Generative AI Education: Will Generative AI Transform Learning and Education

all right good morning everyone it's so great to see you here this morning um welcome um to day two of mit's week of generative AI um in particular uh we'd like to welcome you this morning uh to our Symposium on generative Ai and education um that which is happening here we'd like also like to welcome you to this space building 45 uh the stepen a schwarzman college of computing um just a fun fact um this is in fact the first public event um taking place um in this in this building um so great thanks um

for that we were able to have this here and um and you know hope that um hope that all goes well with the construction um it's a really beautiful space and we're very happy to be here um my name is Chris capizola I'm MIT senior associate Dean for open learning and a professor of History um and together with my co-organizer Cynthia Brazil Who you'll hear from shortly mit's Dean for digital learning we are really excited to convene this conversation um to together with the Office of the President and and so many other sponsors here at

MIT um for those of you who don't know MIT open learning is dedicated to a mission of transforming teaching and learning at MIT and around the globe through the Innovative use of digital Technologies we have uh in our ranks and and through our and distributed through the faculty expertise in digital Education and Research driven teaching and learning and we are excited and have been very active in meeting the moment of generative AI as you all know generative AI is already challenging Educators and students to rethink how we teach and learn in the classroom and Beyond

we're going to ask some of those questions here today first what role could and should generative AI play both in and Beyond the classroom in supporting effective engaging and Equitable learning for people of all ages we're going to explore the possibilities for leveraging these new technologies to improve teaching and learning um we're going to thoughtfully confront the challenges that generative AI poses to academic integrity and the evaluation of student work we're going to delineate crucial steps needed to ensure greater equity and access as new technologies are widely adopted and regularly refined and most importantly we're

going to ask what are the tasks that confront us as Educators seeking to equip learners for a world in which generative AI is ubiquitous now to address these questions we've gathered experts from across MIT both enthusiasts and critics Builders and theorists and I'm going to turn it over now to my colleague Cynthia to tell you a little bit more about today's agenda thank you good morning everyone uh I am so happy to welcome you all here into this amazing event and this fabulous space um I wanted to give you a little bit of uh kind

of the origin story for this event and why we uh designed it the way that we did um and then I'll talk you a little bit through through the agenda in a little more more detail so um when Chris and I started to think about how we wanted to frame this particular Gathering as part of generative AI week um we really wanted to prioritize um bringing together our community so really Gathering our vibrant Innovative community of students staff and faculty uh who are actively innovating and exploring how to use generative AI both in their own

learning in furthering uh teaching practice and thinking about how how can this become a technology that can actually have the positive change that we Aspire uh it to have and what are potentially the challenges in making that happen so it's really a chance for us to network to become more familiar with what each other is already doing to build community to help each other succeed in this really critical aspect of our MIT mission of education so as Chris mentioned our panelists bring uh really deep experience and expertise in Ai and education um with application ranging

from K12 to higher education to Workforce uh they bring careful consideration enhancing learning outcomes for diverse students and learning context both International and domestic and the responsible use of AI in classrooms and thinking about how we can prepare Educators RIT large around this um as well as how to really think about diversity and inclusivity in in all of this um so we also feature uh two 30 minute uh Innovation showcase sessions so you're not only having the chance to hear from our our our thoughtful uh colleagues but you're actually going to get to see and

and and and learn about what is actively being done to see demos to play around with it as well so we think this is going to be a really fun uh engaging part of our program so here uh here is our agenda um our first session is a fireside chat um that is going to really explore Innovative Technologies when they've succeeded eded in transformation and maybe times when they have failed to have the disruptive impact we'd hoped they would and why they'll be with Mitch Resnik and Justin Reich moderated by uh Philip Schmidt then we'll

move on to a panel on Reinventing learner experience covering K12 to Global Workforce and higher education we're going to have faculty staff and students participate on that panel I will be moderating that and will introducing everyone at 10:15 we're going to have our first 30 minute Innovation showcase to kick that off we're going to have half of our innovative Demers give a one minute lightning talk so just that you can have a precursor of of what will be out there all demos will be available for both sessions uh but we want to give folks a

