UDL, Access, and AI

I mean, I already know the answer, but… I am drinking two non-conventional teas. You may have heard of them: One is Diet Coke, and the  other is water with a little bit of Mio flavoring in it, which I think counts as tea. Tom? I'm drinking a lovely decaf rooibos. That sounds good. Sherri? So I actually am drinking my favorite iced  coffee. I have a salted caramel mocha with me. Great. And I am drinking a Tea Forte black currant tea. And I'm back to a good old favorite that I

picked up when I was at Epcot yesterday, which is  English afternoon tea. Our last segment for today will feature questions from the audience.  Because we're managing two sets of mics, as you can see, one for amplification  in the room and one for our recording, we ask that you use the QR code on the screen  to submit your questions for the panelists. Technological advancements like touch interfaces  and speech recognition were initially developed to

provide access for people with disabilities, but  now the standard features used by many. Likewise, early AI innovations, such as real-time  captioning and transcription services and tools like spelling and grammar software,  initially designed in support of accessibility, have been widely accepted and adopted as  useful supports for all students. In what ways can instructors use AI tools to efficiently  develop accessible digital content or design

multiple means of representation or options  for interaction? We'll start with Sherri. Thank you. So I had an opportunity to think about  this question in advance, and in my opinion, based on a couple of decades of watching AI come to the  forefront, I feel like AI tools themselves are, in their current state, something that provides  faculty and students, as well as instructional

designers, with the ability to have different  perspectives than we would otherwise. I kind of highlighted, and I wanted to go ahead with our  questions by highlighting two of my favorite tools that I figured you guys may not have found yet,  because we all know what ChatGPT is, but one of the tools that I have evaluated, and I feel like  is going to change our world in course development is called LearnWorlds, and I encourage you to take  some time to dive into their free 30 day service,

because it is amazing. The one that I use most  extensively with my students is called Goblin Tools, and both of those tools are ones that, over  the last three years, have extensively changed the space of building what is available for us as  faculty and helping us if we don't know anything yet about UDL, it's actually helping us to create  those without having to have too much background

knowledge, because UDL is built into the. Central  mechanism of how these tools work as AI. So I encourage you to take a look at Goblin Tools and  Learn Worlds if you've not looked at either of those yet, because they are very interesting, and  it will get you kind of started with a tool that was pre built on the foundation of UDL. Tom?

Well, the key here is that it used to be up to the  designer or the instructor to create and provide the multiple means of engagement, representation,  action and expression for the learners, the 3.0 version of the universal design for learning  guidelines recognizes that artificial intelligence and other tools now put the ability to create  in everyone's hands. So the verbs changed from provide to design. This acknowledges that we can  both give UDL options and teach learners how to

recognize and craft their own customized  options as well. For example, before large language models came along, if I wanted a text  version of a video clip or I wanted to hear it

in another language. I had to rely on humans for  those transformations. Now I can ask AI tools to create the alternatives I'd like, and then, and  this last part is crucial, do a double check with the humans who have the skills to be able to  say, yes, that's accurate and trustworthy, As noted earlier, AI tools can also help  students with grammar and spelling and basic writing and organizational skills. Will  these uses of AI allow students and instructors

to focus more attention on the development  and higher order thinking skills. Liz? I just want to say that when I was invited to be  on this panel, I thought these people all know a lot more than I do. So when I was thinking about  how to answer these questions, I did what I think many of our students might do, and go to ChatGPT  and say, like, “give me some thoughts on this.” And I found that to be really helpful to kind  of clarify how I want to answer this question

about whether AI tools can help us focus more on  the higher-order thinking skills. So in response to this question, I would say “yes, and…” because  generative AI can create opportunities for us to focus on those higher-order thinking skills, and  to do so as educators, we have to invest time and thought into creating assignments and projects  that do that, that engage those higher- order

thinking skills. And this is something that I've  tried to make clear to my colleagues when talking about generative AI, both in my discipline of  Political Science and with the faculty I work with in our center for teaching. And I'm just going to  do a quick nod to the 1992 presidential campaign, because that's my age, so please don't take this  as any sort of like derisive comment, but it's the pedagogy, stupid. It's always been about the  pedagogy, and John Warner's book Why they Can't

and do the hard work of education. Sherri? So, I agree with Liz on her premise that it  is really about the pedagogy. It's always been about the pedagogy. I am in a position now where  not only am I teaching at my home institution, but I also teach at a couple of other institutions  as needed, and each of them have exceptionally

varying policies and restrictions on the use of  AI. I won't name the one that has a restriction, but I will point out that because my work is  extensively in students with what we would call hidden disabilities, a lot of times there's a  misunderstanding about the need for AI. AI is absolutely a requirement. And for example, Goblin  Tools I use with all of my students who have any kind of learning disability, ADHD, like myself, if  they have any kind of mental health concerns, then

