The Educational Promise of Narrative-Based AI


Dr. Jeremy Roschelle, Executive Director of Learning Sciences Research at Digital Promise, joined the po dcast to discuss the next generation of artificial intelligence for educational storytelling and what they’re hoping to build with their $20 million grant from the National Science Foundation.

The technology is coming along to whichan AI assistant can craft stories, can enhance stories, can customize stories.How can we use that to get back to a narrative center learning experience?You're listening to enrollment growth university from Helix Education, the best professional development podcastfor higher education leaders looking to grow enrollment at their college or university. Whetheryou're looking for fresh enrollment growth techniques and strategies or tools and resources, you'vecome to the right place. Let's get into the show. Welcome back toenrollment growth university, a proud member of the connect Evu podcast network. I'mEric Olsen with Helix Education and we're here today with Dr Jeremy Rochelle, executivedirector of learning sciences research at digital promise. Jeremy, welcome to the show.Thanks, hrek. It's great to be here. Really excited to haveyou here and talk with you today about the educational promise of narrative based Ai. But before we dig into that, can you give the listeners a littlebackground on both digital promise and your role there? Sure thing, Eric.Digital Promise is a nonprofit organization that fundamentally connects educators, researchers and innovators towork together and we spend higher Ed ktwelve preschool, really lifelong learning within digitalpromise. Eric, I lead the learn earning sciences research group and our missionis to conduct fundamental research on the future of learning, but in a veryapplied way. We want to conduct research it's useful to educators everywhere. Soexcited about this conversation because I feel like AI and education gets thrown out alot and often it doesn't really mean Ai. So maybe to kick us off today, can you give us just a high level overview of the current state? Where are we now in terms of AI in education? Eric, ifI could draw a timeline, it would stretch back about fifty years. It'snot new, but we're at a very exciting moment. So let me justtalk for a second about that. Really, since the birth of AI in thelate s early S, people have been imagining what could this mean foreducation, and we have had throughout the years some very successful applications, butthey've been narrow. Those applications are in areas of learning that are more puzzlelike or more logic like. Math is a classic example. Chess, ofcourse, is a classic sort of example. We've all had early warning systems thatmany universities will be familiar with. That given early indicator when a studentmay need some additional attention to stay in rolled, stay engaged. Where weare now, though, is like a hockey stick turning upwards, and that'sbecause the capabilities of ai are suddenly exploding. We have much more capable possibilities beforeus and we all see this every... in our lives. We're startwe're talking to home assistance, we're talking to our watches. You know,we're used to things that complete our sentences, suggest people we ought to talk with, and so it's becoming commonplace. The underlying capabilities, number of patents, number of researchers, just expanding really wildly right now and within the nextfive years that's going to mean new types of applications, not in education,not just those narrow ones that we've seen for the past, you know,thirty or so years. Love that overview, Jeremy. Now can you give usa brief overview of this National Science Foundation Grant and the really exciting programthat you're hoping to build with it right? Happy to do that. I amreally excited in my colleagues are excited to kick off this Ai Institute theNational Fund Science Foundation had a very competitive awards program and selected a series ofpartnerships to come together, and each partnership had to connect two things. Onthe one hand, fundamental and advances in AI itself and, on the otherhand, and application area, in are case learning. So our focus wasreally unengaged learning. New Fundamentals are becoming possible. We wanted to take thatopportunity to rethink engaged learning and the place we've chosen to start for that rethinkingis to think about how stories throughout history have engaged learners, how we,for the whole history of humanity, have sat around campfires telling stories and that'sbeen a primary modality of learning. But somewhere along the way, if learningbecame rigid, it lost the storytelling emphasis and it became something that is offputting, foreign alienating to too many of our students. So we want tolook at the technology is coming along to which an AI assistant can craft stories, can enhance stories, can customize stories. How can we use that to getback to a narrative center of learning experience? All right, you're blowingmy mind a little bit. I do want you to talk about this mentaledged that I'm going to in my own head now, because do we thinkthat learning and education evolved from storytelling toward standardized testing, not necessarily because wethought it was more interesting or better pedagogically, but just in order to scale it? has that been the forced evolution of education? In a nutshell?I agree with that. Eric. You know, standardized testing came up abouta century in large part because the military needed to assign people to jobs inthe military and they needed an efficient system.