chance to be able to to tell you a little bit about their demos so you can seek them out if there's something you particularly want to look at we'll have two of those um at 10:45 we're going to reconvene uh to talk about Reinventing teaching practice um we're going to cover topics from professional development pedagogy classroom instruction um and Innovative tools to support this and then our final panel is uh what we call our moonshop panel so provocative ideas about bigger Longer ranging impact provocative ideas um and how we imagine a vision a further looking

vision of how generative AI can can reinvent and uh uh reimagine how we learn in many different kinds of context um in our final uh showcase we'll wrap up with another one minute uh set of lightning talks and then people are free to uh to roam and convene and network um and of course we're going to give our acknowledgements to all the amazing staff who really helped put this event together so I will turn this back over to Chris all right so before we get started just a few words of thanks um and in particular

um we want to thank the office of the president president cornbluth um for for leading and and charging us um with this work this week um with several of the other organizing staff from Institute events and Institute Affairs who you see noted on the slide um I'd also like to give a special shout out um to the planning committee um for this uh this organization or for this this morning's events um which which includes some of the speakers um who who will be sort of uh serving twice both on the planning committee and as speakers

as well as some of the additional members um and then a special shout out to the staff of the media lab um and of open learning um with one in particular Steve Nelson um if you're here wave your hands he's um in the back um and Steve in particular um has gone above and beyond um in all this but really uh the thanks the list of people we could thank is um is endless and so I want to just keep things short at this point and we are actually um on time um and we're going

to jump right into our very first panel I'd like to welcome to the stage the the panelists and particular our moderator Philip Schmidt the chief technology officer of the axom collaborative thank you I'll grab this one thank you very much is this this working um well it's wonderful to be here this morning uh thank you for the invitation also thank you for hosting this uh what a moment to talk about Ai and education um I mean we we say that every week but it feels like just given the developments of the last 10 days maybe

even more so right now and also I can't think of two better people to have a conversation with about this topic this morning so I'm I'm excited to be here just very brief uh introductions Justin Reich is an associate professor of digital media in the CMS Department um Mitch Resnik is the Lego parer Professor at the media lab and leads the lifelong kindergarten group both of them have written actually books that are very relevant to this top IC life on kindergarten um the four Ps and um uh cultivating creativity through projects passions peers and play

uh on Mitch's side and then um recently both failure to disrupt why technology alone can't transform education and then iterate the secret to innovation in schools by Justin which in some ways when I was looking at the book books I thought actually those are the perfect Frameworks to to use to to think about this topic today so um we had a little preall and decided that you know we're going to be doing so much forward looking what's happening today how is it going to affect the future might be good to start by also looking backwards

a little bit what can we learn from past waves of innovation so um I want to start with a conversation about kind of where are we coming from how can we think about what we've seen before to understand what's happening right now and how can we set the stage and maybe make some adjustments to make sure that the the results of this wave of Technology really benefits all Learners um so maybe I'll start with you and go you just in um but you know just when you think about the link between technology and learning and

maybe actually also you mentioned something about like the MIT view on this I wonder if you could talk a little bit about some of those past waves and how people have thought about technology and education and how you think that might apply to this moment sure thanks Philip and it's great to be part of this discussion and thanks again to the organizers for putting this together and inviting and you know I do think that there have been these sort of ways of technologies that Ripple through society and each time there's a new wave of technology

I think we always need to consider how are we going to integrate this into our learning environments how is it going to affect learning and education and this happened you know you know 40 50 years ago as personal computers spread through Society you know what happened 25 years ago as the internet became prevalent through Society um and happening now with AI and trans and and generative AI and I do think each time we have to think about you know how is it that that this could be integrated and I think there're important choices to make

as it it's not something it just is is automatic of what's going to happen and to me it's really important to consider each time what is it that we want learning and educa what type of learning education do we want for our children for our schools for our society I think that can really guide us and I think too often people don't go back to that basic framework they think what can the technology do as opposed to what we want for Learning and education for children for schools for society um I didn't think my mentor