Goblin Tools is an essential tool for allowing  them to bolster their executive functioning. And I've had to share that research and share that  type of understanding with my colleagues at the institution that has a restricted AI policy that  tells students they will fail if they are known to

use AI. And I've had to tell them that this is  an equity issue. It is a requirement for these students to be able to utilize this and there  was a presentation two sessions ago, earlier this morning, where they shared that less than 20%  of institutions in the United States have college students that disclose a disability. That means  we're missing about 80% of them, and that's an

important fact to keep in mind. They may not tell  us they have a disability, but it is important for us to advocate at all of our campuses,  that AI actually is allowing individuals who may struggle to do these basic skills of spelling  and grammar and outlining without these tools. Students enter our classes with quite a bit of  variation in their prior learning experiences, and that can create a lot of challenges for  instructors. In thinking about designing

outcomes to sustain student effort  and persistence. How might AI tools be used to provide enough challenge to individual  students and enough individualized support that's tailored to the needs of each student? Tom? Well, beyond just creating multiple formats as materials, and that's something that large  language models and predictive AI do very well, instructors and students alike can use AI tools to  create self-quizzing questions about materials and

differentiated study pathways based on learners’  own engagements. The most promising access barrier we can lower with AI has to do with customizable  study methods that learn about you and help you practice, prepare, engage. Of course, many  of us know that the least effective way to study is print out a text and review it with  your highlighter marker in hand. The better way is to read and create application and recall  questions as you go along, and then after a while,

quiz yourself about the materials. Our learners  may not yet be experienced in crafting their own self-quizzing questions; they can ask. Ai, not for  a summary of the reading that would be doing the work in place of the student, but for questions  to ask about the reading that's augmenting and customizing the study skills of the learners.  Liz, you got thoughts on this too, don't you?

I do, and something that I don't hear discussed  very often, although, Sherri, I'm glad you brought it up, is that these AI tools can be really  helpful to support executive functioning by offloading some of the things that executive  functioning is all about. So that can be really helpful for our neurodivergent learners, and also  for us if we're neurodivergent. So for those who struggle to get organized, AI tools can help. If  you need to break a project into discrete goals.

AI can help if you need to create a schedule or  reminders or you want to come up with some ideas to gamify, AI can help with all of those. That's a  really nice affordance of the technology students can take advantage of these tools to help them  identify areas where their work might need some strengthening, or provide feedback on where  they've made mistakes in some of their work, which can help students who might otherwise  feel left behind or feel lost to catch up

and stay with the class. So just two days ago, I  had a study session with my statistics students, and one of them told me, when I get to a question  in the homework that I don't know how to do, I open up ChatGPT. I say, don't give me the  answer, tell me what are the steps to do this question. And that's a really nice way  for them to learn the process and then take those steps and apply them themselves. This is a  really smart use of AI tools by a conscientious

student trying to learn and improve. Continuing on with the principle of designing multiple means of engagement, one UDL  guideline is to design options for welcoming interests and identities with considerations  like providing students with a higher level of autonomy and choice or making sure that learning  experiences are meaningful and relevant. What roles do AI tools have in assisting faculty and  supporting student identity and interests? Sherri,

do you want to start on this one? Absolutely. So I had a different answer for this until yesterday. So I want to tell you that  I'm editing on the fly here because I feel like I have utilized the principles of inclusive design  and UDL, I've built in identity until yesterday, and somebody presented on a wonderful session  about how they submit their syllabi into AI with a specific prompt telling them what your background  and what your history and what your varied

identities are. And it changed my perspective,  because I haven't done that. I've been trying, on my own, to tell students I am also neurodivergent.  I also come from this background, and I've tried to use that to identify with them, but the  solution that was presented yesterday was really unique, because they submit their syllabus and  they say I come from these five or six identities.