So the problem was very much ascale problem that drove standardized attesting. How do we take this mass ofpeople we're bringing into a particular occupation and assign them to the right roles?It wasn't necessarily for the betterment of their learning that the standardized assessments were developed, but once developed they were quite useful at scale, and so you couldsee the education system start to tilt towards being organized around those assessments. Ithink we've lost we know. I don't just think we know, we've losta lot of learners by going too far down that road. And you know, no one starts out with their own child sitting on the SOFA, openinga book and saying Hey, I'd like you to take this multiple choice test. We know as parents we start with stories, but it's also true thebest CEOS. I know you're laughing, but read the Harvard Business Review aboutcultures and huge companies the best CEOS or storytellers. Yeah, Jeremy, that'sso fascinating because I think to your point in terms of the learners that thisevolution of education works for isn't fully inclusive. anecdotally, we've seen that online educationtoday seems to work best for a very, very self motivated students whenthey do not have that benefit of of of storytelling. Is that partially whatyour research is hoping to solve, for to develop adaptation specifically through storytelling,to keep more students moving and engaged in the learning process? Yeah, absolutely. You know, one of the things that's coming along on the foundation sideof AI, and it's scary but it's also promising, is the ability tosynthesize things synthae size, maybe the narrative for how a sports game might playout. You know, it's take a few sentences that someone started and createa complete paragraph that follows the thoughts they started and sounds vaguely like a humanmay have written it. You know, it assistant that can track where you'regoing in a more narrative sense and play along towards that. So what we'reseeing in terms of student engagement is students are engaged in part when they feelthey belong. That's a very powerful human thing, belonging, and stories orsomething that make us feel we belong or can make us feel like we don'tbelong if they feel like they're not welcoming our identity, our personality. Forinstructors trying to teach hundreds or thousands of students, they don't have the timeto make slightly different, variant story experiences for all their students to be partof. But imagine in that we are recontextualizing, for example, a sciencelab experience to be really about trying to discover something that's that's fundamental to aproblem. You know, maybe there's a...

...disease spreading. That's something we're familiarwith right now and people have to address some of the scientific basis in differentcommunities, and so that story about disease spreading. They need to vary alittle bit with different groups of students to let them follow what they are reallyinterested in while still keeping some of the curricular emphasis that is the goal ofthe course, and so problem based learning, work, learning in teams, learningthrough collaborations. There's intrinsically a sense of culture and socialization and story.It intrinsically is very motivating students to be part of that and what we're lookingin at, instead of breaking that come back to a very instructor driven,one size fits all narrative, to try to keep that going a little bitlonger stick with where the students want to take the learning experience. Jeremy,you mentioned being at the the beginning of this hockey stick curve that you're startingto see become more and more visible. What's the easy ask for AI?What do you think ai will be, for sure pretty great at in five, ten years, versus the much harder and long term mission you're helping tostart solve for? Yeah, that's a that's a fascinating question. You know, we've seen a lot of basically machine learning type applications already and I thinkthat'll continue. What these applications do is look for patterns or associations between twothings, and so I do think one thing that's going to become a loteasier is kind of a course recommendations where you may have an AB alternative.Should I start with a discussion or should I start with the presentation? Shouldthe discussion be like this or be like that? I think we have somuch data gathering in platforms that you may see instructors get some pretty sensible recommendationsfrom a course we're planning assistant, for example. That seems very it seemsvery feasible and attractable to me. I just want paint for you what thedifference is between that and what we're trying to do, given this twenty milliondollars, to try to do something exciting. Yeah, and really, Really Eric. The difference is pretty straightforward. Most of that kind of work isrecommending between two things that pretty much already exist or are just very routine,and you're just want to know is it better to do more of a ormore be, or do a first or be first? Or is a orbe better for a particular student? But the a's and B's are the samethings. Yeah, what we're trying to to hear is create a kind ofexperience, this narrative centered experience. That is very rarely done now because it'stoo hard to expensive. If you know to do it any kind of scalehas been too hard and so the only... you see it is a massivelytalented faculty members who can weave their students into a into a story. Butyou're not every faculty member is going to have that talent, and so we'retrying to break through to that. Really is an experience that cannot be createdtoday, cannot be assigned a or B. It's just a different type of learningexperience. Can You alpos picture what that looks like? I think thethe sequential base testing is something that's easy for me to wrap my head around. We think that a, you know, hour and twenty long learning session withmy class could covered these ten things. And what sequential order are learning outcomesmaximize? We a be multivariate test and find out that one I canget my head around. What's an example of what this looks like when finished? WHAT DOES AI base storytelling look like? What what a class look like forone student versus another? Yes, Eric, I have an example.I like to use that. I think it certainly would fit high school,maybe middle school. We know already that inquiry based science is really important totranslate students. Science isn't just facts, it's a process. US With acertain quality, and so I like to imagine a group of students being engagedwith the basic outline that a month from now we are going on a tripto Mars together and we're going there for a scientific project where we want tocollect some data on the Martian soil, and we need to plan this out. What instruments are we going to bring? Where are we going to collect oursamples? How are we going to analyze them? And you know,we may have some different interests about what we want to really look at onMars, just different science, different questions we may want to look out.But this whole course together, maybe there's a hundred students together. We're goingto be in small groups. We're all going on this big ship to Marsin a month. So let's figure out what you want to do when youget there and let's spend a month of our class time learning about the fundamentalphysics or chemistry or whatever it is that we need to know, what thescientific instrumentation is, learning that data is, sampling plans, planning at our analysis, and then, you know, month from today mission starts. We'regoing to get on that rock tip. We're going to go on this ofcourse now simulated mission to Mars and we're going to do it. We're goingto collect our data, we're going to come back to class and talk aboutwhat we what we found when we got to Mars. So to me that'sa kind of experience that we cannot we can see how exciting it would be. We can't produce it right now today. Would just be too expensive to pulloff. Maybe in a museum you could do it today, but doingit in routine online learning to complicated. So that's the kind of thing wehope in five years we've really made some progress towards. And when? What'sthe difference between I want to sign up... job for that class? Soundsamazing. I want to join that class myself. What's the difference between findingthat Master Storyteller instructure to develop that sequential story versus relying on Ai to helpdevelop what that narrative actually is? Well, I think they'll be a blending ofthose two. I think I always really when I think about learning theoryor instructions, a sign. I always want to iterate between sounds, researchprinciples and what the bay best instructures do. And we want to be really humancentered. I think the way to make this understand the feasibility of thisfor ai is that's the customizations, it's the variations that really will be toomuch for anyone structure to do or to manage. But that little bit,it's just what you and I do in conversation. We don't follow a scriptwe made ahead of time. We listen to each other where a little responsive, we take it a little bit different direction. And today instructors, youknow even tomorrow structures under the time to do all that. But I thinkthat's what we're looking at, is within something that's based on sound wisdom frompractice, practice of great instructors, based on sound learning theory. We knowthis is the right overall experience. How can we bring an AI'd make itfeasible to to that right at scale? It's fascinating and so again trying totranslate to myself, the layman, is it still helpful for me to thinkabout it? Is there is still that sequential experimentation happening, but within astory you can't just take chapter for and put it in front of chapter onewithout synthesizing and creating new transitions and and is that the missing piece of beingable to utilize AI within a story? Yeah, absolutely, and then also, you know, bringing it to students in experiential format that that sort offits to for example, a voice assistant instead of a new text to read, perhaps some imagery that may weave in some of their own drawings and soforth. But contextualize that and, you know, just to make it makethis feel realistic, but not just realistic, to get the learning relevant resource inthe right place. So I know for me it would be very excitingif I was part of a scientific mission and I had made some plans inmy notebook if now some of that's notebook pay ages were automatically coming together inan image. For me that looked like it was part of this this rocketship that we're getting on, and I would really feel it. Sometimes thesecues really que you that it's real, and it's just that feeling of itbeing real what a students complain about about...

