Women WithAI: AI in Education: Helping Teachers, Not Replacing Them with Hannah Marr

Hannah Marr: 00:00:04

I knew a little bit about AI. Like a lot of people, I'd started just kind of doing some, some experimenting and kind of working out how it could help me in my role.

But I certainly wasn't any kind of expert, but I thought, well, it sounds like a super interesting role in a really interesting company. Let me just apply and see what happens. And somehow I got, I got the.

Joanna Shilton: 00:00:27

Hello and welcome to Women WithAI, the podcast dedicated to amplifying the voices and perspectives of women in the field of artificial intelligence.

My guest today is a passionate, solution focused publishing professional with senior level experience across a range of sectors and a particular focus on education. But before we jump into talking about how AI is creating opportunities in the education sector, let me tell you a little bit about her.

Hannah Marr leads on Generative AI at Cambridge University Press and Assessment, a world-leading academic publisher and assessment organisation that provides exams in more than 130 countries and is part of the University of Cambridge.

Hannah built her career in educational publishing, working initially on printed books and audio cassettes before focusing on digital resources and most recently AI.

In her role at Cambridge University Press and Assessment, Hannah is ensuring this new technology is implemented safely and ethically to support colleagues, TE researchers and students alike. Hannah Marr, welcome to Women with AI.

Hannah Marr: 00:01:27

Hi Jo, thanks for having me.

Joanna Shilton: 00:01:30

Oh, it's a pleasure to have you here.

I mean you've built a really impressive career in educational publishing from printed books, and as you said, cassettes, and I was just checking, you did mean audio.

Hannah Marr: 00:01:41

Cassettes date me terribly. But yes, my very first job was moving content from audio cassette to CD because that was the big new technology at that time.

Joanna Shilton: 00:01:52

Wow, brilliant. I must say I've just, I've just had to let my old Mini go and that had a cassette player in it and now I'm just, I'm like bereft.

I'm like, oh my God, what am I doing with all my cassettes? But anyway, off the topic there. But yeah, please, can you, can you tell us a bit about your journey into, into AI? What led you to focus on it?

Hannah Marr: 00:02:10

Yeah, so yes, like I say, I've been interested in, in publishing for a long time, all things digital for a long time and education for a long time.

And I've worked for lots of companies who I guess are competitors for Cambridge University Press and Assessment, and always been interested in them as a company, and I saw this role come up, and I knew a little bit about AI. Like a lot of people, I'd started just kind of doing some, some experimenting and Kind of working out how it could help me in my role.

But I certainly wasn't any kind of expert, but I thought, well, it sounds like a super interesting role in a really interesting company. Let me just apply and see what happens. And somehow I got, I got the role.

And yeah, I think for me it was really, you know, this is a really exciting new technology that I wanted to be a part of. I wanted to kind of get a bit, a bit ahead of learning about. I love learning in all of my roles.

You know, I don't want to do the same thing always all the time. I want to kind of continue to kind of learn and grow in, in my career. And this is definitely something that I've been able to, to do in this role.

Joanna Shilton: 00:03:13

Fantastic. I love the fact that your background isn't traditionally techie, you know, not at all.

And you kind of navigated AI or into AI or AI's role in education. So what, what sort of transferable skills would you say have helped with that? You know, people are looking to, to do the same.

Hannah Marr: 00:03:30

Yes. So definitely to, to reiterate, I am not a technical person at all.

And I guess in my sort of personal life, I'm probably not one of those early adopter people in terms of technology either. I don't still have cassettes in my car. But, you know, I'm not, I'm not far behind that.

So, yeah, I think for me a lot of people are in a similar space. We're still learning about this new technology. It is still early stages. So actually a lot of what is supporting people.

So there's a lot of kind of stakeholder management involved, a lot of change management.

So really what interests me, alongside the technology, the technology is fascinating to the level that I do understand it, but it's all the implications for people and how we work with this new technology.

So yeah, my role really is much more around kind of supporting colleagues and ultimately supporting our teachers and researchers and students as well with learning about this technology, getting confident in how to use it and how to get value from it.

