they said
they wanted 100% students to participate in sport and the uh the
teacher the the head teacher at the time, principal said the
problem with PE teachers in the UK was that they were all come from
a sporting background and they all were cricketers, netball
players, hockey players, rugby players and they like their sport
and then if the kids liked their sport like soccer or whatever it
would be great but there was no malleability of the curriculum to
do something like archery or fencing or something like that. So you
got rid of everybody brought in professionals to do listen the
sports teachers were not professional he backtracks very quickly
there lots of my best friends are sports teachers but the um
they were friends
you can hear them all screaming in the in the in the in their cars
now. Yeah. drive safe everybody but uh at the end of the day I
think they ended up getting solving that problem not through
technology but just by changing the staff in so for the interesting
thing when I was listening to you speaking there was the business
problems and we've hit this a few times now the business problems
in schools and universities and education generally have always
been there so would your advice be to the people listening to this
podcast to go back look at the business problems maybe look at the
the strategy papers from the school uh systems that they're working
in rather than the IT project plans. You know, an IT project plan
typically that I see now is okay, we're moving everything to Azure,
we're moving to IAS, to SAS, to pass, all these lovely acronyms,
moving stuff into the cloud, and then trying to move to a
service-based orientation for an IT organization. Sometimes not
really looking at well, what does matter to this the business as a
whole, you know, literacy of boys or whatever it may be.
Exactly. Right. And I think I was listening into a podcast during
the week. I'll put it in the show notes around a data scientist,
you know, talking about the strategy of when things go wrong and
when things go wrong from a AI data science point of view. Often
the project has gone wrong because people have been trying to solve
a technical problem or a theoretical problem, not a business
problem. And so fixate on the business outcome, which might be
retaining more students or it might be more students passing an
assignment. It might be more students feeling more comfortable at
the institution.
Yeah.
Safer, happier, whatever it might be. Focus on those things
because the project is about delivering that outcome, not about
delivering a better algorithm or a better insight. It's about
making a difference to the humans at the other end of it. I kind of
believe that that's often where we kind of lose the plot a little
bit. If I think about a lot of conversations around student
recruitment, for example, you know, the goal is to get more
students in the door. If you're a university, and you're appealing
to international students. How do you get more university students
coming in? Not how do we fix some small part of the equation, but
how do we get more people to apply and start to study and then, you
know, to get to graduate at the other end? And it's that big
problem level that you then break down into parts rather than, oh,
we're doing a data science problem down here where we're just
trying to build an algorithm to predict something. Predicting
something doesn't change anything.
Yeah. And and I suppose using some of those tools like the
accessibility features which talked about a couple of podcasts ago
to generate uh subtitled videos in stream or whatever it may be
would then bring your international students together because
they'll feel more a part of a community be able to access the
materials. So as a knock on effect implementing things that are
already there without thinking about AI even in some cases.
So a couple of weeks ago we had Troy as a guest talking about how
AI is used for accessibility. I kind of feel like we should have
another guest spot coming up.
Oh that would be good.
And we should talk about how AI is used to help personalize
learning and help more students to succeed.
That would segue really nicely into this.
Well, I'm looking forward to that next week. Let's see who we can
find to join us.
Yeah, somebody.
Thanks, Dan.