Volledige transcriptie van AI, Unscripted - Data Privacy, Student Rights, and the Real Cost of "Free" AI Tools

big group of folks looking at trustworthiness of AI in lots of different contexts across society. 00;03;00;06 - 00;03;33;03 Virginia Byrne And I'm lucky to be part of the education team there. And so I get to work with a lot of folks in, information systems and computer science and all across education to think about technology in the lives of young people and educators, both inside of and outside of the classroom. And I think right now there's so much research being done about AI in the classroom or AI as it relates to instruction or grading or plagiarism. 00;03;33;05 - 00;04;08;05 Virginia Byrne What's more interesting to me is about AI in the lives of educators and learners. That has nothing to do with the actual like grading of assignments. So, I study how educators feel about deepfakes and how are admissions officers thinking about the ethics of using AI in their work? And how do we use AI or not use AI strategically to get rid of some of the busywork in our lives? 00;04;08;08 - 00;04;11;00 Virginia Byrne So that's just what's calling me right now. 00;04;11;02 - 00;04;38;02 Jennifer Potter I think that's a great place to start, especially given you know, your TRAILS work and and some of your other work really gets us into data questions around data privacy and some other kinds of concerns. So before we get to those concerns, let's talk about how you actually use AI in your own work. So what are some examples of appropriate or even beneficial ways that you've been thinking about the use of AI? 00;04;38;05 - 00;05;05;06 Virginia Byrne Yeah. So these questions are really grounded in what tools we currently have access to. Right. So sometimes I'm worried that when we talk about like large language models and we're recording in April 2026, when if somebody listens to this in April 2027, it will sound ridiculous. So our conversation's like really grounded in what we have access to right now as, regular people. 00;05;05;09 - 00;05;33;28 Virginia Byrne And so I do a lot of workshops for folks in higher education, K-12 teachers. And I really encourage them to ask themselves some like reflection questions to sort of get their own sense of appropriate use of AI with the tools we have right now. And so I created these based on what I ask myself of, like, I don't want to use this tool in a task that's really high stakes because I can't trust it. 00;05;33;28 - 00;05;59;08 Virginia Byrne That much. So it's always making sure that I'm using it in low stakes situations. I'm always wondering if the task is a good fit for the tool. So like, can the AI tool actually do this task? Can I verify the accuracy of the output? And then what kind of privacy concerns are going to emerge from this that I might not have yet flagged? 00;05;59;11 - 00;06;21;22 Virginia Byrne Because one of the pieces of AI in our everyday lives is helping people realize like, oh, I forgot that that should be private. I forgot that I shouldn't be sharing this with some company I don't know anything about. I think it was years ago when people started to recognize if the product is free, then I am the product. 00;06;21;25 - 00;06;52;09 Virginia Byrne Right. I think that became really common around Instagram probably ten years ago, of people realizing like, this is an incredible product. Why is it free? Oh, because I'm the product. They're selling me things. They're using my data to sell me different things. And we're at this interesting point where I think people are realizing, oh, my voice is valuable, my image is valuable, my search history is valuable. 00;06;52;11 - 00;07;22;23 Virginia Byrne All of my choices in how I construct a sentence is valuable. And that has been super interesting for me of just having people, especially educators, think through what they're realizing is good, interesting data. And how do they think about keeping that private or not. So I get to be in a lot of cool conversations with educators where we talk about what information should be private about themselves, but also about their students. 00;07;22;25 - 00;07;45;24 Virginia Byrne And so you asked a different question, but I always want to start with those, like, make sure that the thing is low stakes to low stakes task. The output is verifiable and the AI tool can actually do this task like there's fit there. And so you asked what are some examples of appropriate beneficial uses of AI? 00;07;45;27 - 00;08;22;27 Virginia Byrne And I think right now these large language models are really good at simple wordsmithing or text generation that makes my life easier, but doesn't potentially damage my reputation or relationships. So I often use ChatGPT as a rage translator or an anger translator. I don't know if you remember that old Key and Peele sketch. But like when I'm trying to write an email to a, to a colleague that I'm angry at and I'm realizing, like my phrasing is just not a professional and appropriate. 