And so, yeah, I've really been able to bring, I guess, those existing stakeholder management, partnering kind of skills that I developed in my previous career into this role. And in a way it helps that I'm still on that journey as well.

You know, I don't have all the answers, but definitely what I've been focusing on is not just what the technology can do, but where it really can add value.

What are the opportunities that it brings and where are the areas where actually, it's not a silver bullet and probably we don't need to be focusing on.

Joanna Shilton: 00:04:55

It yet because I guess in education quite often sounds bad. It's not always at the forefront of technology.

But I can, you know, I'm just thinking because I don't know a lot about AI and I'm learning with every single person I speak to and the more I look into it, but I think at the beginning or certainly a year or so ago, I was like, oh no, we can't let, can't let students use AI. And I guess people were scared of it. So that must be the kind of interesting bit because you're, as you say, you're learning as well.

How is it sort of creating opportunities that you're sort of making sure people aren't scared of?

Hannah Marr: 00:05:26

Yeah, I think that's it.

I think there's just such a spectrum of people who are either just not interested slash terrified, you know, AI is going to take my job, or yeah, specifically in the education space, AI is going to prevent students from learning or help students to cheat or whatever it is right the way through to people who are so enthusiastic and so excited and think that, you know, AI is going to change the world, change education as we know it's, and can't wait to kind of get out there and use it, which is exciting.

But also then I'm needing to kind of make sure that again, we're just kind of doing it in the right way and the safe way and everything else and not kind of getting too far ahead of ourselves. So, yeah, it's interesting working with people right the way across that spectrum.

And I think what's interesting, particularly in education about AI, like you say, typically or generalising massively, but quite often big new technologies aren't necessarily adopted in the education industry straight away or kind of early, but I think with AI, because it was chatgpt was, was a couple of years ago now and everybody was just able to experiment with that in their own time. Not necessarily if you were a teacher using it in school, but just using it, I don't know, to ask instead of Google or something like that.

So it kind of quite insidiously became part of people's awareness and we all know how kind of time poor teachers are and that that is such a barrier in, in education in terms of teachers just have so much to do, a lot of which is teaching, but a lot of which isn't.

And I think quite often teachers were just using ChatGPT quite hackily and going, oh, you know I can, I can use it to help me plan a lesson or I can use it, you know, just really sort of simple things like, I don't know, find out more about this topic that I'm going to be teaching. And I think that's what's maybe a bit different about this is you don't have to have at first loads of training to use it.

You just have to get in there and experiment. And not all teachers were doing that by any stretch of the imagination, but some really were and that they were sharing what they were finding.

And it's been quite organic that way. So, yeah, it's an interesting one. I think that's why it's maybe felt quite different up to now.

Joanna Shilton: 00:07:43

Do you have to check what people are using it for, or, you know, like getting them to fact check it, or is that kind of the guidance that you give?

Hannah Marr: 00:07:51

Definitely, so there's a, there's a term that you've probably heard on this podcast a lot around human in the loop.

At Cambridge, we adapt that to teacher in the loop. So the teachers are the experts. They know their subject, they know their children.

But AI is there, like any technology to be used as a tool, but it's the teacher's expertise that is ultimately the most valuable thing.

So, yes, if you're, if, if they're using it to create a lesson planner, we're saying make sure that the teacher is then also checking that, that lesson planner and checking that they're happy with it, checking it's the right level for their students, all of those kind of things. So I think AI is, is a, is a productivity tool. It kind of helps you do things more quickly when you've got loads of things to do at once.

But again, it's not, it shouldn't just be left on its own to go and teach students. That's definitely not, not what we're saying.

Joanna Shilton: 00:08:38

Okay. We're not going to get it anytime.

Hannah Marr: 00:08:40

Soon anyway, let's put it that way.

Joanna Shilton: 00:08:42

And it's that thing, because AI and education, you sort of automatically. I did before just think of the students.

But you're right, the teachers are just, you know, just as important and it's got to support them with what they're doing.

Hannah Marr: 00:08:52

And particularly at that sort of K-12 level.

