AI-Empowered Instruction at Duke University

ABOUT THIS EPISODE

Jon Reifschneider, Director of Masters Studies of the Artificial Intelligence for Product Innovation at Duke University joined the podcast to discuss how their artificial intelligence assistant empowers faculty to provide more personalized instruction.

I started using the tool also on an individual student basis to look at what I call it a mastery map for each students. So that's basically so of a map of all the topics were covering in the class and which ones they have reached the level of mastery on. You're listening to enrollment growth university from Helix Education, the best professional development podcast for higher education leaders looking to grow enrollment at their college or university. Whether you're looking for fresh enrollment growth techniques and strategies or tools and resources, you've come to the right place. Let's get into the show. Welcome back to enrollment growth university, a proud member of the connect Evu podcast network. I'm Eric Olsen with Helix Education and we're here today with John Reif Schneider, director of Master Studies of the artificial intelligence for product innovation at Duke University. John, welcome to the show. Thanks, sir. Great to be here. So excited to talk with you today about how artificial intelligence can help empower faculty to provide more personalized instruction. Before we dig into that, can you give the listeners a little bit of background on both Duke and your role there? Sure so. I'm in the faculty at the Engineering School at Duke and my main job there is to run our new program in artificial intelligence. So we have a graduate level program that's designed for early career folks from engineering or science back grounds who want to learn about machine learning in AI and specifically how I can apply and in various industries. And so I run that program is as well as teaching it. Love it. John, to help set up this conversation today, can you highlight this common problem faculty have of being able to broadly understand where their students are academically based on test scores, but not always understanding where a single student specific competency gaps are? Sure. So a little bit the background on this problem. When I first started teaching and when I joined the fact with the Duke,...

...one of the surprising things to me, and it was also a source of frustration, was that I had a really difficult time knowing whether my students were were understanding what I was trying to teach them. And it was only once we would get to assessments and I was able to assess that through a homework or quiz or test and look at the results was I able to determine that. But the challenge of determining that, even from assessments, was that the tools we as instructors have our disposable, like our learning management system, will is our general designed to just give you aggregate scores on things. So homework, one student got an eighty five, quiz one they got a ninety two, but that doesn't really give you insight into what a student is understanding and where they're struggling. To really figure that out you have to dive much deeper or really engage the student in discussion to figure that out. And so as I start teaching, I really struggled with this. But I also realized that now, more than ever, we have such a vast amount of data at our disposal about our students through online homeworks. Everything these days is online lectures, quizzes and tests, etcetera, and I started looking at how can I use all of this data that have available to give me, as an instructor, more insight into both where my class and aggregate was was understanding things well and was struggling, but also where each individual student was under standing things well and and struggling, so that I could provide a much more personalized experience and more personalized guidance for all my students. Yeah, so in terms of digging d print the data, can you give us this high level overview of the intelligent classroom assistant that you and your colleagues designed to help solve for this challenge? Sure. So I started building a tool. The genesis of it really was last summer when we started prototyping things, and we call it the intelligent...

...classroom assistant and what it does is it is uses all of the data that we have available to us as instructors. Again, whether that's, in my case, for the classes I teach, it's weekly formative assessments, simple low stakes quizzes that we do. It to weekly homework assignments, it's our weekly lectures, it's any test that we have during the class. We take all that data into the tool and and what the tool fundamentally does is it maps everything we do in the class across all of those things back to a set of learning objectives or topics that the instructor has defined for what they want the students to get out of the class, and by doing that we're then able to evaluate how students are doing on each of the topics that they're supposed to be understanding. So now we can start to see things on a sort of a topic or learning objective level rather than just looking at aggregate scores which are really difficult to distill an any sort of meaning from. So finally, that's that's what it does, and the reason for doing that is to provide that air insight to me as an instructor in terms of how it classes doing on each of the topics and where I need to spend my time, and then also to give students that insight for themselves so they can see individually how well they're doing and where they're struggling on certain topics in the class. Yeah, now let's talk some use cases here. Can you give us an example of how this AI assistant would inform your classroom teaching in front of a larger group? Sure, I am highly adaptive in my teaching style in that each week I start preparing for the following week and when I start my preparation I spend time in my tool, the intelligent classroom assistant, and I look at the aggregate picture as well as pictures of individual students and and that helps me identify a couple of things. First of all, it helps me identify certain topics where students are struggling and then, secondly, I can actually dive one layer...

...deeper and then identify individual so the subtopics or questions where students are struggling. And the way I utilize that then, as I prepare for the following week is in two ways. One, I will make sure that we spend class time working through that topic or particular aspects of that topic that that my software has helped me identify as areas of weakness for a broad portion of the class. So I make sure we spent on reinforcing those things when we're together in class time. And then, secondly, I'll continue to bring those things back in new ways through assessment again, whether that shows up later and homework or shows up later through additional questions on quizzes, so that I can see, as I'm dedicating more time to that topic and as I'm flagging it for students, or maybe they're in the tool faking it for themselves and identifying it as a weakness and then spending additional study time on the topic. By bringing it back over and over again through the following weeks, I can then start to see our students progressing in their knowledge of that topic or they still struggling, in which case I need to continue to make additional adjustments and spend more time with it. I love that. I love that. And how about the more personalized use case? How can this AI assistant better inform your one on one help sessions with individual students? Yeah, so this was a another bit of frustration when I early in my teaching days that I would have students show up to my office hours or to one on one sessions for help and I wouldn't really know where the students needed to help. I'd have to ask them and I have to hope that they were self aware enough to understand that for themselves. So I started using the tool also on an individual student basis to look at what I call it a master map for ECH students. So that's basically sort of a map of all the topics were covering in the class and which ones they have reached the level of mastery on and which ones they have have knowledge gaps and are struggling on. And by spending time on that in...