SE more papert U sometimes frame this in terms of two different schools of thought about what the role about you know different approaches to learning education he described one as instructionist and that was how is it that we can deliver instruction in a more efficient and more effective way um and of course there there a lot of advantages being more efficient more effective of how you deliver instruction and then another school of thought uh that that he term constructionist there was more about how can we Engage The Learner in designing creating experimenting working on projects

based on their interests and collaboration with others so they develop type of you know you know you know creative skills and curiosity and you could use technology in either of these ways to support them and in fact they have been used in both of those ways and now going back to the early days as computers first start to get through Society you know 50 or so years ago uh you did see these different ways and this gets a little bit different universities took some different approaches uh like the instructs Approach at least I see it

back then was really uh Carnegie melon became a real Pioneer and leader in that that they saw the technology could be used to make different types of you know tutoring systems and the computer could deliver instruction and do it we based on the responses it could then go tailor the next question so was a more efficient effective way to deliver instruction and to assess what the student was learning and and guide things appropriately and here at MIT led by Seymour and those working with them took a very different approach say how's this a new type

of you know material that kids can design experiment and and express themselves with uh and there's been a continuing threaded mit2 through that I see in the front Hal abon who's going be joining later today you know worked with seamour on the early version of logo and the continuing work on his projects like Apen vener and our own projects with scratch have sort of tried to continue in some of that tradition um and I do think maybe just to end this with you know the role of AI in this because it was the like in

those two approaches AI got used more strongly in that sort of intelligent tutoring approach and I do think traditional forms of AI were more sort of easily integrated into that instructionist approach uh that it was sort of possible to use it as a type of tutoring system so it got more adopted that way um and now as I see the next wave of AI with generative Ai and sort of this latest wave a thing that encourages me and makes it and I think is an interesting opportunity I do think there are greater possibilities now of

using this new wave of technology to support either instructions or construction so I think there's a lot more to that we can make some decisions how we want to make use of it so it's encouraging that we have opportunities to use in different ways I will though say what's discouraging when I look at what's happening in my mind in the Ed Tech world most of them I think are using this technology just to reinforce traditional approaches to education so it's discouraging how it's getting out there in the world in my mind there's just reinforcing some

traditional approaches to education when I think real big changes are needed but I think there's great opportunity but also some signs of discouragement and things that you about about the way it's actually getting adopted in schools yeah and maybe um uh as a segue to you Justin you know Mitch's point there are always when these new waves appear there always these Grand pronouncements about how this is fundamentally going to change everything how it's going to improve everything and then we often kind of the results end up looking a little bit different as you point out

in your book so I wonder if you could maybe talk a little bit about some examples from the past but then also is this one going to be different or you know where where do you see this heading yeah I well so I was I was was at a conference at Harvard um where a fellow named Ethan mik um who's been very enthusiastic about AI was talking about his sort of optimistic views um and he was saying uh um that AI was going to democratize education that everyone is going to have free access to these

kind of personalized tutors um and and someone and someone was like um well weren't muks supposed to democratize education he's like a no muks M muks were Silly muks were just a bunch of you know you just watch like people sitting by themselves watching a bunch of videos like how could that be interesting and I was thinking you know 10 years ago we were like well of course textbooks weren't going to democratized education you need these really like carefully made personalized videos humanistic videos that's what was going to transform education and I'm sure 10 years

from now there's going to be someone sort of watching recordings of Ethan mik's talk being like of course you're not going to transform education with a bunch of kids sitting in front of tutor Bots like no one wants to talk to their computer um it doesn't matter how compelling it is um you know when a kid is sitting down trying to learn algebra or something like that um you know you can build a great instructional sequence with textbooks you can build a great instructional sequence with videos you can build a great instructional sequence with a

torbot but none of that matters because the kid doesn't really want to learn algebra um what the kid wants to do is build a relationship with their teacher and the peers around them um you know learning is just a fundamentally social uh Enterprise um now there are a relatively small portion of people um in the world who are pretty good at teaching themselves using anything um lots of those kinds of people end up at places like MIT um lots of those people develop the privilege to be able to develop education Technologies um but the vast

majority of people when we learn especially when we learn thing that Society asks us to learn not the stuff that we're most passionate about um we really learn it because we're in a social Enterprise with other people um you know I I couldn't agree more with Mitch I was I was thinking of Seymour papert's book Mindstorms and trying to think like go back and read that book and everywhere it says computer just replace it with AI and see if it still has a pretty good argument you know one of the things that seem more paper