00;08;23;00 - 00;08;42;03 Virginia Byrne I use ChatGPT to help me make it kinder, make it softer, make it more agreeable, make it more professional. And I think that is a great example of what we can currently use these tools for, because the output is super easy to verify. I can read the four sentences and it'll be fine. It's a good tech like task fit. 00;08;42;05 - 00;09;06;17 Virginia Byrne And it's not going to ruin my reputation if I send an email that sort of looks AI-generated in this particular context, right? To me, that makes a lot of sense. One of the things I like to do from my other research is we found that students love getting Canvas announcements, or if you use blackboard or whatever, they love a learning management system announcement at the beginning of each week. 00;09;06;19 - 00;09;25;21 Virginia Byrne So what I have ChatGPT do at the beginning of the semester is I put in my schedule and I say generate an entire semester's worth of announcements. And then I sit down on my couch and I copy and paste them and pre-scheduled them in

Canvas to auto send every Sunday. And on the first day of class, I tell students this exact thing. 00;09;25;21 - 00;09;46;24 Virginia Byrne I'm like, here, I'm going to be upfront with you. I cannot remember to send these announcements. So my friend ChatGPT and I made them together. I scheduled them to you and they are. That's what you're going to get in your inbox. And they love it because they're getting the announcements. But I think they also like that I'm upfront about what's appropriate use. 00;09;46;27 - 00;10;05;17 Virginia Byrne And I think that that's where I would encourage folks to use these tools right now is like as a crutch to help you when you're forgetting something or it is a low stakes but important thing. And you just need a little extra help, or else you might avoid the task entirely. 00;10;05;24 - 00;10;38;19 Jennifer Potter So we're going to switch gears a little bit, maybe, and talk about some of the concerns, particularly around data privacy that you mentioned. So, most of us think we understand what data is. But there's a question if we really do understand what data is. And so can you talk a little bit about that. And also, you know, just sort of more broadly like what what should faculty be thinking about or know about data privacy when it comes to AI in particular? 00;10;38;21 - 00;11;04;27 Virginia Byrne Yeah, the field needs better digital and data literacy overall. Right. And part of data literacy is understanding privacy concerns. And how everybody has their own sense of what they want to keep private. And we all have the right to feel differently about our data privacy. And that's something that I think not everyone is okay with that. 00;11;05;00 - 00;11;36;07 Virginia Byrne Just because you're okay sharing your video with ChatGPT doesn't mean I'm okay sharing my my voice with ChatGPT. And I think the field needs to get better at having conversations about an individual's right to their own ideas, voice, image, way of writing, and way of speaking. And so you ask a good question of like, we need to have a better conversation about what we mean by data, both for ourselves and for our students. 00;11;36;09 - 00;12;05;01 Virginia Byrne Because it's a lot more than just our names and our IDs. Right? It's also about, all the ways that we move in on campus. It's the ways the things we sign up for, the things we go do, our engagement on campus. There's been some peculiar studies out of Turkey with the mandatory facial recognition software in all higher education spaces, of which college students are going to class and who are not. 00;12;05;04 - 00;12;36;22 Virginia Byrne And I think when we talk about video like that, it makes sense that we might not want to share our videos with people or with these AI tools, but it's also about how we construct a story and how we construct a sentence that these large language models are so desperate for more text and more original ideas that it's really important that we start to recognize that students own their ideas as much as they own their sentence structure. 00;12;36;24 - 00;13;03;17 Virginia Byrne I would love to see the field have a reckoning with students do not give up the right to that important data. If a faculty member wants to use AI to help them grade right? Like, I see a lot of scary situations where faculty are uploading student essays to something, to a ChatGPT to help them grade and find feedback. 00;13;03;19 - 00;13;30;17 Virginia Byrne And the student didn't consent to that. The student didn't consent to their ideas and their sentence structure being put into that software. And I think that we're just going to have to keep having these conversations. And I think students are going to really start to speak up for themselves and demand some ground rules. And then I think faculty will start to realize that these free tools are not free. 00;13;30;17 - 00;13;36;22 Virginia Byrne They are absorbing all of our data. Right? They're not free. We are the product ourselves. 