Small Steps, Big Impact Episode 23 | EDI Conversations around Generative AI

NODIA MENA: Welcome to Small Steps Big Impact, a podcast conversation about equity, diversity, and inclusion in our classrooms. Our guest today is Dr. Nir Kshetri Professor of Management within the Bryan School of Business and Economics. Our conversation today focuses on the intersection of EDI and artificial intelligence. Hello, Dr. Chhetri, how are you today?

NIR KSHETRI: I’m good. Thank you. How are you?

NODIA MENA: I am doing great. Thank you. Would you please introduce yourself to our audience?

NIR KSHETRI: Yeah, my name is Nir Kshetri This is my 21st year at UNCG in Bryan School. And my research focuses on the emerging, latest emerging technologies. Like long ago, about 20, 25 years back, I studied internet and e-commerce. Then after that, about big data and cloud computing and then blockchain. And, recently, more about, in the past few years, artificial intelligence and, recently, about the generative artificial intelligence, which is really kind of everyone is talking about that after the US startup OpenAI introduced the chatGPT last year.

NODIA MENA: Thank you for that introduction. Can you share a little bit about your research and experience studying or using artificial intelligence with students in higher education?

NIR KSHETRI: So not that much, currently, because after all these things came last year, I’m teaching mainly online. But I encourage my students to use as much as possible, all these generative AI, like chatGPT. But in the classes I teach fish to fish. Actually, I basically divide the class into small groups, and the books and the materials I give to them have really complicated concepts. And I ask the groups of students, small groups, to basically, OK, use the generative AI, chatGPT, and find what you, maybe they explain the concepts or theories, and all the things, in a simpler way than have done in the materials that I’m giving.

And they do a good job, actually. OK, the students actually look at ask the chatGPT and, OK, that’s what chatGPT said about that. And so in many cases, actually, that is simpler and they find that more engaging. And so that is also alternative way of learning new concepts, and they do a good job, actually.

BRAD JOHNSON: So I think we throw around AI and artificial intelligence a lot because they’re buzzwords. I think, especially here recently in our society. What exactly do we mean when we say artificial intelligence or AI?

NIR KSHETRI: Artificial intelligence has been here for a long time, since the 1940s. Actually, two mathematicians, McCulloch and Pitts, wrote a paper about what they call the neural network. And, basically, that is exactly how human brains work. And simulating machines to work as closely as the human brain. That is the basic idea behind artificial intelligence, basically.

Machines are behaving, working, thinking and like the human brain. That is the idea behind artificial intelligence. But recently, I think, most people are talking, artificial intelligence until maybe 1940s and 1950s, and it requires a lot of processing power. And they did not have computing power to use artificial intelligence. And they were not user friendly.

And, basically, it was only for big companies, only these techies in big companies. Even big companies, they did not use for every function. But it really democratized all these artificial intelligence after the generative AI came last year. Generative AI was kind of launched by the US technology startup OpenAI.

So generative AI basically involves generating, OK. They use generative model. Generative model is kind of easy thing to understand. Generative model means they are, means the artificial intelligence programs are trained with a large amount of data. Large amount of data means if we, again, go back to the chatGPT, it is trained with about 300 billion words.

So based on that, the artificial intelligence generates things. It is trained to do that, OK. It resembles the things with the data that they are trained in.

So when we talk about the chatGPT type of thing and they are special type of generative AI, we call them large language models. And large language models means inputs are text, means we can ask things in text. And output is also text, means they generate your response also in text.

But it is not necessarily the case. They generate text. They can generate images. They can generate video or they can generate any type of media, depending on the prompt and how they are trained on. So, basically, they generate, generative, AI generates, definitely, text or image or video or any type of data when a prompt is given.

The prompt means it could be simple statement that you like or it could be a question that you ask and give some prompt. Or it can be even, not generative, like chatGPT, but other types of generative AI applications. Like there are image generating type of things, like the midjourney or dall-e, and other type, and they actually generate emails for you or generate video for you, depending on prompt, either text or you can even write some type of emoji as prompt. So, basically, give some input to generate some type of data for you, depending on how they are trained on.

So this is kind of basic thing about AI and more specifically about generative AI. People say AI, but they are mainly talking about generative AI these days, more.

NODIA MENA: That is great. And in that same of line of information, how does AI relate to accessibility, equity, and universal design for learning, or UDL?

NIR KSHETRI: I think generative AI plays a really very important role in contributing all the goals of UDL that you are talking about. When you talk about accessibility and, for example, many students find it difficult to understand mathematics. Mathematics is really difficult for them. And that is one of the main reasons, actually, many students drop out.