famously sort of phrased is like are we going to use computers to program kids or we going to have kids program computers I think you could ask the same question like are we going to have ai program kids we're gonna have a kids program AI that to me you know 40 or 50 years later is is still a great question um you know maybe the thing that I would add to Mitch's list of sort of fundamental considerations like especially if you look back at where technology has struggled to make change um is that our existing

education systems are complex social Technical Systems um they like people complex brains that learn in certain ways they learn with other people our schools have multiple competing purposes that they're negotiating between about preparing citizens Preparing People for work you know providing pastoral care just supervising kids during the day things like that um and you know historically technologists have not done a great job of understanding those complex social Technical Systems so you sort of build stuff for the way you wish students learn as opposed to how they actually learn you build stuff for the way you

wish schools were as opposed to how you actually were and then these things get dropped into systems and they don't do what we want them to do and they don't work the way that they want them to work and so really try you know um I mean I this is like this is a pitch for for failure disrupt it's a book and I teach a class in the spring called learning media and Technology where the basic thesis is like if you don't understand these systems that you're building things for um you're not going to build

things that work in those systems yeah so it's probably combination of what's what systems building for and then what type what do you want young children to grow up as so I think those are sort of two things to make it work it has to work in the system but to go in the direction you want so you could things that have things that are transformative but transformative in a bad Direction I was thinking the title of this is will it transform learning education uh even if we said yes to that there's then a question

is it transforming the direction that we want it to transform it in and and the technologies will not transform it that I am 100% clear of like if you think you're going to make a thing that you download onto someone's phone and that Transformer system like I promise you that will not work um when systems change it is because we partner with Educators we partner with Learners we partner with families we build their capacity um for a new technology to be actually useful you often need to change a whole bunch of other things like if

you like a I made the best possible technology for Project based learning well then you need a new curriculum and then you need new professional development you need actually a different kind of schedule because you can't do that kind of learning in the sort of class periods that we have in most places so there's a whole bunch of other parts of the system that have to change and organizations you know like the lifelong kindergarten lab like other places that are really good at at building technologies that actually have the capacity to to to make changes

in how learning happens it's not just about building this thing and downloading it on people's phones or their tablets or whatever else it is um it's about helping those organizations make the kind of systemic change you know what a new technolog is only as powerful as the community that guides their use um and so thinking about how those communities change is a central part of of anything that will look like transformative I mean you know or even just you know it's also fine not to transform education it's also fine to have like iterative improvements that

sort of systematically you know in small ways over time accumulate to make things better um my colleague Ken kadinger who's at CMU says that step change is what 25 years of incremental change looks like from a distance is it okay for us keeping each other modation job I've ever had yeah but one thing that's come to mind is that you another thing that might be a little different now it's not the technology but the way the world is right now might require some different types of things we have a greater recognition in some ways to

me when I was sort of laying out sort of this what seemed more quite a constructionist approach to me it feels it's more needed now than ever partly because of advances in Ai and many other things I think everyone would agree that the world is you know changing more rapidly than ever before and that kids are growing up and there's going to be you know a never- ending stream of unknown and uncertain and unpredictable things and the proliferation of AI systems and tools is just going to speed it up more and add to more disruptions

so it's even more important than ever for young people to grow up with the ability to deal with the unexpected things that arise in their lives so it'll be more important than ever for them to develop as creative thinkers we'll come with innovative solutions and so I do think that a lot of traditional approach to schooling which was really about helping learn particular Concepts and particular skills is less relevant today although I mean some of those might be important as part of the picture of what you're doing but it's so important to make sure that

young people grow up as sort of creative curious you know collaborative thinkers in a fast changing complex world that's becoming even more complex and fast changing because of AI so it's not just what AI can bring as a tool but how it's affecting Society leads us to take children to grow up to be successful in the workplace and in society and as Citizens will need to develop these types of you know you know become more creative curious collaborative in order to succeed which I think are some of the goals of this approach of trying to

enable kids to create design collaborate work on projects based on interest is the way is towards that so it feels more important now than ever to keep a focus there I wonder yeah I mean I think there's you know how does the education system make sense of this moment and how we've talked a little bit about how do the technologist make sense of this moment why building an app may not although octo studio is an app so why building an app may not be enough to to Really transform education but I wonder Justin I know