Help me make sure that my language in my syllabus  is open and inclusive to all identities. That's a technique I'm now going to be implementing as of  yesterday, because I've been trying to do this all on my own with thinking of the other identities.  But the real value of artificial intelligence… and just allow me to nerd out for a second as  a psychologist… is intelligence is just what

the general community knows. That's how we define  intelligence. Artificial Intelligence, they know a lot more than us, because it's big and broad,  and it touches every culture and every continent on the entire world. And so from that perspective,  I will always get a less biased response from AI, instead of trying to rely on what my own  brain can tell me about trying to be open

and inclusive of everyone's identities. So that's  a tip that I'm taking directly from our colleagues from yesterday, of just even starting with a  syllabus and saying, “These are my identities, help me make sure that my work is representing  all other identities who might be in my class.” It's a great example… continuous  development, continuous improvement. Liz?

Yeah, so I want to echo what Sherri said, and just  say that one of the uses of AI that I think we often overlook is this idea that it can help us do  some reflective practice about our own potential cultural gaps or lived experiences and those of  our students. And so if we want to support student identity and interests, we probably need to have  at least some fluency with that, or the ability to get that, and especially for many of us who are  not the same age and even close to the same age

as our students, or for those of us who teach very  large classes, this can be very hard. So I would like to suggest that AI tools can help. Sherri,  don't want to go off on too much of a tangent, but I was listening to another podcast episode  this morning with a conversation with Maha Bali of the American University in Cairo, and she  was talking about the implicit bias that comes from large language models, because they're  reflecting back the text that they have been

trained on. And so if you ask these generative  AI programs, what is terrorism? or who are the people who are terrorists? It won't answer.  But if you ask it to give you five examples, they'll all be of a certain kind. And so I think  that there are opportunities here as well for students to think about and for us to learn  about implicit bias in the culture by doing a

critical analysis of the stuff that generative AI  gives us. So Bonni Stachowiak talked about asking ChatGPT or some other program to give her a  picture of a classroom that was a philosophy class at Harvard, and they were all men,  and the women in the class noticed that,

the men did not. So these are opportunities for  some critical reflection. I also just want to say one more thing, AI tools can be really helpful for  our students who are not native English speakers, because they can help them build fluency by  parsing text and correcting their own writing. Traditional online assessment techniques such  as discussion boards and essay assignments on traditional topics in our disciplines may  not align well with the diverse interests

and lived experiences of our students. How might  AI tools be helpful in designing well scaffolded assignments that better connect to the diverse  interests and lived experiences of students, providing them with multiple means of action  and expression? We'll start with Liz. So, from instructors’ perspectives, AI tools  can help those of us who are subject matter experts and may have a hard time adopting a novice  mindset, break down projects into smaller steps so

that our students can approach those in a more  accessible way. If someone asked me to write a journal article in my discipline, I would know how  to do that, soup to nuts, right? No problem. But a sophomore or junior in a class who's asked  to write a term paper may not even know where to get started or what kinds of things will help  them get there. So AI tools can help instructors who need some help returning to a novice mindset  and understanding what those discrete steps are

to create that scaffolding that you mentioned.  From the students’ perspectives AI tools work well here, if and when, instructors allow students  at least some choice over what the topics of their assignments or projects are, and especially to  create assignments in different formats. So be that, video, audio, visual, written, etc. In those  cases, AI tools can help students identify topics,

refine ideas and then create structures for  their eventual work products. That helps the instructors meet the students where they  are, and as an instructor, it makes grading those assignments far more interesting, because you're  not reading 100 papers on exactly the same topic. I love getting to know my students through their  assignments, and I suspect many others do as well.