Again, we're very much seeing is it as a tool to help teachers help students at the moment, rather than necessarily kind of going and supporting students straight away? Obviously, university students they're using it in different ways.

They're using it to help to help their learning directly and we support that as well. But yes, just again, it's such an exciting tool. It has so much potential. But let's make sure that we're putting all of the right guardrails in place.

I'm going to sound like a broken record on this because this is the conversation I have with people all the time. But also this is what we're hearing from our teachers and from our researchers and students.

Well, less so from students, but from our teachers as well, is please help us to use this technology in a way that we can be confident with it.

Joanna Shilton: 00:09:42

That's great because I was just speaking to someone in my office last week and they were saying they had all these slides to produce and they had to do this presentation and they were sort of drowning and I said, have you asked Chat GPT? Have you? And they were like, no, no, no, I don't, no, I don't. They won't understand. It won't know what I need to say. It won't understand the project.

It won't, it won't. It will just take me too long. No, but it's that kind of, at the start you do have to get your prompts right and know what you're saying.

And I guess also where in your role in the organisation you've probably got your own Chat GPT or your own, like AP or something, you know, so you're not just putting out, I won't.

Hannah Marr: 00:10:18

Say which one is, but we do.

Joanna Shilton: 00:10:19

Yes, but I think that's the way because again, that was something that I hadn't ever thought about at the beginning that, you know, you're happily putting everything into. If you've got the free version, I suppose you need to sort of keep it locked down so no one else.

Hannah Marr: 00:10:30

No, exactly. And like again, any sort of technology really.

I mean, if you think about, I don't know, you might keep your financial information in Excel, but you probably password protect that Excel. It's, you know, what are you putting in and using the right amount of sort of security and sensibleness based on the data you're putting in.

So, yeah, at Cambridge we have, we do use one of the big LLMs and we have our own kind of private instance of that for exactly that to make sure that it's secure. And then obviously if we were actually creating tools for teachers, etcetera, we put whole extra levels of security on top of that.

And I think what's really one of the interesting things about this technology as well is the more feedback you give it, the better it is.

So that's another key kind of principle that we have when we're, when we're designing AI tools as well, is that whoever your end user is, say it's a teacher again, that you're giving them that opportunity to provide feedback on what the tool is giving them so whether that's just.

Was this lesson plan useful, thumbs up, or whether it's, you know, it's output something that actually wasn't the right level and the teacher can literally kind of type in and say this was too high level or whatever because again, it's feeding back into the tool, it's making the tool better. And I think that's. Yeah, that's really, really exciting for me.

I think again, just as a principle, that's something we always try and do with our customers, is make sure that we've got that conversation with them, and this is another way that we get to do it.

Joanna Shilton: 00:11:51

And I think as well, like for you, for whoever's doing that as well and feeding back, I think that's really helpful because you're like, well, actually I am the expert and, you know, this is what you do. We're putting the data into AI so that it can get better and it can give us back what we need and it's learning from us, rather.

Yeah, because it is, I think that's the thing, it is a tool and that's, it's just how to sort of best use it. And that's why I think more places need people like you to be kind of pushing it forward. Will it also be used for exams?

For marking and that kind of thing? I can imagine it's going to save so much time if it's like a multiple choice.

Hannah Marr: 00:12:29

Definitely. So, I mean, I guess there's two or a few different versions of that.

One is actually again, helping teachers who we all know have loads of marking to do so, you know, giving them the tools to help with their marking. But then, yes, at Cambridge, as an assessment provider, we're also looking into all sorts of different ways that AI can support with exams.

Whether that's actually writing the questions or then. Yeah, whether that's marking with them.

Obviously, particularly in kind of high stakes exams, we need to be incredibly careful and make sure that everything is at least as accurate as a human would do it and hopefully more accurate.

So, yeah, that's going to be a sort of fairly slow burn in terms of making sure that we can build the right tools in the right way and not obviously disadvantage students in any way, shape or form.