OK. And, actually, they’ve done a lot of research surveys in the past, maybe 11 or 12, 11 or 10 or 11 months, after the chatGPT came. And so all the students that they have interviewed or they have asked, they consistently say that those chatGPT and other types of generative AI explain all these very, very complicated mathematics concepts better than their professors. Professors are good, but professors, maybe, they understand just mathematics, like if we are talking about mathematics professor.

But many of them may not understand how to explain that to people, students who do not have certain level of understanding. But chatGPT does that. And so you can ask chatGPT, for example, you are doing, you are in your undergraduate program you are doing, but your level of mathematics is, let’s say, the middle school level.

You can ask, please explain this concept at a fourth grade level. It can explain for you. Explain it eighth grade level. It can explain. It can exactly, right, the same thing in a different way, simpler way, explain it.

OK, you really, you are undergraduate but you think that you are at the PhD level. And you can explain, you can ask to explain explain this for a student at a PhD level. It can do that for you. So, basically, accessibility here is depending on the level of the kind of learner.

It can explain the concepts in a way because it is trained on, again, hundreds of billions of words, like chatGPT or other types of programs. So that is one thing. Another thing is many professors, they want help now. But if you write emails to professors or they have to deal with a lot of things, and it may take 24, 48 hours, in some cases, one week to get back. That might take time.

And this open AI is a type of, generative AI available 24/7 hours a day, seven days. You can ask anybody you like. And so and they are free, actually.

Many, not all of them, but many of the things they are available are free, like chatGPT or Google Bar. They do a good job, and you can ask anytime. They are free. And so those are some of the things here about the accessibility things. I think the UDL, basically contributing to the goals of UDL.

BRAD JOHNSON: So talk to us a little bit about how institutions of higher ed and maybe, more specifically, UNCG, can leverage this new AI technology in their EDI work or UDL work in the classroom.

NIR KSHETRI: Right. Yeah, again, in classroom maybe one possibility would be to kind of divide class into small gropus and depending on the level of students. And so maybe generating different types of study guides and learning programs. And asking them to basically generate something, asking the generative AI to get some output and then asking students to look at the things and what looks good.

And generative AI, actually in many cases, there is the thing. And so they just, if we ask them something, there is a complete the sentence based on the data that trained on. They do not know that is wrong or that is right. And they don’t know what they are being asked. And so basically, and there is the thing known as hallucination. That means, in many cases, actually– they say sometimes 10% to 15% of the times. And maybe they might be improving over time. And there is always some degree of hallucination.

They might give you answer in a very confident way, the artificial intelligence programs, where they are actually wrong. And maybe ask the students to fact-check. What did you find here? And what looks reasonable?

And how did you– and OK, and also, there is a way to write prompt in a better way. And if you write prompt in a good way, the rate of hallucination decreases, actually. And so if you ask to write more– if you give them more context and if you ask, also, give me the source of information that you do.

And so basically, students might look at the things in a more critical way and then find what looks reasonable and what doesn’t look good. And so they understand the limitations of the program and artificial intelligence program. And they understand their capability. That may be one thing.

But we might have to go beyond classroom. Maybe one possibility is– one thing is tutoring. Tutoring right now is probably students who have a lot of money they can spend on tutoring and learn things. But this is free tutoring available to everyone. Anyone can learn after the classroom in AI-based tutoring. And basically, asking generative AI to kind of explain things for them. And that might be one possibility.

NODIA MENA: Thank you for that. And how can faculty and staff better engage in the process of educating themselves around AI and how these experiences play out in the classroom?

NIR KSHETRI: Yeah. Basically, it is very, very important for everyone, like professors teaching in the university, to be familiar with the generative AI, and what generative AI can do, what are their limitations, how they can improve their materials. And so basically, people who have used and have basically read about a lot of surveys and they have done in companies, actually, and companies in any type of setting,

and all these artificial intelligence programs– recently generative AI– their interfaces are really easy to use.

And they are really user friendly. They are intuitive. And it doesn’t take a lot of time to learn. They don’t even have to learn anything. It is really important to know about artificial intelligence, learn about things and how they know, and basically, trying something, generating some materials, that study guide or a lesson plan and even planning a course or how to actually deliver a course.

And some concepts and maybe what generative AI is giving them and how that is– if that is helping students to understand the concepts that they are trying to teach. And if that is really correct, again, there might be some things– some of the things may not be correct. And so are they relevant? And do they meet their learning goals.

And so they might try. Spend a little bit of time, maybe 30 minutes, one hour. And I’m sure they do– I mean, like, artificial intelligence does a good job. And they can use artificial intelligence to plan their course. And they don’t know what to teach.

