Explorations in Data and Machine Learning
Using the power of data organically generated by users of the Canvas ecosystem, you should be able to come up with new and exciting solutions for gaining insight into personalized learning. This talk will give insight into the possibilities of the Canvas environment combined with machine learning, NLP, AI. big data applications and nudging. Different examples of in-house and incubated startup ed-tech developed applications will show the power you can unleash based on the already available information.
(upbeat music plays) - Good day to you all. - Yeah, good day. - We're Eric Slaats and - Martin Ruissen - From Fontys University of Applied Science, Department of IT. - In Eindhoven. - In Eindhoven. - Yes - Yes.
And we're gonna tell you a story about explorations in data and machine learning. Which we have conducted with students and which will lead to new add-ons and startups for campus plugins. - Yeah. Hopefully we have a nice talk for you and we have some special guests later on in this talk. So, well here we're gonna start.
- Here we go. Well, let's give a small context description of what we are doing. At Fontys ICT, we promise our students they can become anything in IT. And we have about 5,000 students and every one of them gets a personalized curriculum. This personalized curriculum makes sure that they always on par and that they're doing the things they want to become.
So we are basically educating who they want to be. And we use canvas to make sure these personalized curricula fly. We did some canvas talks before about that one and the backgrounds. It can be found on the canvas con websites. What these personalized curricula do, is they generate personalized data on how a student student is performing, what his habits are, what his knowledge is and it's knowledge base but also how he does things.
And we are tapping into what we can do with this data at this moment. How do we do this? How do we deploy canvas? Just a small recap. We give them total ownership. In our curriculum, students have ownership of basically everything that has to be connected with their education. They data mine their own content.
They data mine, their own didactics and the criteria in which they're tested upon. So to make this fly, we had to get rid of something. We don't have any schedules anymore. We don't have tests anymore. We don't have classrooms.
We don't have classes. We don't have courses, but what do we do? We even give them total ownership in canvas. - So what we do, basically we give every student all kinds of rights in canvas. And that means that we have only, well basically, two courses in general, one guiding course which we use for mandatory things. But basically every student gets his or her own personal course.
And she is in charge or he's in charge of most of this personal course. That leads to a very rich learning environment. Where in the end, we have a personalized, validated portfolio because students can make their own assignments. They can make their own rubrics. They can make their own criteria.
They can invite who or whatever they think is applicable to their situation that they're working on. Then can, they can add their products. They can add their assignments. So basically, well the whole course is from, or is the student. So that leads to a lot of student owned assignments with what was already said.
Their own criteria, rubrics but also different iterations, different versions of those assignments, where you can look back at the comments, which were given by teachers, by tutors but also how it was graded. And we don't do any kind of testing. So it is not one grade. So these submissions can grow and grow over time with all that data, which is already gathered well basically by design through the canvas system, we are able in the end to do a thing that we call competent profiling that we can show how a student did perform but is also performing on the development on their skill sets, on their professional skill sets. And that will in the end give a rich dashboard and rich insight, okay.
On the development and growth, even we are then also able to give out personalized diplomas. So in our system, every student has its own course but also gets a personalized diploma. - We're not giving students back data about their development only, but also about how they do it. The process itself, we're gathering a lot of data with feed builds and together with this content data it gives a perfect picture of how a student is performing and what he's doing. And what we are basically wanting to do now is tap into that massive amount of data that a student is generating.
All these documents, all of his process data, how he is performing, et cetera. So wouldn't it be nice if we could use AI to cater that? So, for example, ask questions in the canvas course and the AI will answer them. So you can basically ask anything and see if the student complies with that, with what he has done. We also have something in our courses that students can create their own startup. And that's also curating content that blends in their personal course.
And we have startups doing all kinds of things like autonomous drones, et cetera, but also a rather exciting startup who's doing AI within EdTech, and basically they are creating stuff like that. Let me introduce to you the guys of Open Maze. - Hey, there, I'm Ruben. - I'm Niek. - And I'm Max.
- And together we are Open Maze. We are a student startup in the AI and EdTech. We are working on several concepts where we are deploying machine learning on these applications are applications that are deployed in canvas. I would love to tell you a little bit more about three concepts that we are working on right now and is currently in the testing phase. The first application I would like to tell a little bit more about, is the knowledge profiler.