But again, I think the Potential there is absolutely huge and it's not just helpful for us as an awarding organisation, but it means students will be able to get their results quicker. It will mean things like more personalised assessments will be possible.

So, yeah, there's all sorts of exciting opportunities I think that this technology opens up.

Joanna Shilton: 00:13:38

That's really. I'm just starting to think now in the future you might have an eye that would recognise the student as well.

Hannah Marr: 00:13:43

And be like, well, definitely down the line. Definitely. Yeah.

So I think, I mean, and I, again, I don't think this is in a couple of years time or anything like this, but potentially down the line, what learning looks like, certainly what assessment looks like, I feel could be very, very different in the, in the sort of, let's say, medium to long term future.

Joanna Shilton: 00:14:04

And also, and this is, this isn't anything against, you know, teachers or anyone that's marking anything, but, you know, everyone might have an inherent bias or something.

Hannah Marr: 00:14:11

Just an off day.

Joanna Shilton: 00:14:13

Like, yeah, if something really awful's happened or you're feeling really sick or you just need to get through it, you know, the AI will sort of take away from that. As long as you know that.

Hannah Marr: 00:14:22

Absolutely.

And again, when you think about kind of high stakes assessment, of course we have all sorts of protections in place to sort of prevent that happening already.

So, you know, there's all sorts of different levels of QAing that happen and we already use technology to kind of make sure that, you know, to spot any outliers and things like that.

But that's all so manual at the moment, you know, it's so labor intensive and actually, you know, if AI can help with any of that, again, always having that human in the loop still. But it's. Yeah, there's just, there's a lot it can help with, I think.

Joanna Shilton: 00:14:56

So what excites you most, do you think about that potential?

Hannah Marr: 00:14:59

I guess I think it is. I love still being able to kind of talk to our teachers and our researchers and hear about how they're starting to, to use it.

So not thinking that we as Cambridge need to come up with all the answers, but, you know, really understanding how people are starting to use it again, acknowledging its early days, because I think that that kind of gives you ideas and things that you would, you know, I'm not, I'm not a teacher, I'm not in the classroom every day, so it gives you kind of ideas that you wouldn't have thought of otherwise.

So I was at a conference a few weeks ago with a load of universities and it was just really interesting a hearing, just the general challenges that they have.

We all know there's lots of complexities in our university system at the moment, but again, how are they starting to think about how AI might be able to solve some of those? And then of course, we kind of take that away as Cambridge and go, okay, is there things we can do then to help with these problems?

So, yeah, it's not that I necessarily have a specific thing that I'm most excited about.

I think what's so interesting about this technology is it does have such a wide range of potential applications and that is everything from just, I suppose, quite simple in a way, sort of product of productivity things and then right the way through to does an assessment look completely different in 10 years time because of AI? You know, it's. Yeah, it's, it's moving so quickly. So, yeah, you ask me this again in six months and I'll have a different answer for you. It's.

It's just changing all the time.

Joanna Shilton: 00:16:25

Yeah, because you, I mean, just even the sort of cassettes to CDs, that's changed, hasn't it? I do still do lots of printing as well. Do people like having books?

Hannah Marr: 00:16:36

Absolutely, yeah. Yeah. No, printing's still a big part of what we do.

And again, where are the sort of opportunities to make that more efficient with AI as well at the moment, you know, we don't necessarily have robots that could do the printing for us.

Not saying that wouldn't be possible in future, but again, behind the scenes, there's so much that's still quite surprisingly manual in a lot of these processes. So. But yes, books are here to stay, I think it's fair to say.

Joanna Shilton: 00:17:00

So you sort of touched on like, sort of responsible implementation and kind of like you're looking at that kind of thing. Like, how do you do that? Is there a team of people sort of doing it in the background? And then do you. Who spots it like, how does all that work?

Hannah Marr: 00:17:13

Again, everything's still really early and I think, you know, lots of organisations are still working out kind of what the best ways are to do this. Interestingly, Rolls Royce have some really robust security documentation that I found the other day. So just shout out to them.