And artificial intelligence, generative AI can generate for them. OK, I want to teach this class at the college level for 16 weeks semester. Generate for me what exactly it is, exactly it generates for you. And OK, now go to the class. And I need to do– I need to teach this in class. And my students are of this level and suggest me a way to do that thing. It suggests you have to do this way and that way. And it does.

And if you like, follow that. If you do not like, follow only the parts that you like. And so basically, and all these professors, teachers, they basically use a lot of time in grading. Grading is the things that they do not like.

Actually, it helps you grade your things. And grade your things, you have to– you just generate some rubrics. If they cannot really generate rubric for what they are grading, that’s fine. They can even ask the generative AI to get rubric for them. Give them some rubric.

OK, this is the thing I’m giving you. And these are the labels. And if the students do this, and this will be excellent. And they will get a grade of 90 to 100. If students satisfy this, then they will get 70% to 80%.

And so divide into different categories. And they will do it for you. And the generative artificial intelligence do it for you. And they will do grading for you. You will save a lot of your time. And they can prepare study guide for you.

And they can prepare quiz definitely. And just give the label of the students. And I’m generating quiz for a calculus class at undergraduate, this label. And I want to know this essay question. So the multiple-choice questions, it will generate for you.

And sometimes, you have to spend a lot of time. And sometimes, I get one-page-long email. And it takes a lot of time to process that one-page email. And so OK, just give ChatGPT, reply to this one-page email. It writes a very good reply for you [INAUDIBLE].

And of course, you do not send that. Read that, and maybe delete some of the parts. And add some of the parts, and add context. But you save maybe 90% of your time replying to your email.

And also, a lot of administrative things that it can do. And it is really, really helpful. But you have to spend a lot of time. And good thing is they’re user friendly, many of them. And many of them are freely available, currently at least. I don’t maybe in the future, they might charge you to do the things.

And so basically, I’m sure a lot of people can save a lot of their time. And they can use the generative AI, especially to do routine things for them that they are spending a lot of time. And they might find it helpful.

And also, even if they are not actually, it is very, very important for students. Students here are not really– here, they have to pay a lot of money to go to university and spend four year, five year time to learn things. And that is noticeable in their life.

And so basically, artificial intelligence and generative AI is something that every student needs in the future. If they want to go to a job and everyone is looking for– like 90% of the companies in a survey actually, according to the ResumeBuilder company survey found that we give preference to people who have generative AI skill and ChatGPT type of skill.

And actually, they pay more money, tens of thousands of dollars more money per year to students. So even if you do not like and if we think about the student’s future, and it is very important for them to learn and understand what generative AI can do for them and what are their limitations, and so how to write a better prompt. And those types of things are important.

And so OK, it takes a lot of their time, professor’s time to use. Number two, even if they do not like and they have to use to help students, and so that they can have more experience about this generative AI, and they will get– they will be more attractive for the job market. And that’s why it is important.

BRAD JOHNSON: Yeah. I know I have used it for my own teaching. I have entered prompts in to check to see if what I’m talking about is a good summary, if I’ve captured everything. I’ve also used it to generate case studies because a lot of times, it is hard to find case studies about certain topics that we teach in. And it’s done a wonderful job.

And in terms of the email, I actually– my cousin was having to send an email out to his daughter’s softball teams and everything to say your daughter did or did not make it. And he was sitting there doing it. And I basically just put a prompt in that said generate an email to my daughter’s softball team telling them that their student– their child did not make the team. And it did like a five-paragraph, very elaborate email. So yeah, it is amazing and, at times, I think, a little bit frightening at times.

NIR KSHETRI: Right. And sometimes, it is really difficult subject that we have to reply– an email, the student is failing the class and is not doing good. And so it is difficult to write that email. And ask ChatGPT, this is the case. And ask– I mean, write an email in a polite way. And it does pretty good because, again, it is trained on about 300 billion words. And if it does better than people who process– who cannot process 100 billion or 200 billion words, definitely, in many cases.

BRAD JOHNSON: Yeah. So I’ve set it up. I’ve done my settings. And I said, so I teach master’s students in a higher ed student affairs program. And I want formal writing because it said, do you want formal or informal responses? And it’s very intuitive.

NIR KSHETRI: Exactly.

BRAD JOHNSON: Well, as you know, UNCG is– community engagement, especially for faculty and staff, is very big here at UNCG. So how could faculty and staff members utilize AI to establish partnerships out in the community?