We experience that students get more and more freedom in their education about what they learn. This means that students get very different skill sets compared to other students. So how do you get clear knowledge profile of your own education and share this with others? The knowledge profile uses artificial intelligence to analyze the submitted assignments and create a visualization of your own unique knowledge profile. You can see here, a generated knowledge profile of the course from Max. Every student has instant access to their own visualization of their knowledge profile, which is validated which they can use.
For example, for project group making, for company matching or for their own resume. Besides that, this concept can also be applied in other context, called the expert finder. Every student has their own skills. So you can also try to find each other and see who has experience in this field where I have a question about and expert finder helps you with finding the correct students that have experience in your location. - For the next application we want to talk about the Q app.
With the Q app, we want to redesign the way that you're going to be reading or grading your documents. What this application does, is it allows you to ask questions to handed in documents. You can, for example, grab a paper or a document about your topic. Let's say machine learning, you can then ask our Q app. Does this document contain info about machine learning? Does it answer questions X and Y? If it does, the bot will look it up and actually show you the results.
If it doesn't, it's not there. It's not present. This way, teachers can use it to more dynamically and interactively grade documents and students can use it to see whether or not their documents contains the essence that they want to convey. - The final two we are creating is about AI generated feedback and using canvas documents to get feedback on. Continuous feedback is so important for a student's learning journey, but it's hard for teachers to get feedback on time let alone multiple times.
We use state-of-the-art machine learning models to generate feedback on documents. There's no teacher configuration necessary and the student can use it whenever they want and how many times they want it. You can imagine that this also saves time for teachers. Our tool can give feedback on grammar but also on document structure and writing style. And if the research questions that are posed, if they are answered and how good they are.
This allows for the student to then improve that document before it goes to the teacher and then they can have a meaningful conversation about the essence of the document. Would you like more information or to get in touch? - Okay and there you have it. This is a really nice project. We always like to do stuff with students when we are developing new concepts, et cetera, and what these guys are doing, they are creating a startup and we are gonna use that product but we are also doing more. We have an ongoing project that is called Quantified Student, to make sure we use student data to improve the way they're learning.
For example, we're doing an experiment now with biometric data, et cetera, sleep data, to see if we can improve their studying and nudging, et cetera. And what these guys are building, fits perfectly in this picture. Well folks, that was it. Thank you very much for watching from Eric, - Martin - and - Open Maze.
And we're gonna tell you a story about explorations in data and machine learning. Which we have conducted with students and which will lead to new add-ons and startups for campus plugins. - Yeah. Hopefully we have a nice talk for you and we have some special guests later on in this talk. So, well here we're gonna start.
- Here we go. Well, let's give a small context description of what we are doing. At Fontys ICT, we promise our students they can become anything in IT. And we have about 5,000 students and every one of them gets a personalized curriculum. This personalized curriculum makes sure that they always on par and that they're doing the things they want to become.
So we are basically educating who they want to be. And we use canvas to make sure these personalized curricula fly. We did some canvas talks before about that one and the backgrounds. It can be found on the canvas con websites. What these personalized curricula do, is they generate personalized data on how a student student is performing, what his habits are, what his knowledge is and it's knowledge base but also how he does things.
And we are tapping into what we can do with this data at this moment. How do we do this? How do we deploy canvas? Just a small recap. We give them total ownership. In our curriculum, students have ownership of basically everything that has to be connected with their education. They data mine their own content.
They data mine, their own didactics and the criteria in which they're tested upon. So to make this fly, we had to get rid of something. We don't have any schedules anymore. We don't have tests anymore. We don't have classrooms.
We don't have classes. We don't have courses, but what do we do? We even give them total ownership in canvas. - So what we do, basically we give every student all kinds of rights in canvas. And that means that we have only, well basically, two courses in general, one guiding course which we use for mandatory things. But basically every student gets his or her own personal course.
And she is in charge or he's in charge of most of this personal course. That leads to a very rich learning environment. Where in the end, we have a personalized, validated portfolio because students can make their own assignments. They can make their own rubrics. They can make their own criteria.
They can invite who or whatever they think is applicable to their situation that they're working on. Then can, they can add their products. They can add their assignments. So basically, well the whole course is from, or is the student. So that leads to a lot of student owned assignments with what was already said.