you talk a lot to leaders in the K12 systems what's the response for them what are the questions they're asking right now what are the concerns and how do you respond to that probably like faculty here lots of other places the the number one entry point that school staff are currently encountering AI with is described as cheating um that is that you know other things they could actually take a couple of years to figure out for for you know like how how should AI change the way we teach social studies how should AI change the

way we teach earth science like if we didn't figure that out until 2027 like we'd be fine um but actually this year school superintendent Schoolboard School leaders need to figure out like what are we going to do about the fact that um Sil actually I was joking with them I you know to Schoolboard members if you're Schoolboard member elected in 2019 you had the pandemic in 2020 and then you had actually a year in most schools that was harder than the pandemic in 2021 2022 when all the kids get came back super angry and they

weren't really sure why they were angry but they were whoever nearby that happen to be teachers and then Silicon Valley invented this machine that can do all of the kids homework for them it's like it's like give me a break you know like can we cut us some slack here um you know but I um so that is the most urgent concern I've encouraged School leaders not to think about to stop thinking about as a cheating problem and think about a problem as bypassing useful cognition like we had invented a bunch of exercises as teachers

um that are useful for helping young people develop certain skills and proficiencies content knowledge and things like that and then we invented this machine that can do most of those kind do can do many of those kinds of things for them especially in writing however over the last 50 years maybe a hundred years we've done that all the time um it used to be a great idea to ask people to summarize lots of things and then we invented encyclopedias it used to be a good idea to ask them to compute math problems and then we

invented calculators it used to be a good idea to ask them to translate uh you know documents from one language to another and then we invented Google translate so schools have basically faced for decades um tools that let people bypass useful cognition um and schools then have to figure out like okay so what of that do we not need to do anymore like what are the kinds of I don't know a student was telling me the other day I guess it was a year ago I was like what are you using chat GPT for and

they were like to format latch in my papers I was like great never think about that again just like if a machine can do that for you that is not useful cognition um but there are other kinds of things you know where we we're asking students to write not just to generate a document but because writing is thinking um and bypassing that Pro process may be bypassing useful cognition so we're going to have to go back you know um Math teachers figured out pretty quickly that like you can't ban these Technologies Banning calculators was a

bad idea they also figured out that there were times when Walling th things off was quite good um to say like actually you should be able to do this without a calculator um because you won't be able to do later kinds of math thinking um if you don't have some automaticity with these sorts of skills so for this part of the lesson the week the homework whatever don't use those tools and then we'll figure out where to bring them in later um that's the most urgent thing that I think um School leaders are working on

and I'm sure that there are ways that we could help them um think through those kinds of challenges um a big part of that will just be continuing to think of like what are interesting ways to to have people do useful cognition that can't be bypassed with chat GPT if there anything way to figure out how do you do useful cognition you know without using Google without using Spark Notes without using calculators all these kinds of things um you know and then I think the next thing they have to figure out is like well how

should there schools change um based on like what should we be teaching differently you know Mitch pointed to one sort of thing which is that a lot of what effective schools do um is they help young people build capacities where we have comparative advantages over computers um that was work that like you know Frank Levy did here in the economics Department David Artur in the economics Department Richard Rene over at Harvard said like if a computer is a good at a thing um then probably you know think carefully about whether you not you want to

spend a lot of time teaching humans to do that thing um because they probably can't get a lot of work doing that um however there's lots of things that computers and AI maybe are not going to be as good at and we should figure out what those things are and really try to emphasiz um skill development in those areas and Mitch gave I think a pretty good list of what some of those kinds of things are um the other thing is for schools to think about is you know I think I think over the last