...higher education that seems so detaxed fromreality. So what can we do to make this feel like this is areal experience and you're learning in a real I think a little bit can goa long way. Jeremy, I'm not sure I can wait five years tosee what your team builds. This is too exciting, but for those ofus who have to stand on the sidelines and cheering you on. But butwant to start thinking better about ai today. Any next steps? Advice for institutionswho are looking to leverage where I ai U is today and and preparefor where AI is going to improve their students learning outcomes? Where should theystart? Yeah, great questions. You know, there are some nice readingsout there that just summarize the state of the art of where AI is.That's one thing. It's just to understand what the technologies are. Ability todo naturalistic input, like voice, like making sense of a sketch, abilitiesto sense patterns, abilities to synthesize constructive action or text. So, anyway, some reading about what the capabilities are. I think it's important to compliment thatwith some readings about human center Edai. Sometimes it's called responsible ai, sometimescalled Ethical II. We have issues of bias, because there are alot of challenges with this technology as well, and so tracking some of that,those issues and becoming aware of that is super important. Starting to thinkabout policies and safeguards and really how they'll be informed consumers of anything that mightbecome available and don't just take promises for granted. Going down that path alittle bit more. Something I think is important is just developing some ability tolook under the hood or have someone on your team who can, because thereare superficial promises being made out there that something has ai and it's good,and sometimes what's in it is really trivial and the team doesn't really the teamthat built that thing really hasn't gone particularly deep so as it kick the tiresprocess. So, anyway, those are the three things I'd say. Youknow, read about these fundamentals that are exploding and open your eyes that,think about the ethics and think about how you're going to build a team thathas enough capability to really kick the tires. Jeremy, thanks so much peer timetoday. What's the best place for listeners to reach out of that theyhave any follow up questions? Yeah, I would be happy to engage listenersby writing to me at my email address. It's j Rochelle at Digital Promise Dotorg, J r Os ch Elle at digital promise that Org, and youcan also pretty easily find me on linked and if that's easier for you.Jeremy, thanks so much for joining us... Best of luck to youand your team. We're all counting on you. It's been a pleasure hopeto see you on our rocket ship error. Attracting today's new post traditional learners meansadopting new enrollment strategies. Helix educations data driven, enterprise wide approach toenrollment growth is uniquely helping colleges and universities thrive in this new education landscape,and Helix has just published the second edition of their enrollment growth playbook with fiftypercent brand new content on how institutions can solve today's most pressing enrollment growth challenges, downloaded today for free at Helix Educationcom. Playbook. You've been listening to enrollmentgrowth university from Helix Education. To ensure that you never miss an episode, subscribe to the shown itunes or your favorite podcast player. Thank you somuch for listening. Until next time,.

In-Stream Audio Search


Search across all episodes within this podcast

Episodes (231)