AI in Action | Active Learning with AI

We've known for a while that active learning works. Active learning creates durable learning. And when students have durable learning, they're able to utilize that knowledge, that content that they have in their head to solve really complex problems in creative ways. So active learning is incredibly important within the learning process. And now that we have artificial intelligence with GPTs and even agentic AI, we have the opportunity to create asynchronous active learning opportunities that help make that learning stick as well. So what is active learning? And active learning is really the interaction between the learner and the content. How they interact together. So when they actively engage with content that's an incredible piece because oftent times learning is a byproduct of our activity of our actions. U most of the time it's a byproduct. So learning is a byproduct of studying. So we can study to learn but now active learning gets us engaged with the particular content. And as we engage with this content, the reason why it makes such sticky learning, such durable learning, is because it brings our schema, our previous experience, and it engages with those as well. So we literally have a framework already set up, a mental framework within our minds that helps us absorb this type of learning. So, when we're actively engaged with the content in multiple different ways, I'm going to show you in just a moment how we do it at Los Angeles Pacific University because it's been a uh it's it's been it's been an evolution of the way we've integrated active learning within our asynchronous online courses. And that's why this is such a game changer because we can now have active learning within our asynchronous and online courses. how we utilized AI within our courses at LAPU transformed as we implemented. It's been a fantastic process and we started with Spark and Spark was there as an AI course assistant. It had all or it has we still use Spark. It is in every course and it is specifically trained on the content within that course. And what's interesting about that is that was the beginning. That's where we started with a pilot back in spring 2024. So we've implement been implementing for a ye over a year now. And we've kind of changed how we haven't changed spark, but we've added so we've used additive measures to get up to active learning within our asynchronous environment. So it started with spark and spark was a form or is still a form of active learning. Why do I say that? Because as earlier stated, Spark is trained with all the course material. So when a student says, "Hey, for assignments XYZ, it's asking me to do this. Can you explain it to me?" Or, "What is assignment whatever it might be?" Spark is specifically not allowed to create content for a student to be able to copy and paste. Spark is designed to be socratic to enter into a conversation. Oh, okay. So, I understand that you're not that uh the assignment seems a little bit confusing. What's one part that is? And it'll go in and it will explain in different language. It's been act very accurate every time, but it'll explain to the student in a slightly different way. So that's been one way where they've and then engage with that student back and forth because Spark will explain it in a way it'll say how do you think you can implement that within your assignment. The student will come back. So that's engaging with the context of the content right there. So that's really how we started and spark is able to you know if somebody asks a simple question what what when is assignment three due or whatever it may be it will give uh the dates and things when assignments are due. Uh but if they ask about specific things it enters into a socratic conversation that creates active learning right there. So we went from that and we went from a pilot to full implementation relatively quick quickly in a matter of weeks. So by uh summer 2024 we had full implementation. So the end of spring we knew we wanted to continue because the results we were getting. So boom we move into uh the full full implementation in in summer. And within that then we started thinking okay in our master of science and instructional design we wanted to implement an assistant an instructional design assistant and we did. So Desi is the name of the instructional design assistant and it acts like an assistant like somebody helping you design. You ask it questions, it responds. Ask you other questions and it still that socratic nature, but this time it's focused on the work and creating whatever you want to create within the uh within the course that you're in. So, Desi is specifically designed on instructional design. So, it is expands throughout the whole program. So there's one instructional design assistant for the entire program and it's trained on agile, atti, SAM, whatever it is, it is completely trained on that. So it can help the instructional design student gather their thoughts, put things together really well. So we went from Spark, which is individually trained on all course on individual course content to Desi that's now kind of like workflow sort of stuff. And now we in this year or this past term we implemented um uh think pair share where you think about a topic you pair up with the AI assistant and you discuss the topic. So given something to think about pair up with your AI assistant and give them and then in the discussion you come back now and you put what you learned in there. So we focus on the student voice. So now the discussion questions where we use think pair share are what we like to call AI resilient. So now students are getting more from it and they're able to use their own voice and describe what they did. So we've noticed that there's been a lot more engagement within the discussion forums um uh that have the think pair shares where they use AI and we also have AI assistants where especially in we started in psychology and history where they go in and have conversations like one assistant is named Miguel and Miguel is I want to say a six-year-old boy that witnessed a tragic accident his brother got hit by a car. They're in the hospital. Now, the therapist has to have a conversation with Miguel. And Miguel's trained to react uh in multiple different ways. And when it comes back, it states the uh physical actions and any any verbal actions that it has. So, it's