Tom, I think you have some more thoughts. Yeah, to build on what Liz is talking about here, we're most often engaged in creating  assessments of learning: tests, quizzes, papers, exams. AI can help us construct those  kinds of things and suggest alternate ways that learners can show those skills. But where I see  the greatest potential for artificial intelligence

is in thinking about assessment as learning.  In our 300-person online lecture courses, they're a terrible format to begin with… don't  get me started… there are scarce opportunities for engagement and for showing what you know.  Designers can craft ways for learners to use artificial intelligence tools, almost like a  private tutor. Ask AI to create study flash cards, self-quizzing questions, like we talked about  earlier, fill in study guides and the like. The

key is to make sure that what the AI produces is  good information. Vet the content yourself during the creation process or design in a tech check  as a course activity so that the instructional team can assess the quality of the self-  assessment materials that AI is generating. So we've talked a lot about how AI can facilitate  learning, but haven't yet addressed some of the equity issues in accessing and using AI tools.  What barriers might some students face in using

AI tools in their online learning experiences?  Tom, do you want to start this one? Yeah, I can take this one. There's a lot of  ethical considerations to using generative AI tools in online education. I'll spotlight four  of them here. First, we already see the haves and the have nots, the folks who can and can't  afford to use customized targeted data sets and tools. We have customized tools for the C-suite  and hallucinations and errors for the rest of us.

Second, water and electricity usage… for every  image that you ask AI to create of a butterfly, unicorn, kitten flying through space eating  a cheeseburger, three liters of water and 10 watts of electricity are consumed. Third,  we preach respect for intellectual property, while the most common LLMs and generative AI  models have been trained on oceans of copyrighted

content without consent. And fourth, and perhaps  the biggest barrier for widespread adoption of AI tools for UDL purposes, the high prevalence  of racist, sexist, and pornographic inputs to the most general models. Ask AI for an image of a  doctor, and it will always create a white, male, older doctor. These are all elements that we  should share with learners about the tools that they're using. That was a bit of a pessimistic  turn. Sherri, do you have something different

here, or you want to follow in the same way? Mine is going to support what Tom said and actually what Liz brought out earlier. So  at one of my universities, because I use AI as a teaching tool in the field of psychology,  I actually build into week one and two of our classes how to use AI effectively. And then,  after my first semester, because so many of my students love and are familiar with the free  tool, Canva, that you might be familiar with,

Canva now has a free functionality to  generate images. However, Canva is, to date,

the single most biased image generator I have  ever seen. And as soon as I recognized that when they input things like mental health  or ethnic diversity or poverty or crime, that it always provided the same specific images  of individuals, then that led me to needing to narrate and modify the way I teach them about  which types of AI tools and what to trust, because as future counselors, they do not need to  be creating content or using AI's incorrect LLMs,

as Tom and Liz have pointed out, to create a  belief instead, that only these individuals are representative. Tom mentioned doctors, and they  always have one particular group. In my area, we're talking about mental health disorders, and  it is so, so problematic as a professor to have the student potentially create a presentation  that only represents certain ethnicities and

certain genders when we're talking about mental  health disorders, when that is incorrect. And Liz, you highlighted this so well, even without knowing  what I was going to say, because the intelligence behind AI is growing. It is growing every day.  And one of the statistics I wanted to share with you that just came out last month was some of the  original developers of AI estimate that AI is in

its infant stage, it's not even a toddler stage,  but it currently has an IQ of 160. That's five points below Albert Einstein, but it's only  a baby, and so once it grows, it, we hope, will become less biased, because it will receive  input from us, but it gets its knowledge from us. So if we are inputting biased information,  it will continue to output biased information. So it's only as good as we are good at not being  biased about the information that we're publishing

as well. So thinking about that diversity and  the equity, I think that we cannot encourage students and we cannot utilize AI ourselves  without also making sure we're teaching them about the inequities. And if they understand the  dynamics of how AI is created, its intelligence is based on the fallacy of humans, sometimes are  also biased inherently, then they know to be a little bit more conservative about evaluating the  quality of the material it presents as well.

So as we move into the audience questions segment,  we'd like to remind you that this session is being recorded, and if you'd like to ask the panelists  a question, please do so using the Google form that we provided. And we do have some questions in  there. Do you want to ask the first one, John? Our first question is from Elizabeth  Blythe-Lee, from Arizona State University Online. Her question is: “How do you see AI  being used to support personalized learning

and supporting UDL? What will that look like?” Alright, I raised my hand so I get to go first. So one of the things that I see it doing… if you guys  have ever heard of the concepts of choice boards, or of the idea of student agency, where  students have a choice, sometimes it's difficult for new faculty, new instructional  designers, to come up with enough choices to

make learning personalized. And so one of the  things that I've seen happen, whether you're using the built in AI idea generator or content  generator in Pearson or Cengage or LearnWorlds, or any of the ones that are about to come out  in our LMS colleagues, they will give you ideas