00;13;36;24 - 00;14;11;10 Jennifer Potter Can you go a little further on that? I'm curious about, like the tools at the campus level that are enterprise tools in quotations that that we're often told, you know, they're fully secure. Because they're enterprise, they're walled off, kind of all of this discourse around how secure and private they actually are and whether or not that's true or, or kind of how you're thinking about those enterprise level tools that supposedly come with this data privacy. 00;14;11;12 - 00;14;38;15 Virginia Byrne Yeah. Well, I have real trust issues. I have witnessed us trusting things like Google Docs only to have all of that be fed to the machine. People storing all their photos in Facebook just to have that be fed to the machine. So forgive me that I have trust issues, and I don't believe a single thing about a company upholding my privacy, my personal privacy expectations. 00;14;38;17 - 00;15;01;08 Virginia Byrne And when I talk to educators, there's definitely this sense of hopelessness there, that like, well, it's already out there. They already have it. So why does it matter anymore? There's, a degree of, like, of loss. People feel like they've lost control and the right to their own image or their right to their own data. 00;15;01;10 - 00;15;25;19 Jennifer Potter Yeah. I also think there's something really interesting about, you know, if for faculty, for anyone but faculty who may not sort of be deep in the discussions around data privacy, you know, if you're told or if you, you know, have some kind of messaging that says, of course, you can upload, you know, you can upload your students' work because it's secure. 00;15;25;21 - 00;15;40;13 Jennifer Potter And so I think that sometimes folks aren't doing it. You know, for any they're not trying to do something wrong. They really think that those are secure, that they are private, and they're trying to improve their own grading time. Right. 00;15;40;15 -

00;16;02;27 Virginia Byrne When we think about these types of privacy violations, it's because we are anticipating a flow of information in only one particular way. Like when I upload a student's paper to turn it in, I expect it to TurnItIn to just check it to other papers and then delete it and move on. But TurnItIn might be keeping all those papers and selling them or doing something I don't know. 00;16;02;27 - 00;16;18;21 Virginia Byrne Right. And it's those violations of the assumed flow of information that I think is why I have trust issues. Right. Like I don't trust anymore that these companies are going to do what they say they're going to do with our data. 00;16;18;24 - 00;16;35;07 Sam Collins We don't know really what what they're doing with data. We're not sure how the data that they, have have, you know, harvested is used to train, their various tools. And we don't really know, like what even that training data consisted of? 00;16;35;09 - 00;17;04;02 Jennifer Potter Yeah. Well, and just to take it maybe to the next level of thinking about how corporations are using our data and how how we're interacting with corporations. I'd love to talk a little bit about sort of the work that you do around equity and racial justice in this regard. Right. Because I think there's, you know, all of this research coming out about how AI perpetuates harm against marginalized communities, particularly Black students. 00;17;04;04 - 00;17;26;12 Jennifer Potter And so, you know, I know you've done a lot of research here. So kind of what are you seeing? But then also, how can faculty interrupt that and, you know, make wise or pedagogical decisions in, as they're using AI in their classrooms that, that don't, you know, harm students in these ways? That's a big question I know. 00;17;26;15 - 00;18;03;14 Virginia Byrne No, no, no. It's good. And I think you name it up front, right. That I these they're not neutral. They're they reproduce racism, sexism, other forms of social injustice, that because they're trained on, large amounts of data that has racism, sexism cooked into them, their output is going to perpetuate those harms. And I know that these companies are doing what they can to try to remove some of that, but it is still cooked into the machine. 00;18;03;16 - 00;18;40;04 Virginia Byrne And so there's little we can do except for keep humans in the loop. Right. And I think at this point there's when talking to educators about AI use, it's making sure that there's always a human there, being the expert to look at what's coming out of these either recommendation systems or what the output is from an LLM, to take a discerning eye and make sure that it's not suggesting things that are harmful or, perpetuating misunderstandings or policing people in a way that is unjust. 00;18;40;06 - 00;19;15;01 Virginia Byrne And there's been, a lot of cases of this in the field, right. Because many of these LLMs were, trained on a white American -entric database that their