been kind of amazing to be able to get people to utilize these skills in an asynchronous environment. And the way we did that was very uh purposeful. We started with Spark within every course as a course assistant. We said, "Hey, this is really powerful." But underlying Spark, so nectar.io is what we is who we partner with for all of our in all of our assistants within courses. So we found Spark was great for the course. I said, "Hey, this is Nectar's connected to an LLM and that's our ability to do that." So now we created Desi which was almost like a workflow help for instructional design. It's a literal instructional design assistant for the program. And then we said okay now now that we've done this and this we stack one more on top which is the active learn the definitive active learning where we use think pair share and we engage in different types of conversations with these AI assistants. And it's been kind of incredible because we found that when students engage with this, they increase their critical thinking within any range that they're using. So they know they can question the content that they're getting. So that's been a very powerful way of ethically using artificial intelligence within um courses and actually having students become more AI literate and be able to use it in ways that are very powerful. And that power is attainable in many different ways. Like I said, we use nectar.io because they're furpa compliant. Uh everything everything is in place. So we don't we we don't have any ethical issues with the use of uh AI as well. Uh now if you you create a GPT and it's not doing what you want it to do, it's probably not the GPT. probably not the large language model, it's the prompt that you wrote. So, as you're writing, as you're creating and things you want to do, you have to really embrace agile and iterate. Write something out, see how it works, test it, get the feedback from it, redo it. So, you're you're going to create, test, feedback, create, test, feedback, create, test, feedback, and it's just that constant loop. until you get it where it needs to be. Because if you're thinking you're going to create the perfect prompt and you spend a lot of time on that initial stage of writing, you're going to be behind because it's uh it's probably not going to get you what you want. You can use uh GPT to help you write the prompt and that's a great place to start and then test it, test it, test it. Um, one of the prompts we use for training that we have with for one of our teams is it engages in a conversation and as it's engaging in the conversation, it gives feedback to the individual that it's engaging with on how they're responding to the fictitious person. So, it'll respond to the participant and it'll also coach them and give them different ways on how they're responding. So, it's incredibly important that you first define what you want the AI assistant to do, what you want the GPT to do, and then begin prompt writing. And if it's not doing what you want, it's probably your prompt. And we've been finding some very interesting results with the use of our AI assistants. And first is better grade outcomes. And when we ran our pilot, we noticed there were better grade outcomes when we directly measured those who used Spark versus did not use Spark. So that was a big reason why we wanted to move to implementing Spark fully in every course quickly because we noticed better grade outcomes, increased sense of self-efficacy. Now, the self-efficacy and increased uh intrinsic motivation to learn are self-reported. The grades, of course, are not self-reported. they're the grades for the courses. Um, and what we found is that now on a larger scale, now we're able to look two, three, uh, terms down the road when we look at who's used Spark, has not used Spark within a course and their grade outcomes, it is still significantly different. those who use Spark have a significant statistically significant higher grade outcome with about a medium uh effect size compared to those who did not use Spark. And what's kind of interesting is Dr. Brewer, Professor Hunt, and Dr. Tong did a uh a study uh recently that's going to be up for um be up for publication in Christian Higher Education Journal. is that when we looked at great outcomes without Spark, grade outcomes with Spark, and then great outcomes with specific AI um active learning assistance, the grade outcome increased with each one. So those who used the uh active learning assistance, so like the think pair share and had the direct conversations where they were practicing different therapeutic techniques, motivational interviewing, whatever it may have been, they had increased great outcomes compared to those who use spark and compared to those who did not use spark at all. So and the effect size between each one was about a me it was small from spark to the authentic AI assistance but it was uh large if we look at LA or looked at uh the AIS assistant active learning assistant and not using uh any AI assistance at all the outcomes the effect size of the outcome was large and that's important because remember Spark does not create content for students. It engages in conversation with them about the content and lets them immerse and engage with the content. So that's kind of pretty pretty big deal. And then we also found that they had a greater ability to concretely explain how context works within a professional and personal context. So they they were able to those who utilized Spark, the AI assistant, were able to better articulate what they were learning. And we found that through a study that we did where some students uh self reported that they utilize Spark, didn't utilize Spark, and then they wrote uh end of open-ended end of course research or end of course survey responses. And within those responses, we did a semantic and thematic analysis and found that it was a very large it was a it was a medium to large effect size. The difference in the way people were able to communicate about the content and how it affected their individual lives or professional lives as well. So the use of the active learning and the AI assistance is having a large uh effect on students if you use it correctly. If you're just using like a copy and paste chat GPT will not have that effect size. But when you create the bot that lets them actively learn and engage with the content, these outcomes happen.