40 years as have been technological changes a pretty good question to ask is how are the disciplines changing and how should that change our instruction you know I was a high school history teacher in 2003 when the world's archives and governments were digitizing every primary source document they could find um and the field of History was changing rapidly and so our instruction as history teachers changed to incorporate many more and many more diverse Source primary sources in our instruction at least in in places that I think did that well um you know so we could

start asking questions in schools like what Fields seem to be changing fastest with generative ey you know it's it strikes me that most software engineers in the next couple of years won't do any programming without some kind of co-pilot um that that will just become a completely normal part of instruction so if three years from now the way we taught people to code always had a little co-pilot helping them that seem like a good way maybe to follow what the disciplines are doing um you know if you're teaching a 12th grade career technology education class

in design you know the field of design is changing incredibly rapidly because of generative Ai and if you're not teaching some of those skills even if you don't totally understand them yourselves um you're you're probably not preparing young people for for that world on the other hand there's going to be other disciplin um you know poetry Memoir social studies um you know when I when I taught uh US History um I was seeing out in the hall that there was a whole lot of emphasis on sort of personalized education um when I when I taught

history there was very little that needed to be personalized um what I wanted to do was have young people with very different Minds look together at the same time at really important primary sources there's there's very little that personalization helps with that like I don't need kids moving at their own speeds or I actually need them at the exact same place at the same time thinking and talking with each other about the same thing because that's how we build the skills for deliberative democracy um so the you know thinking about how the disciplines are changing

can help us think about how schools should change before I get to you MIT um just reminds me of this fantastic quote that I think Chris capazo shared with me some time ago about um writing at MIT and I don't know who actually said this originally but it's something along the lines of the creation of text is incidental to the teaching of writing at MIT I thought there was such a great way of thinking about writing as it's it's something it's a muscle we develop or it's a skill we develop that actually represents critical thinking

persuasion understanding the world around us like these are all those important things and writing is beautiful and is an important skill but it is kind of just one manifestation of these other things that we hope people develop and so strikes me as a good example of rethinking how we're approaching this teaching and disciplines now that technology comes in but I know you want to get in and also I wonder if it's related to what you just shared with us before went on stage this response from the education sector from another country that you mentioned you

know I wonder if that's maybe there's an openness there or an interest there to engage with this in a way that that is now created by this new technology although let me start with first just to reinforce something that the just was was I liked when you talked about how it becomes even more important to help you know you know young people develop the things that machines aren't necessarily so good at but um and see this is a great opportunity for us to sort of focus more on what makes us most human you know the

ability to think creatively to engage empathetically to work collaboratively because I do think those are special things that are really important and will be important things for young people develop but also for educators and teachers to really lean in on those things that are most special because I do think there are a lot of things that human teachers are especially good at as you mentioned earlier the social side of forming relationships relating to lived experiences creating a caring Community those are things I don't see gender AI playing you know big role and so and I

do think that having leaning in to make sure that Educators have the opportunity to develop those skills as one reason I get really frustrated with the way things get positioned with a lot of these commercial things where it describes as a worldclass tutor the same as a teacher it's not the same as a teacher it's not say it's not very valuable you can do good things but I I wish people would stop comparing because I do think we should we should see what it is that AI can do really well in that context what do

the people do really well and make sure that we highlight that and then help Educators prepare on that me one other thing that came to mind as you were talking with uh the concerns that people have when people start using technology to do things like you know to be able to just you know as chat G and hand that in as your assignment and how do you get away from that quote cheating what it reminded me of was having a conversation here at MIT this must be maybe 20 years ago when videos of class were

first starting to go online I was in a meeting where people said this is a real problem we put the videos online and the students aren't showing up to classes anymore and my reaction was well that means you should be doing something else in your classes if they find those videos as useful as your classes you shouldn't be running your classes that way and I feel somewhat the same way with writing assignments if you give a writing assignment the Chach P2 can do a good job on that might not be the best writing assignment um

and I thought it made me remember think I was in third grade uh and I handed inside we had to the biography of someone we admired and I did just look up in the cyclopedia entry and I didn't copy it but it was just a single Source I did it and there was a strong critique from the teacher about you know you should make it more personal about why it is that you think that this person was important and I thought that was really important because that was something that wouldn't be automatically generated but she

was really asking me what it is that I found important about this how that I relate it to my life so there are ways you can ask different things that I couldn't just do from the encyclopedia or these days just ask chat GP you could like tell chat GPT like a few basic beliefs you have and then ask it to generate your opinion um on these things which is sort one of the problems are coming up I can I say one made me think of was just uh coming back to the idea that we're creating

social environments and this this notion that like what we should be doing with tutor Bots um is getting them to replace teaching functions because we don't have enough teachers and sort of one-onone um a question that some students and I came up with in class A few years ago as I always really enjoyed is um what is the humano human interaction generated by a new technology um so when you add a technology to a learning environment not what is the human technology interaction you generate but how does that technology generate a humano human interaction um

