Generative AI tools have made it easy to create custom chatbots. In this episode, we discuss the process of creating chatbots and some potential uses in higher education. Thanks for joining us for Tea for Teaching, an informal discussion of innovative and effective practices in teaching and learning. This podcast series is hosted by
John Kane, an economist... ...and Rebecca Mushtare, a graphic designer... ...and features guests doing important research and advocacy work to make higher education more inclusive and supportive of all learners. Our guest today is JeVaughn Lancaster. She is the Manager of Instructional Design at Lynn University. JeVaughn’s work using generative AI to support effective course design resulted in a Best in Track award at the 2024 OLC Innovate
Conference. Some of her more recent work on this topic was presented at the OLC Accelerate Conference in November 2024. Welcome, JeVaughn. Hi. Thank you for having me. I'm so excited to be here. It's good to see you. I saw you at your session, which is what resulted in this invitation. So we're really glad to have you join us. Today's teas are:... JeVaughn, are you drinking any tea? I actually do have tea. It's been my heart and
soul for the past couple weeks, I should say. It is a ginger honey tea, but I added a splash of soursop. I don't know how to explain what a soursop is. So you can drink just that one. …sounds good. I have a London Strand breakfast. And today I have a Darjeeling tea. So, today we invited you here to discuss your work developing GPTs to serve as chatbots that support faculty. Can you start first by describing what a custom GPT is, or maybe even what a GPT is?
Oh, yes, that is, I feel, the perfect way to start before we talk a little bit deeper in what a custom GPT is. GPT really stands for generative pre-trained transformers, and a lot of people really get that confused. But when we say generative, that's exactly what it does. It generates content based on a multiple of things, like patterns, responses to questions. A lot of people know ChatGPT as the essay writer, lo and behold, but that's really generative. So
it generates things. Then when it gets into the pre-trained aspect of it, it is trained by a huge data set of information that people feed into the GPT. So every time somebody enters a prompt, it continuously trains it and builds that data set to be a little bit bigger. And then the transformer part. I like to always say that the transformer part is like a multitasker, so it puts all those little data bytes and pieces together to analyze it. So it's almost like using a learned language.
It identifies a variety of patterns, and again, like I said, it's a multitasker. So it brings in all of these bits and pieces of information and then generates the response to your prompt, making it a GPT. So with that said, a custom GPT is really dedicated to a specific task of something that you do often on a daily basis, that you just want to make your life a little bit easier so you want to customize something for your immediate needs. So I really think that's
what a custom GPT could be, in a nutshell. So in order to create one of these custom GPTs, you do need a paid chatGPT account, and initially to use one, you needed a paid GPT account, but that's changed. Are there any restrictions on the use of GPTs that others create for free users? Yeah, you're right. So before, when someone created a GPT, in order to share that,
the end user would also have to have that paid subscription. But now with the updates, it's so easy, because what happens is that the paid user can create a GPT and then share that and as long as you have an account set up, a username and password, it doesn't necessarily have to be paid,
you're able to then use that customized GPT. So that means that instructors can create GPTs for their students, and people in professional development roles can create GPTs to share with their faculty, which is what you were presenting on there. Exactly, and I feel that it makes everything so much easier, because you have those that are really into it, and they're creating all different types of customized GPTs on so many
different topics. And then those that may not be as great with technology that they're not really able to put these things together. However, it's so easy to use by adding in those prompts, it's already built and customized for whatever it is that they need, so it really makes life a little bit easier. Now, there are a couple things that a free subscription might struggle with,
even when using a custom GPT, and there are a few restrictions that they have. So, for example, tokens, they won't be able to submit as many prompts as someone else with a paid subscription. Or if there are bandwidth issues, if there's a lot of people using ChatGPT in one instance, someone with a free account may struggle. They may have a lot of slow responses to their GPT. At OLC Accelerate, you provided a presentation on creating GPTs? Could
you describe the process of creating a GPT? Absolutely. Oh, my goodness, that's the fun part. And I think a lot of people feel that it's a lot more difficult than it really is to do. It's so simple. So when you're logged into your ChatGPT account, there's a little button on the left side that says, “Explore GPTs.” Once you click on that, it'll open up this toolbar, and on the right side, you click on it, and it says “Create.” And it has a list of prompts that you can follow along
to really help you build your GPT. It's going to ask a description. So there's where you say this is what this custom GPT is going to do. You can keep it short, sweet and to the point, because it is what it is, it's the description. And then after that, it's going to ask you some instructions. So what are you going to instruct that GPT to do? And this part is really important,