interpretations of world events have that view, that their understanding of what is correct English, correct in quotes here, has that interpretation and view that they are going to identify different dialects of English is incorrect or sloppy or other harmful adjectives there. 00;19;15;04 - 00;19;49;12 Virginia Byrne And and that is harmful to students and to see their language be slashed up by Grammarly when it is, you know, a meaningful dialect of English, that's harmful for both the student to see, and a faculty member could misinterpret that as, oh, the computer's correct. I'll just give them this feedback. So it's really important that educators say like, are very cognizant that these tools can be problematic, and that they are the first line to make sure that this isn't perpetuating harm against students. 00;19;49;15 - 00;20;00;03 Virginia Byrne It's not giving bad recommendations. And then we're we're aware that these black box technologies, well, we know one thing that's in the black box, and it's harm. Well. 00;20;00;06 - 00;20;23;00 Jennifer Potter Let's talk about your work. So you're working in sort of you're positioned extensively in student affairs. So you're thinking about advising and missions conduct. Right? All of these pieces. What's happening with AI in these spaces? Sort of the non-teaching spaces in student affairs in higher ed, like what's the landscape look like? 00;20;23;03 - 00;20;50;25 Virginia Byrne The landscape is being filled with vendors who are selling solutions somewhere between meaning well and a real grift. These companies are coming in, and I went to this great talk, at AERA, the big, American Education Research Association conference, where this speaker who will rename remain nameless, said, like, you know, they tried to sell these chat bots to companies and industry didn't accept them. 00;20;50;25 - 00;21;15;00 Virginia Byrne So, you know, where they're taking them to? Education. And so one of the things I do is I work with a lot of educators and, we do all these AI literacy trainings. We're AI literacy to mean just that doesn't mean being able to use AI. It means being able to have a meaningful conversation about the appropriate use of AI and like being able to push back when these vendors come to your office. 00;21;15;02 - 00;21;44;21 Virginia Byrne And it is helping people articulate like financial aid officers are cannot be replaced with a chat bot because financial aid officers are the front line social workers to college students having huge crises in their lives. Right? Financial aid officers deal with more student trauma and crying and breakdowns than almost anybody else on campus. Maybe academic advisors are second. 00;21;44;23 - 00;22;09;24 Virginia Byrne Right. And so these folks, the idea that they could be replaced with a chatbot is so offensive. And, it is now just my mission to figure out how to, equip these folks with being able to explain why their work is so valuable and how they they cannot be replaced with an AI. And even having an AI come in as a, as an aide is dangerous. 00;22;09;27 - 00;22;42;28 Virginia Byrne And so I know that sometimes after I do these workshops, folks come up and they're like, I appreciate your your take because I'm a real Luddite and blah, blah, blah. And I always have to go on my little tangent of like, Luddites are amazing. They're badass feminists. They were incredible people who made sure that workers weren't replaced by machines, and that they didn't just stand against technology, but they were for the rights of workers and making sure that work is done correctly, not just fast. 00;22;43;00 - 00;22;53;03 Virginia Byrne Right. And so I always like to say, like build up the Luddites, because Luddites sometimes are the best critical thinkers of like, we need to do work well, not just quickly. 00;22;53;06 - 00;23;06;03 Sam Collins And I think that goes to what you were saying about AI literacy. Being able to think critically about AI is AI literacy not being a consumer of AI. That's just simply being a consumer. 00;23;06;05 - 00;23;14;16 Virginia Byrne Yeah, yeah. And so I think when we see all these trainings about AI literacy and it's about what to click on, that makes me very concerned. 00;23;14;18 - 00;23;57;07 Jennifer Potter I, I love this conversation because I think, you know, there's so many faculty that are that are wanting to think about AI in the classroom in the right ways. And, you know, I say in the right ways very meaningfully because I think that most faculty are not just looking for the easy answer. And so they're thinking about their own literacy, you know, for for faculty who who want to experiment and they want to think about how to teach students about AI literacy, and they want to also, you know, sort of use it to maybe do some new things pedagogically in their own classrooms and do some of that experimentation. 