there's a great researcher at CMU named Carolyn Ros who's worked on intelligent tutors for a long time so there's there's nothing there's it's EX extremely unlikely that you will come up with something that people have not worked on in the last 70 years um since we've had computers the size of your living room learning scientists and computer scientists have partnered um to find all kinds of ways of getting computers to help teach people um and studying that history can be enormously helpful because there's lots of ideas that we've already tried and things that we can

learn from those ideas we've already tried so for instance Carolyn Ros who's been building tutor brats forever um very quickly lost interest in any thing that interacted directly with a student she was much more interested in tutor Bots that you put in places where two people were already conversing with one another it was basically an aid to help two people think more expansively solve problems when they came up with shortcomings in their knowledge or other kinds of things like that so that I think is you know a potentially really interesting way of of using technology

that will enhance social interactions rather than replace social interaction and there's already a bunch of stuff that's written about it um and there's probably ways a generative AI can do it more easily or better or more creatively or other kinds of stuff um but there's a lot of fundamental thinking that's already out there so that's actually a perfect segue to think a little bit about how can we design the Technologies to do things that are more in line with what we're trying to achieve in learning and education and one of the things I've appreciated about

both of you is that you're so thoughtful and and sometimes critical of what's happening in technology but at the same time you are building Technologies and you're tinkering with Technologies and you're trying to use them in in effective way so I wonder if you could both talk a little bit about what are your current activities I know the million tutor moves project I know you're experimenting carelo and others in your group are experimenting with how to integrate AI so like you know what are some ways that you're bringing this into your work sure um so

uh maybe I'll try to do it a bit in Mitch's framework we just say like what is the learning goal that I have so I'm really interested in how teachers learn and I think one challenge that teachers have is they have insufficient opportunities to practice um so when teachers learn they go to seminar rooms and talk about teaching and they go into classrooms where 26 kids have to learn how to fact their polinomial that day um and there's not really a practice field so we in my lab we build technologies that create these simulations um

where teachers can rehearse for and reflect on important decisions in teaching and we think they're interesting ways um that generative AI can add value to those simulations um we're starting playing around with some feedback agents um so you say some stuff um and some coaches predict the kinds of things that people might say and they predict some things that might be useful to say you know given given people's responses and then AI figures out how to match you know novel responses to kind of pre-existing or modified pieces of feedback that's been sort of fun to

play around with it's been particularly fun because we tried to do it a couple years ago with things like Bert and gpt3 and it didn't really work um and now it seems to be working better and so that's kind of fun um another project that I have um is there's a whole bunch of people who are interested in tutoring um an unusual thing about our education system is that we have very few machine readable records of what teachers do so we have we've terabyte zillions and zillions of rows of data about what students do we

record students all the time we don't record teachers nearly as well um so with some colleagues I'm sort of organizing a group called The Million tutor moves project um that would look at tutoring platforms and classroom recordings and would try to collect data um about what it is that teachers do um you know one of a simple way to frame our hypothesis is that if you build a bunch of tutor bots on web scale data um they might not work very well because because most of what happens on the web um is that people answer

questions um but actually what really good tutors do is question answers um and so we might need a different kind of data to be able to do that although I was just talking with someone who supervises lots and lots of tutors um and uh they've been doing some research on them they said one you know one of the main things that effective tutors do is just keep students talking they just say keep going what did you mean about that you said something there keep going keep going keep going and the reason why that works um

is that when human being says to another human being that was pretty good keep going say more about that and then stops talking there's this awkward silence um and the student is like I really should keep talking just to fill this awkward silence when a tutor bot says that was pretty good keep going um the student going to be like no I don't want to keep going you're a computer I just told you what I know about it now you tell me um and uh I to me it's just a great example of where uh

you know our our our hopes for what you know text can be generated by machines or even voice can be generated by machines may be swamped by the fact that uh what people really want to do when they learn is to interact with other people yeah as as I think about you know worthwhile things to work on it all as I was indicating before it's all what can we do in the service of project-based design oriented uh interest driven Collaborative Learning so always saying that's the goal and what can we do to help support that