NIR KSHETRI: Right. That is very important thing because we are here for doing things for the community in the long run. And so basically, many companies, really big companies, pretty much, they have started using generative AI in different applications, like marketing or human resources or in operations and many things.

But I mean, maybe many small companies haven’t yet started. But there are a lot of things, but they have no idea how to start and what to do with [? JI, ?] and all those things. Basically, one possibility is just to do

some type of outreach program, literacy program about generative AI so that communities aware of different benefits or maybe some limitations of that.

And also, another problem currently many companies are having is that what they need is not this old-fashioned type of skills that universities, many universities, have been teaching all over the world. Maybe the skills that the future workforce need, they currently need.

And maybe training students in those types of programs, short-term training, giving them. And basically, they can help them and maybe some type of internship program that our students who have skills in generative AI kind of do with them. And it is kind of a win-win situation for both parties– our students and the community.

So these are, in the long run, definitely kind of producing manpower that is kind of skilled at using this generative AI, when they actually go to the workforce. Those are some of the things. And maybe doing some type of joint research. And those are some of the things that I think the universities and community work together.

NODIA MENA: I had a conversation with my son not too long ago in regards to ChatGPT, AI. And sometimes, it is hard to bridge generational gaps in understanding, how artificial intelligence benefits society nowadays, and how there is still this thought about, oh, traditional ways of doing things, traditional ways of knowing, or more modern ways of doing things.

You see, there is this dichotomy in understanding how one can benefit another and whether one is better than the other. And for that reason, there have been a lot of– some resistance from some areas in utilizing AI. So a question is, what wisdom would you give faculty struggling with how better incorporate accessibility, equity, and UDL work in their technique or the teaching pedagogical practices, especially as it relates to AI and working with our college students?

NIR KSHETRI: Generative AI, those interfaces are really, really user friendly. They are really easy to use. And actually, if all these faculty who are now struggling with and who do not think that generative AI really doesn’t have any value in their teaching and learning and even research and all those activities, and so maybe that may not be true. And they have to try.

And so they definitely find all these applications really useful for teaching and research and like teaching and learning type of activities. So again, the most important thing is to try spending a couple of minutes of time and open an account with a ChatGPT, OpenAI. Or if they already have a Google account, any type of Google account, and they can automatically use Google Bard. Actually, I think it is almost as good as OpenAI ChatGPT. And they’re even making it better.

So spend time, maybe 10 minutes, 15 minutes, 20 minutes. And then I’m sure they will find it useful. And again, even if they do not, we are here. Probably, our teaching and learning activities are kind of– we are here to help our students and prepare them for the future of their job.

And so even for them, we are required to learn and teach them how to do that and how to use generative AI as much as possible in the college. And so they will have competitive advantage when they enter the job market. So basically, we do not really have a lot of choice. And everyone has to use these ChatGPT or any type of generative AI applications currently.

NODIA MENA: So it’s best to get acquainted with it.

NIR KSHETRI: Exactly.

BRAD JOHNSON: Get on board.

[LAUGHTER]

NIR KSHETRI: And one important thing, basically, going back to the main reason why many students kind of drop after their first-year program, is if they do good in the programs, and they will continue. And if they fail a lot of things and maybe math and this type of difficult concepts, then they will fail.

And so again, doing research with a lot of these students, and they say that these open– the generative AI programs, again, help them understand the concepts better. And study has indicated– research study published in one of the academic journals has indicated that if the students’ mathematics grade increases by about– by a one standard deviation, the dropout rate decreases by about 32 percentage.

So basically, OK, again, these professors we have, they are really good. But they might be good at mathematics. Or they might not be meeting the level of the students. They may not understand the level of the students. And then students, maybe they might be able to learn better with that.

And we need all these teachers to understand better what their students need. But the generative AI kind of personal– help personalize them depending on the level of their knowledge. And we might be able to increase their GPA. And they might be able to pass the class. And they might be able to continue the program and make more money in their lifetime because of all these things. So that’s why it is important.

BRAD JOHNSON: Well, Thank you, Dr. Kshetri, for joining us today to talk about what I know is a topic that is thrilling for some people and scary for others– so the topic of artificial intelligence, generative AI. So thank you for sharing your knowledge and your wisdom with us today.

NIR KSHETRI: Great. Thank you so much for having me.

[MUSIC PLAYING]

BRAD JOHNSON: Thank you for joining us today on Small Steps, Big Impact. Our goal is to have continuous conversations about equity, diversity, and inclusion in our classrooms. To learn more about EDI, please go to our website at go.ucng.edu/smallstepsbigimpact. Feel free to leave us suggestions for future topics and resources, or join us on an episode for a conversation.

[MUSIC PLAYING]