because it's almost like guiding the GPT on what to do. You're training it. And just like any other GPT model, you have to train it and you have to fine tune it and customize it to fulfill your needs. So because of that, you want to add in things like, if you're building rubrics, what are those specific criteria that you want if you ask it for a rubric, if you're using it for grading, you're going to want to have to put in those specific details: “Well, I don't want my feedback
to be more than 100 words,” or “I want it to be in a nice tone.” You really have to give it these specific instructions so that your GPT will give you the output that you're really looking for. So after you do the instructions, there's a section called knowledge, and that part is really important, because there's where you can do things like upload specific documentations that you want
your GPT to pull from, and that'll give you more straightforward responses. So rather than ChatGPT searching random articles or information that it finds on the internet, it's going to first go by the knowledge you provided. It's all part of the GPT training. So for example, at OLC, we created a GPT that was almost like a teacher's assistant. And it was important because we found that a
lot of faculty members struggle with learning objectives and how to create it and align it. So with that, we took the Bloom's Taxonomy chart, and there's some people that still like the old Bloom, some folks like the new Bloom’s. Then there's the digital Blooms. So to make everyone happy, we uploaded all three , so there's no discrepancy at all. And then there's many specific things that you can also add in. So if you're looking for things that have to do with universal design
for learning, you can link to the UDL page from CAST. There's so many things that you can do so it can give you that information that you want when you ask it as a prompt. So all of that is going to be built out into the knowledge section. So you're building something that you can utilize to help you with your information. If you are conducting research, or if you're using it to help you with research, you can link specific databases that you would use
so you find those peer- reviewed articles. There's so many things that you can do to train it, and then, hey, if you find out something's not working, you can edit it and fix it and go back. So after you finish adding your description and giving it your instructions, you can then also give it some talking points. You know, when you log on to ChatGPT, sometimes it'll give you
four different options. You can add in question prompts. You can do that, and then you can say, “Okay, now that we're done, what are we going to name you?” And it'll offer you suggestions for a name, it'll offer you suggestions for an icon image, and you can generate it, save it, and you have your custom GPT. Again, it's a lot easier to accomplish than it seems. I know it sounds so intimidating at first, like, make a custom
GPT is like,” Ah, I'm not qualified.” Yeah, it does. It does. And then when you get into it, it's just like, “Wow, I really just did that all by myself,” and you can customize it down to how it's responding to you. I have a friend and a colleague that has it sound like a Southern belle every time it responds to them. It's amazing, the things that you can do with your GPTs. You can really customize it and personalize it. And it's really no more difficult than just
interacting with ChatGPT in other ways. When would you use a custom GPT rather than just interacting directly with ChatGPT? I think if you want specific tasks done, because there's so many times when you are working with just regular ChatGPT, you have to prompt after
prompt after prompt until you really get the output that you want. If it's a task that you're doing 10 times a day, it'll be more stressful sometimes, I think, in my opinion, to continuously prompt to get things down to right where you want it, rather than have this custom GPT that I can upload an assignment that I just created and say, “Hey, can I have an analytical rubric off of this and it'll just give me the output that I need immediately with one question.
Are these custom GPTs less subject to hallucinations? I think so. I think that they really are, because once you feed it that information, and once you feed it that knowledge, it's always going to pull from there first. Now, there's something that I have noticed, when a lot of faculty members try to look for content, it'll definitely hallucinate a lot of links. So a lot of the links are always, I feel, fictitious.
You would click on it, you’d get the 404 code. You'll get “this video no longer exists. However, if you actually search for that author, you'll find it, 9 times out of 10 you'll find it. So I feel that it's definitely reduced the amount of hallucinations. And then in your custom GPT also, if there's a database that you frequent that some people will link it to Google scholar or something
that may not be behind a paywall, you'll be able to go right to it as well. Or I've even suggested that with those links and titles that the GPTs might give you, log into your institution's library and see, 9 times out of 10, you might find it there too. So yeah, in my use I've
noticed that it reduces hallucinations. Going back to your earlier response, where you suggested that it's good for things that you might do repeatedly, you've also suggested that it could be good if you want to create something that you share with others. So could you talk a little bit more about how faculty have responded to the GPTs you created to assist faculty? Have many faculty been using it, and how have they reacted to it?