But how we, we've approached it is we've got what we call our Safe Use Panel, which is a group of various different stakeholders, technology and business stakeholders, and any new use case that anybody in the business comes up with, whether that's for something internal or whether that's something to support our customers, basically goes through this safe use process.

It's a two-stage process, initially just a sort of fairly light touch review of what tool they want to use, what data they're planning to put in the tool, to give the team approval or not. But usually we do with some guidelines to experiment with that tool, because that is such a key part still in this space.

We need to make sure that we're supporting people to be able to learn and experiment with this technology.

And then depending on the outcome from that experiment, if the team want to move ahead with actually kind of putting it into production, then there's a more in depth review that we do at that point which takes into account all sorts of things, the sustainability impacts for example, and just generally thinking about, you know, again, responsible use of this technology in all, in all the different ways that that kind of means. So yeah, that's how we approach at the moment.

It's all, we've got it kind of centralised in a tool so that we can kind of track things and all that good stuff.

But yeah, it's, it's a huge focus for us and I think it's, it's one of the areas that I'm most proud of at Cambridge that I think we were ahead of the curve to be honest, in, in setting this up and we're really kind of seeing the, the benefits from that.

Joanna Shilton: 00:19:00

Brilliant.

I love that you mentioned earlier, like one of the skills is sort of, well, just change management, and so that, because I guess you probably have still got people that are unsure about. So how do you handle that?

Hannah Marr: 00:19:13

So I guess in parallel we've got a big learning and development program that we're running, which is a mix of I guess quite introductory sort of support.

So again, just, you know, here's a productivity tool that you can use in your everyday work and kind of giving people guidance on like you say, what good prompting looks like, what the basics are of kind of safe use, all of those kind of things and then right the way through to some more specialised training for different Personas in the business, including our technology teams. So that's a huge one.

I think it's, you know, you're going to get resistance to things where people feel like they don't have the information or they're just not confident. You're always going to get people who are still like nope, this isn't for me.

You know, and interestingly, even some technology people who are like, nope, I like doing my coding how I've always done my coding. I don't want you to know, AI is coming and interfering with this. Less, less so probably in technology, but you do get them. And so, yeah, I think it's.

How do you bring people along the journey while also acknowledging again, we do not have all of the answers yet. This is not where we can go.

Okay, this is definitely where we're trying to get to in three years or whatever it is, and this is the exact journey we're going to take to get there.

So it's change management around the technology, but it's actually also supporting people to get comfortable with uncertainty and, and just change in and of itself, which again is something that I find really, really fascinating.

And it, you know, you have to think about in your, you know, in your life when you go through a change, what do you find difficult and what kind of helps you and, and it is having being confident that you're being told as much as, as possible, you're being given as much information as possible and that, yeah, you're all kind of working towards the same goal. I suppose ultimately no one wants anything.

Joanna Shilton: 00:21:06

Or change just sort of thrust upon them, do they?

Hannah Marr: 00:21:08

Exactly, exactly. Yeah. And everybody, you know, everybody's involved and everybody can have a say.

We are very definitely a very sort of collaborative organisation, which maybe means we're also not the speediest organisation always, but I would rather, I would rather go on that side than do things super quickly but leave people behind.

Joanna Shilton: 00:21:25

Definitely. Because you mentioned as well, keeping the teacher in the loop.

It's maintaining the human input, which I guess is really important in education because that's the thing, you know, of, you know, we're sort of AI is only as good as the data that has been given and everything that's fed into it. But things are changing and things do change. The world's changing, you know, everything is sort of is happening.

So you kind of have to feed all that in there. But you still need that humour because I guess the thing is it's prediction, isn't it? A lot of AI is, you know, is that sort of mathematical toy.

It's predicting what's going to happen next, but you can't always do that.

Hannah Marr: 00:22:02

Absolutely, absolutely.

Joanna Shilton: 00:22:04

Definitely reaching. So how do you.

Well, I suppose we sort of, we've talked about this or sort of thinking about it, how AI will shape the future of sort of publishing and learning and assessment. I don't know. Have you got kind of, I guess.