00;23;57;09 - 00;24;15;17 Jennifer Potter What what's your advice to them in terms of sort of, you know, ways that they can do this in the most responsible and ethical of ways, but still kind of maintain some of that interest and experimentation. And, and continue to, to move in this space. 00;24;15;20 - 00;24;38;21 Virginia Byrne I think that faculty should start if they're interested in playing around with AI in their teaching space is to use. So I'll keep using ChatGPT as an example, like use it as a coach for yourself. So put in your assignment descriptions and say like help me make these clearer, or put in your rubrics and say like, what about this doesn't make sense? 00;24;38;21 - 00;25;12;05 Virginia Byrne Or put in your whole syllabus and say like, are there any broken links? Right? And like first doing the most basic things with this learning management or LLMs of like have it, see where you can improve before doing some type of big mandatory assignment where all students have to use LMS to do blah blah blah, right? I think first getting a good handle on it by

having it critique you is a is better. 00;25;12;07 - 00;25;42;12 Virginia Byrne I also think it's nice when faculty are clear on their syllabi about their own AI use and what they expect students to do. I know that that's a contested space right now about appropriate use and plagiarism, but just having your own appropriate use policy for yourself as an individual, I think it's a good, activity for faculty to do of like telling students, I won't use this to grade your assignments, but I will use it to improve my assignment descriptions. 00;25;42;15 - 00;26;07;24 Virginia Byrne I use that for myself, and I've realized that I'm not the best at writing assignment descriptions. My assignment descriptions are pretty wonky, and when I have it, make tutor like or even a worksheet like AI does it really clearly. And then I go, oh, that's what I should have been doing. That's cool. And that's a useful thing. But I would encourage faculty to have it start with them before they require other people to use it. 00;26;07;26 - 00;26;47;03 Virginia Byrne Because I do think as we talked about before, there's this this growing concern around students, that they're being asked to use AI or being policed with AI in ways that they don't consent to. Last summer, we did a three day teacher professional development with a bunch of teachers here in Baltimore, and one of the activities was tell this particular AI, this particular LLM, what kind of make a lesson plan with all the materials for this particular like eighth grade science thing and then critique it and they had it make something with all the slides and the worksheets and the assignment. 00;26;47;05 - 00;27;07;11 Virginia Byrne And these eighth grade teachers just like were rolling on the floor laughing like you think an eighth grader is going to do this. This is so silly. It doesn't make sense from start to finish. It's boring, not it. Like it was so encouraging because at first they were like, oh, this could replace me. And then they were like, oh, this is junk. 00;27;07;13 - 00;27;33;19 Virginia Byrne And I feel like that's where I think faculty could also live is like, have it you and have it make slides for you. See if that makes any sense. They are going to be so boring and vague and weirdly catchy, and students are going to see right through that, because students are also very like the average Gen Zer can see AI generated content from a mile away and they are cringed by it. 00;27;33;22 - 00;27;46;21 Jennifer Potter Well, I think that brings us to the end. I that's a great last line to capture. So thank you so much, Virginia, for joining us today. This has been a really great conversation. 00;27;46;24 - 00;27;49;08 Virginia Byrne Thank you so much. Thank you for having me. 00;27;49;11 - 00;28;14;29 Jennifer Potter As we've heard, there's no single script for exploring new uses of AI in our classrooms. Each educational context presents unique opportunities and challenges, and we're all learning as we go. If you have questions or want to share your own experiences with generative AI and teaching and learning, we'd love to hear from you. Please email us at AIunscripted@usmd.edu. 00;28;15;02 - 00;28;38;11 Sam Collins This limited series is brought to you through the collaboration of the University of Maryland, Baltimore, the USM Kirwin Center for Academic Innovation, the USM Council of University System faculty, and MarylandOnline. Join us next time for our wrap up episode of this encore limited series. Until next time, keep experimenting and stay unscripted. 00;28;38;13 - 00;28;51;27 Scott Riley Thank you for joining us today on Moving the Needle. Visit us at umaryland.edu/fctl to hear additional episodes, leave us feedback or suggest future topics. We'd love to hear from.