Oh, my goodness, my faculty absolutely love it. They love it. And then also they've had suggestions to take it to the next level. It's incredible. So first and foremost, I mentioned those pesky learning objectives. For whatever reason, it's hard for a faculty member to think about alignment in the sense of using those specific words. They're definitely experts in their own areas and maybe teaching and pedagogy is not one of them, so sometimes it could be a little
bit more difficult. So one of the main things that all the faculty love is those learning objectives, because they can say, “Hey, I need this aligned properly.” Now, one thing that I did receive was that some faculty members noticed that when I first created it, they noticed that it wasn't aligning to the correct course level. So maybe a 100-level class was getting “create” or something in the higher levels of objectives. So I had to go back in, I refined the GPT and say, “Hey,
if a faculty member asked for 100-level, I want it to align with these levels of Bloom’s. And if it's a 400 level, I wanted to align with these levels of Bloom’s.” So then once you regenerate it and you save it now, hey, as soon as somebody reopens the link, they're getting all of those updates in real time, so now they're able to help with their alignment a little bit better. So their suggestions really come in handy. I appreciate it. I listened very carefully at those.
Can you talk about some productive uses for GPTs to support instructors and students? Yeah, definitely. So for faculty members, I already mentioned those rubrics, there's rubrics, and that's such a time saver right there, because I know that that's really my pain point when it comes to course development. So yeah, there's a lot of things that faculty members and students can use customized GPTs for. So, for example, I've seen a lot of faculty members use it to
create lesson plans. We have two and a half hour block scheduling, and there's many times where some faculty members teaching four classes in a block, it's tough sometimes for them to make sure that they have their daily lesson plans. They create timed daily lesson plans to keep them on track for their two and a half hour block. Quizze, is straightforward. You can upload a PDF chapter and say, “Hey, can you create me a quiz with 20 questions?” And you'll be able to get those and
of course, you'll have to tweak them a little bit. I would never say, to take whatever it is that a regular GPT or a customized GPT, and just copy and paste and put it in your class. We don't want to do that. As educators, we always want to fact check anything that any type of AI gives us. But again, it's a time saver. You want to help save your time and your brain power to do tasks
that will more benefit the students than staying and creating rubrics. Discussion questions is a really big use case, and I've seen faculty members do this for both asynchronous and face-to-face classes. For the asynchronous that's really self explanatory, but for the face-to-face classes, I've seen it really help the discussions flow nicer in the classroom, because now, if the faculty member is discussing an article or a chapter, they take that same article, run
it through ChatGPT, and they ask for discussion questions. And it really helps enhance the course, because they get brand new course discussion questions every day, or if they're teaching it multiple times throughout that block, each class has a different question and a different discussion, which really, really helps. Feedback and grading, I already talked about that. A big
use case is accessibility needs. Now accessibility is growing, and UDL, we preach all the time, and there's many times where you just can't figure out how to make an assignment accessible, but you can get those suggestions from ChatGPT as well. Now, on the student side of it, it's
awesome because students can use it as tutors. And I know that I was telling them at OLC that I created a custom GPT for my son in math, and it's just an adaptive learning so it'll provide, depending on how you create it, the one that I have will provide multiple questions, and he'll have to answer it, and if he gets it wrong, it's adaptive. So it'll break the question apart and say, you got this part correct, this part correct. But let's focus more on this step,
and it'll take him on a step-by-step journey. And then after he completes 20 questions, he gets a report, a spreadsheet saying all the ones he got correct, all the ones he got incorrect, and it showed him the steps of where. So really good for student tutoring. Another thing is research help. It definitely is great for students. I'm a dissertation chair myself,
and I help the students use AI to leverage their work. So I just had a student that had Zoom focus groups and interviews, so the ChatGPT was able to transcribe the audio, it was able to create pseudonyms for each attendee, so it'll help the privacy, and it was able to pull out themes, so now the student is able to spend that brain power on other parts of the research and that content really and not spend hours of time trying to transcribe items. It's really interesting,
and I feel that this is just the tip of the iceberg. There's so many things, so many different use cases that all of these GPT models can do. In terms of creating a tutor, one of the first things I did after seeing your session was I created a tutorial for my intro micro
economics course, because the book I'm using is tied to an OER textbook. So I could upload the whole textbook to it and ask them to use it, and also have it ask students initially, what they know about the topic, ask them what they have questions about and then have them ask their question, but then guide them to an answer, rather than directly answering the question. And it works remarkably well. And because it's trained just on the textbook that I'm using,
I haven't seen any substantial hallucinations. And so far, it's been working well. It's incredible when you're able to really dive so deep, to be so specific, and accomplish those
things with AI. And I mean, I can even say after one of the faculty members created a GPT focusing only on assignments, so he has the GPT assign students, I think it was like six different research papers, and what happens is that everything is uploaded into that knowledge base, and then from levels easy, mid, difficult, by difficulty, and then once the student enters their name, it will never give the students the same assignment twice, and it will always give
the student one of each level of difficulty depending on what week of instruction they're in. It's amazing how far and how creative everyone has become customizing these GPTs. Should we be careful in uploading information about our students in terms of personal information? Might perhaps we want to use pseudonyms or something else for students so that they're not being trained on information specific to individual students?