Hannah Marr: 00:22:21

I mean, you hear all the time our education system fundamentally was designed for the Industrial Revolution, you know, that. And you kind of hear people say that it's not Fit for purpose and everything else.

I don't know that I would go that far, but I do think we're probably overdue a bit of a shake up in terms of, yeah, just how we think about learning and what are we preparing students for ultimately?

And again, it's not that AI has all of the answers that we need to answer that question, but maybe AI is that opportunity given that by and large schools are kind of buying into this, you know, voluntarily. They are kind of genuinely excited about this. It's not something they're being told to be excited about.

Maybe this is that opportunity to kind of start to shake up education a bit more and really look, you know, if ultimately AI is going to be able to give us the answers to most of the questions we could ever want to ask it, well, what does that mean for learning then?

It's not probably, and this is, this is my opinion, this is not necessarily the opinion of Cambridge, but it's probably not going to be as valuable to focus purely on knowledge because actually if you've got a tool that can just tell you things, do you have to have all of that knowledge in your head? So what that means maybe is that it's focusing on skills.

So it's focusing on how do you communicate with that AI effectively, how do you fact check it and kind of make sure that you're aware when it's telling you something that isn't accurate or maybe biased or whatever it is. So maybe it's that chance to shift to more of a kind of skills-focused education rather than purely a sort of knowledge-led education.

But I don't know, I could be completely wrong.

Joanna Shilton: 00:24:04

No, that's made me sort of start to think about it differently as well. Like really has opened it up because you're right if you've got all that knowledge there.

I suppose it's a bit like with medicine or in the medical industry, you know, it's being used to look at all the scans and the results and it can just do it so much quicker. And if it's something like that, then is it freeing up the doctor or the teacher?

And this says to sort of, to do more, to learn more, to be more hands on with students or, you know, with the learning and looking at it.

And then it sort of comes to those sort of skills jobs where it isn't about that, that's where there'll be a rise, and maybe it's education and how to do the jobs, jobs that.

Hannah Marr: 00:24:43

We don't even know what they're going to be, yet I mean it's definitely not. There are no easy answers still. But I do believe that AI is going to be a help, not a hindrance. I guess that would be my underpinning thought on it.

Joanna Shilton: 00:24:57

So for anyone else, well, anyone that's working in education that's AI curious but doesn't know where to start. What would you recommend?

Hannah Marr: 00:25:04

Well first and foremost, if you haven't already, do experiment. That is the best way to learn. I mean you know this, your teachers.

But yeah, just do go and use ChatGPT or an LLM of your choice and just put a few prompts in there. Don't put anything confidential in there.

Just kind of, you know, a few put, I don't know, ask for recipe requests or whatever it is, just to get a sense of, of kind of what it, what, what it can do. I would recommend Claire Zhao has a Substack, which is AI and education, which is really good. She does a really useful kind of helpful.

I think it's a weekly roundup of stuff that's going on in AI in education.

Definitely recommend that one, and the DFE have actually just released some generative AI product safety expectations, which again, particularly if you are in a school or a university, those are good ones to look at and be aware of some of the pitfalls to watch out for, I guess.

Joanna Shilton: 00:26:01

Thank you, Hannah. And do you use AI in your personal life? Do you use it for recipes and that kind of thing?

Hannah Marr: 00:26:09

Quite often, for creating images, I must say to send as kind of gifts and things like that. So you know, I don't use it in a particularly high powered way, probably in my, in my life. But it just fascinates me what it can do.

And you know, almost every month these tools are getting better and better, so you have to kind of, if you know, if you tried something and it didn't work six months ago, try it again because chances are they'll have got way better by now.

Joanna Shilton: 00:26:33

That's great advice. So finally, where can everyone listening sort of follow on what you're doing or sort of learn more about AI in education?

Hannah Marr: 00:26:40

I am on LinkedIn so do please follow me on there. Get in touch if you've got any questions.

Joanna Shilton: 00:26:46

Brilliant. Thank you so much Hannah Mart. Thank you for being on Women WithAI.

Hannah Marr: 00:26:50

Thank you so much Jo.