...advance of meeting one on one with a student, I can show up to that discussion of the student much more prepared and we can talk through together the areas that have been flagged as areas of weakness and then I can I can provide them a much more personalized and invaluable experience when we get together. Is there in the future, potentially a student facing use case for this AI assistant, helping students to see for themselves and better understand their own learning analytics? HMM, absolutely so. One of the things I'm prototyping right now that I build over the holidays is a student facing aspect of the tools. So our tool now lives on a public url, intelligent classroom dot org. My students in my spring class are now able to go to the website, log into their class us with with their individual credentials, and then view their own data, so they can see their own mastery map of how well they're doing to date on each of the topics we've covered in the class. They can dig in on any given topic and see how their performance on that topic has been trending over time as they've gone through the weeks in the class. And I'm also able to provide them personalized help recommendations, and the way I do that is for questions that they have gotten incorrect on the homework or the quizzes. I use my machine learning algorithm to analyze all of the lecture transcripts. So that's one of the great things about teaching online, is that we're able to all their transcribe all of our lectures. So I can use my algorithm to actually direct the student too exactly which lecture and and specifically where in the lecture they should go for help on whatever particular topic they have a weakness, and I can do that automatically without me having to spend hours and hours going through and figuring all that out. So so we're piloting that now in my introduction the machine learning class that I'm teaching the spring and it's a good, good...

...response from the students so far. It's still pretty early in the semester, so they'll be very interesting to see as a semester goes, how valuable it is for the students. That is such an exciting one. I think of of a just a very simplistic use case with dual lingo. It knows that I'm a four out of five, you know, with knowing how to give someone to actions to the discotheque, but that's all I see. As well it's like, what am I really misunderstanding here? My understanding, you know, conjugates verbs like like what is the thing that I'm missing? And you're solving the you're creating a situation here where both the faculty and the student would both have a broader awareness of where those those death its. It's are super exciting, right John. For anyone listening with concerns about the integration of teaching and artificial intelligence, any response to those concerned that AIS could end up outsourcing teaching all together? Yeah, that's an interesting topic and I would say there's a number of websites and software out there that provide sort of automated intelligence to theres and that type of thing for students that they can use outside of the setting of any type of normal classroom or or engagement with human instructors. I really focus my research and the development of my software is is designed to support the classroom experience and specifically to support instructors to be able to provide a better experience for their students and for the students to have a better experience learning in the context of a class. It's not designed to to replace the instructor but really to help the instructor deliver a better experience but also help the instructed to scale. It's a wellest I wish, problem in education that there's a sort of, at least what traditionally considered to be an inverse relationship between scale and quality of the educational experience, and that's one of the things that makes a text so so challenging is that inverse relationship. And we know that there's there's incredible value to great teachers, but when a great teacher can only teach ten...

...or twenty students at a time, we're really not able to get a lot of leverage out of that particular teacher skill. So part of this is thinking about how can we take that same great teacher and we can enable that teacher to give world class quality educational experience to fifty or maybe a hundred students at a time versus ten or twenty. So I'm starting to look at ways that we can enable that and again it is all designed to support great teachers and what they because I am myself a teacher and I firmly believe in the value of the human interaction between instructors and learners. I think that's not something we should seek to replace, but we should look at ways to scale it and and to increase the value of the experience for learners. It is incredibly, incredibly exciting work that you're doing, John. Any Next Steps, advice for institutions listening looking to leverage or look into leveraging AI empowered instruction? Where should they start or where should they look first? Sure in my opinion the first place is to look are actually don't really have anything to do with Ai. There's two so the prerequisites in my mind to be able to use a tool like this. The first one is good teaching practices. So you know specifically are we using formative assessment in the classroom to create this feedback, a loop between between learners and instructors? Because without any type of formative assessment, whether that's low stakes quizzes or questions in class or classroom discussion or homework or something else, without that feedback loop, it's impossible to know what's in a student's brain and until we figure that out there's really nothing we can do with any type of analytics. And it's surprising how you know how much that varies between instructor and instructor. So following best practices in pedagogy is a key so the prerequisite here, and then the second one is making sure we're capturing all that data in some way that it's available, so through a learning management system or...

...through some type of system, that we have data available to instructors to be able to use. Once you've done those two things, then then you're you're in a spot where you're able to start to apply some machine learning or other types of analytics to extract value from from the data that you've got. And then you have options. You can build things yourself, as I did. There's, you know, various commercial software out there that's either available or in the process of being developed, as book of the number of at tech companies and a lot of folks are out looking at how we can accomplish things like this. So there are a lot of things coming on the market and there will be commercial options available to I'm sure, but again, the immediate focus is on those two things I mentioned as prerequisites before you even start thinking about what the right you know, tool is. Superread advice. John. Thank you so much for your time today. What's the best place for listeners to connect with you if they have any follow up questions? Sure feel free to contact me. Probably the simplest thing is that I have a personal website. It's John Reichnither dicom's about Jo N R E I F SC H and eidercom, and that is my email linkedin excether on there. So feel free to reach out by any of those methods and happy to talk. Awesome. Thanks so much for joining us today, John. Thank you. It's been my pleasure. Attracting today's new post traditional learners means adopting new enrollment strategies. Helix educations data driven, enterprise wide approach to enrollment 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 fifty percent brand new content on how institutions can solve today's most pressing enrollment growth challenges. Downloaded today for free at Helix Educationcom. Slash playbook. You've been listening to enrollment growth university from Helix Education.

To ensure that you never miss an episode, subscribe to the shown itunes or your favorite podcast player. Thank you so much for listening. Until next time,.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (259)