Absolutely. Yeah, data privacy is really, really important, especially when we're talking about students, because we have to be so careful. Remember, anything, especially if you're not in a paid account that you upload to ChatGPT is then used to train future models and versions of it. However, what I found out is that according to ChatGPT now, both the free version and the paid
version allows you to say if you're okay with sharing your data or not. On the Settings icon, there's a data control button, and one of the options under data control is to improve the model for everyone. Now, if that is turned on, they're taking all of the data that you put in and using that to train but if you turn it off, it won't share your data, and it'll delete those
conversations, so it won't share anything. Now is that 100%? I don't know. Who knows? Personally, I don't 100% trust it, so I don't feel like I would ever upload actual student data, God forbid, financials or anything like that, but I would still be careful on it, but that's just an added step of security to turn that off and make sure that you're not sharing or you're not giving them the right to use whatever information that you upload.
It also seems like if we're having students using GPTs in our class, that talking about data privacy and things might be a topic that we might want to cover with our students. You want to talk a little bit about ways to maybe do that with students? Absolutely, and I think that really goes with
the whole topic of AI literacy. I always recommend to faculty members that, hey, number one, if you're going through some of these information sessions and you're learning how to use AI in your classroom and you're using AI to create those discussion questions, let's teach the students the proper way of also using AI ethically in their academic work. So I think that a conversation that should be had. I'm sure that high schoolers are already having
those conversations. I know that my son, who's in 11th grade, definitely is with his teachers, but I think that that's a conversation not only to have right on entry when the students are entering the institutions, but constantly, it should be everywhere. It should be in the syllabus. It should be in that very first class of every semester. You want to make sure that students are aware that whatever it is they upload is fair game unless they protect themselves. I mean, we had to
have those conversations with social media also. So I feel like this is just the next layer. What's your advice to instructors about disclosing their own use of AI to their students? I think it's important, because I feel that students are really less motivated to use it for non practical purposes. If they know that their instructor is using it as well, they know that their instructor is well versed in the use of AI, they'll be careful, and they'll be very focused
on their content, rather than putting in a prompt and submitting it. My advice to faculty members is always, “Hey, you can go ahead and disclose that you're using AI, but also give your students the opportunity to focus their brain power on what you feel is more important.” So my example is that, for me, I'm all about the content. I want to make sure that my students know the content, forward, backward with their eyes closed. I want them to know that, however, if there's a presentation,
go ahead and use AI to create that PowerPoint. I can care less as long as you know that content, you can focus that brain power on solidifying that content and using AI as an assistant for you. And I feel like that would be one of the better use cases, or the better discussion that a faculty member can have with their students rather than saying, edit your paper using ChatGPT. One concern that a lot of faculty members have is that students may be using this as a substitute
for learning instead of as an assistant for learning. What can we do to encourage students to use AI to facilitate their learning and not to short circuit that learning?
Yeah, I think that it will really enhance the overall conversation and relationship between the student and the faculty member, because now the student is able to go off on their own and ask all of these questions of inquiry to their GPTs and then come back to their faculty member, or their professor, and say, “Hey, so I was thinking about x, y, z, and this is what I found.” So now the faculty member would have the responsibility of saying, Oh, my goodness, this is
wonderful. Let's discuss it more, and it'll help the student boost his confidence, or the faculty member can say “Mmmm, you don't have that quite right. Where did you find your information? Let's figure this out now together that you've already went out on your own to try to ask the question.” And I feel that these days, a lot of students are lacking that wow moment in their educational career. I think most of us just had a wow moment that's memorable. It's always important to make
some type of learning memorable, and I think a lot of students are losing that. And I think a lot of students, and maybe even faculty members, are losing that even more with all of these AI, but if we have the ability to let the students go fish and learn on their own without having that negative point against them saying, oh my goodness, I'm using AI, am I a terrible person?