Their own criteria, rubrics but also different iterations, different versions of those assignments, where you can look back at the comments, which were given by teachers, by tutors but also how it was graded. And we don't do any kind of testing. So it is not one grade. So these submissions can grow and grow over time with all that data, which is already gathered well basically by design through the canvas system, we are able in the end to do a thing that we call competent profiling that we can show how a student did perform but is also performing on the development on their skill sets, on their professional skill sets. And that will in the end give a rich dashboard and rich insight, okay.
On the development and growth, even we are then also able to give out personalized diplomas. So in our system, every student has its own course but also gets a personalized diploma. - We're not giving students back data about their development only, but also about how they do it. The process itself, we're gathering a lot of data with feed builds and together with this content data it gives a perfect picture of how a student is performing and what he's doing. And what we are basically wanting to do now is tap into that massive amount of data that a student is generating.
All these documents, all of his process data, how he is performing, et cetera. So wouldn't it be nice if we could use AI to cater that? So, for example, ask questions in the canvas course and the AI will answer them. So you can basically ask anything and see if the student complies with that, with what he has done. We also have something in our courses that students can create their own startup. And that's also curating content that blends in their personal course.
And we have startups doing all kinds of things like autonomous drones, et cetera, but also a rather exciting startup who's doing AI within EdTech, and basically they are creating stuff like that. Let me introduce to you the guys of Open Maze. - Hey, there, I'm Ruben. - I'm Niek. - And I'm Max.
- And together we are Open Maze. We are a student startup in the AI and EdTech. We are working on several concepts where we are deploying machine learning on these applications are applications that are deployed in canvas. I would love to tell you a little bit more about three concepts that we are working on right now and is currently in the testing phase. The first application I would like to tell a little bit more about, is the knowledge profiler.
We experience that students get more and more freedom in their education about what they learn. This means that students get very different skill sets compared to other students. So how do you get clear knowledge profile of your own education and share this with others? The knowledge profile uses artificial intelligence to analyze the submitted assignments and create a visualization of your own unique knowledge profile. You can see here, a generated knowledge profile of the course from Max. Every student has instant access to their own visualization of their knowledge profile, which is validated which they can use.
For example, for project group making, for company matching or for their own resume. Besides that, this concept can also be applied in other context, called the expert finder. Every student has their own skills. So you can also try to find each other and see who has experience in this field where I have a question about and expert finder helps you with finding the correct students that have experience in your location. - For the next application we want to talk about the Q app.
With the Q app, we want to redesign the way that you're going to be reading or grading your documents. What this application does, is it allows you to ask questions to handed in documents. You can, for example, grab a paper or a document about your topic. Let's say machine learning, you can then ask our Q app. Does this document contain info about machine learning? Does it answer questions X and Y? If it does, the bot will look it up and actually show you the results.
If it doesn't, it's not there. It's not present. This way, teachers can use it to more dynamically and interactively grade documents and students can use it to see whether or not their documents contains the essence that they want to convey. - The final two we are creating is about AI generated feedback and using canvas documents to get feedback on. Continuous feedback is so important for a student's learning journey, but it's hard for teachers to get feedback on time let alone multiple times.
We use state-of-the-art machine learning models to generate feedback on documents. There's no teacher configuration necessary and the student can use it whenever they want and how many times they want it. You can imagine that this also saves time for teachers. Our tool can give feedback on grammar but also on document structure and writing style. And if the research questions that are posed, if they are answered and how good they are.
This allows for the student to then improve that document before it goes to the teacher and then they can have a meaningful conversation about the essence of the document. Would you like more information or to get in touch? - Okay and there you have it. This is a really nice project. We always like to do stuff with students when we are developing new concepts, et cetera, and what these guys are doing, they are creating a startup and we are gonna use that product but we are also doing more. We have an ongoing project that is called Quantified Student, to make sure we use student data to improve the way they're learning.
For example, we're doing an experiment now with biometric data, et cetera, sleep data, to see if we can improve their studying and nudging, et cetera. And what these guys are building, fits perfectly in this picture. Well folks, that was it. Thank you very much for watching from Eric, - Martin - and - Open Maze.