goal and you do see examples of people using chat GPT towards that goal as they're work on Project they use it as a resource so I do think you know in the short term there's a lot of things that can be done in the short term if we just help frame it of giving people giving the learner control over using this that is there the same way that they look up things on Google online and they look at YouTube to get at videos about the what when they're work on a project and an issue comes

up and they're not quite sure how to deal with it or you know they're not quite sure what to do there are a lot of resources to use talk to another person is another one and all of these are things we can do there's a an additional resource it's a special type of new resource so I think that's one thing that I think trying to get people a better understanding of how they can make use of this new resource and balance it with its advantages and disadvantages over others I guess I also been thinking about

in our group you know as you know we do a lot with you know developing programming languages for kids and what's this going to mean for the future of programming languages and coding a lot of people wonder whether it's going to sort of you know be the end of programming as we know it um and I do think it's important for us to be open to to change because we shouldn't hold on tightly to things just because we're familiar with it uh so it's important to look at different things I do have colleagues who are

starting to explore this space uh former student Eric Rosen bound who now works at the scratch Foundation is doing things of integrating you know some image generation tools inside scratch uh so that you know if you're right now you could sort of look online to get an image you do the library or use the paint ater but also you should just be able to say you want an anime style purple frog and it generates it now if your main goal of your activity was to learn how to draw a frog that's not a good thing

to do but if you have if your main activity is to make an animated story that seems like a good thing to do um I think more ambitiously I talk with my colleague Brian Silverman who's been involved in development of logo and scratch and other programming languages and he's really very interested to explore this you know conversational interface that we made a transition you know 20 years ago with scratch and other things to a block from texts to blocks and he's saying well maybe conversation is the next thing to be doing will we in the

future just have a conversational interface and for me as I think about that I do want to be open to it but I also want to think carefully about what are the qualities that we appreciate in our current ways of doing things there's certain things that I really appreciate about the tool the coding tools that we've built coding environments the ability that they provide people with the ability to create things that they really care about and to be able to refine their Creations in a way that sort of matches what they want for them to

have a feeling of a sense of control and agency so they're in control of it just their relationship with technology I think it's important for young people to feel that that it's not just there you know they're interacting with they're giving something to them but that relation with technology is something we've always wanted U I think we want people to learn about the process and strategies of design uh to go beyond coding but how to break complex projects problems into simpler Parts how to iterate when you try something and iterate and refine it those are

all things that are important to learn about the design process and also I think we always appreciate just many people find joy in doing this and I would hate for that to get lost uh we do a lot of work with the Lego company and they have this phrase Joy of building pride of creation and I think a lot of kids where they're building a castle with Lego bricks or an animated story with scratch do have a joy of building and a pride of creation when they share it I don't want that to go away

and and I do worry about that sometimes when people just use you know an image generation tool they don't feel like they really were the ones that created it so there are a lot of these things that I don't want those things to go away or or one thing we've already talked about is help kids think about their own thinking and right now with writing programs as they are I do think that as we investigate new approaches like conversational approaches for telling the computer what to do are there things that are being lost does it

meet all of those criteria and what my guess is and I don't know I think that's yet to be known I think what's going to be challenging is my guess is these new interfaces will meet some of those well and might even enhance it and some of them they will degrade and then it's a challenging thing because they'll make certain things easier and will remove some barriers and other things will lose from that and then how do we make decisions about what what to support but it's an ongoing process to get a better understanding of

that and I think that's an ongoing conversation that hopefully we can have throughout the rest of this day this half hour kind of flew by I thought it was a great introduction to some other ways to thinking about Ai and learning and education and we're I just want to remind myself also we're sitting in the College of computing many of the people in the room are building these tools or will be building these tools will'll be applying them in creative ways we'll be sharing how you're using them and so um you know for me what's

special about MIT is always this tinkering building things to explore what the future could look like and it's it's alen K's quote kind of summarizes it perfectly for me which is the best way to predict the future is to create it or to invent it and so I'm hoping that this conversation was maybe inspirational a little bit for all of you as you're thinking about how to bring this technology into teaching and learning and looking forward for the to the rest of the day and want to thank the two speakers for this fascinating conversation thank

[Applause] you