Rather than helping the students utilize it the right way. That might just be their moment of making that educational moment memorable for the student, because you're now on their level in how you can adapt to their educational needs. And I think that's really important. And our students come in with really diverse prior learning. And one benefit of AI is it provides an opportunity to bring everybody up to a higher level, because it can address their
specific needs. It's available 24 by seven, and it allows students to get help when they need it, where they need it, on the specific concepts and topic they need, as long as they're not
doing that as an alternative to actually learning. So with every aspect of AI, there's some really strong positives, but there's also some really serious concerns, and I think we all have a lot to learn about how to use this effectively, but for me, at least, it strikes me as being really silly to try to ignore AI and assume that this is not an issue we need to deal with, because students are going to be working with AI in some form for the rest of their lives, and not
preparing them for that world strikes me as being not particularly helpful for our students. Yeah, I agree. At this point, we've all realized that it's not going anywhere, it's going to evolve and move on to so many different things. So we have to be adaptable to technology. AI, it's updating as we speak. It's learning more and more as we speak.
And I think that one of the things that really strikes me is that you were pointing out motivating curiosity, and when someone is indicating that they were curious, to not smash it, regardless of whether or not their curiosity led to a correct answer. And I think one of the things that I've experienced with students, even prior to the surge of AI is that sometimes when students are asking questions, even when they were just Googling answers, is the way that they were
asking the question was where their misconception lies. And so sometimes having inquiries about how they were asking the question, or how they were using AI to answer their own questions could be an interesting way to engage with students, and opening the GPT and exploring with them
could be a great exercise to do with students. Absolutely and having these types of discussions with our librarians, I found out, and I didn't know this until maybe right before that conference in November, that actually, when you're working with APA and student writing, it's an APA requirement that the student attach the entire conversation with the GPT as an appendix. And that part I didn't realize. So if the student puts in their paper and say, I want this. I want that,
they have to include every word of it so they can't delete those conversations. I was floored by that. I had no idea. I'm like, well, that's interesting. I never knew that APA required that, but according to our librarians, they do, But also not a bad thing for faculty to request. Sometimes, if we're asking students to use AI or use a GPT as a way to get to an answer to use as a learning tool to show their work, it's something they can save.
It's true, and most institutions use tools like Grammarly, and I know that a lot of our faculty members require students to check their work through Grammarly to make sure that the grammar and punctuation. But then there's the new Grammarly, where it does just that. It records every keystroke that the students are typing to say, “the students typed this,” this part came from AI, this part was copied and pasted from this website, and all of the rest of it seems
to be human text. And then the students can create a share link and share it with their instructor, and you see it in real time as the students are drafting their papers. It's amazing. I mean, AI has come such a long way from where most people think it began two years ago.
Feels a little creepy, but okay. AI has been going on for decades, and it was always something that was going to happen soon, and then all of a sudden, it arrived, way before anyone expected anything this powerful to be released with an interface so friendly. Yeah, no warning, just boom, here we are. …right before finals. Yes, yeah. Alright, so I always wrap up by asking, what's next? As far as AI or GPTs is concerned,
I think about it all the time, again, from where it started to where it is now. I was just reading where there's a K-12 school in Colorado that has two-hour intensive AI instruction, and it's completely done by AI, the instruction, the curriculum, everything, and the rest of the day it focuses on life skills, which, to be honest, that's dwindling a little bit with all of
the gaming and all the other technologies, there's got to be a gap somewhere. So focusing on life skills isn't bad, but I think in the near future, I feel that we'll see a lot of enhancements when it comes to that data security, because it's such a hot topic right now. We're already doing so much with accessibility, but I think that's going to be a huge push when it comes to AI and ethics and the
ethical concerns behind it. Who is actually behind those keys training. So we really have to think about that. And honestly, I feel that’s where it's going. Of course, it's going to advance when it comes to the databases and the types of knowledge that it provides, but I think in the near future, we'll see a lot of ethical concerns and a lot of data privacy concerns being targeted.
Well, thank you. This has been a fascinating conversation, and your presentation inspired me to start creating GPTs because, at the time, I didn't realize that free users could access them. But now that that's there, it opens up so many possibilities for supporting students and for supporting instructors and anybody more broadly. So it's a really nice improvement. Absolutely, I couldn't agree with you more. It's so exciting of where we're going.
Thanks for a great conversation. And it really is important for folks to understand that it really isn't that hard to create them. It sounds really scary and really intimidating, but it really isn't. Yeah, the platform is already built for you. You just have to feed in that information. Well, thank you. This has been great. Thank you guys so much for having me. So much fun. If you've enjoyed this podcast, please subscribe and leave a review on iTunes
or your favorite podcast service. To continue the conversation, join us on our Tea for Teaching Facebook page. You can find show notes, transcripts and other materials on teaforteaching.com. Music by Michael Gary Brewer.