of potential projects that may suit different  types of learners. And what I've always done in my classes is I've created a final project  that has allowed students to make a choice: which modality suits your strengths best, pick  this modality versus that modality versus that modality, but I can only come up with so many, and  so using AI to not only come up with the ideas,

but say, “AI come up with these ideas and build  the rubric.” It's a huge time saver for me. And so I already had the ideas about potential projects,  but AI has enhanced that and truly made it more personalizable for my students going forward. Many of you might know the concept of differentiated instruction, or  DI, it started in the K-12 realm, and a lot of us in higher education are  doing this as well. It's very difficult to do

differentiated instruction when you're teaching  a lot of people. Differentiated instruction is paying attention to patterns that appear among  the learners who are in your class now and then doing design work in how you respond to them. To  pay attention to those patterns and respond to those patterns. Universal Design for Learning is  what we do before day one of our online courses, it's how we design, not knowing who's going to  be there, and so we assume that there's going to

be wide variability. Differentiated instruction,  on the other hand, is what we do after day one, and artificial intelligence helps us, not only  in the proactive design in terms of UDL, how do we give people more on-ramps to get started? How  do we give people more representative samples or methods or means for the content? How do we give  them more than one way to show what they know?

The flip side of the coin is also intriguing. We  can design in opportunities to ask students to use artificial intelligence to do personalized,  differentiated instruction for themselves, and then share that information with us so  we start seeing the patterns more clearly. Our next question comes from Marcus Popetz  from Harmonized Learning. And the question is: “How do you feel about the AI note takers  that record the class and offer recall and

quiz questions to help the student? Is the  note taking and creation of questions the important part, or the quizzing and recall.” I appreciate this question a lot. So first of all, I just want to acknowledge that we are having this  conversation in a context where being recorded in our classroom can feel unsafe for some instructors  and some disciplines. And so I think that there is reasonable concern from faculty when a note  taking AI tool might be recording everything

they're saying, getting a transcript and then  generating questions. Because many of us feel quite rightly, under increased scrutiny. And so  I sympathize with the idea behind this question, that the summarizing and the question  generation is an important part of learning, but it may not be as accessible to all students as  we would want it to be. And I think to what we've been talking about here throughout this panel,  is that we want to be accessible to learners

at different stages of their learning. And so  to say no, you can't do this, because question generation summarizing is an important skill of  learning, might then preclude some students from ever getting to that higher-level thinking. So  I'm personally not comfortable with saying no, but to engage in some reflection about what  is our resistance, where is it coming from, and how can we think about this in an as inclusive  of a way as possible, of different learners?

So I was sharing with Tom right before our  presentation, that I am doing something for the first time in 24 years of teaching that  I never thought I would do. I'm excited to no longer be in administration, but this  semester, I'm teaching eight classes, and somehow I've survived it. And so I just want  to point that out, that in those eight classes, across eight different classes themselves, I  may have 60 students with ADA accommodation

letters. It's a lot. It's grown significantly,  and out of those 60 accommodation letters, at least 75% have note taking as a core component  of their varied accessibility needs. Again, I want to throw this back to executive function,  which, regardless of what type of disorder you have, executive functioning, your ability to pay  attention in class and take notes is likely to be impacted if you have any kind of diagnosis, and if  you don't have a diagnosis, and by chance you have

trauma in your background, your executive function  is also negatively impacted. So you might not have a formal diagnosis, but you need these note  taking functionalities, as Liz kindly pointed out, in order to even have a level playing field. So  yes, in psychology, we talk about some things that get very uncomfortable, and more than myself, I'm  more protective of the other students’ comments

and those being shared in the note taking. It's  not myself, it's what's being disclosed in our classroom, because these are individuals being  trained to become clinicians, and they learn best

by sharing their own stories. So I'm protective of  it, but in the same way, I want to recognize the fact that it's important for everybody to see that  these are useful tools for individuals who may and may not ever be diagnosed in order to have a level  playing field in their learning experience, We have a question from Richard Powers from City  Colleges of Chicago, and his question is Beth Stark and Jérémie Rostan developed Ludia, an AI  tool that reviews lesson plans, syllabi and other

documents through UDL lenses. Reactions have been  really good. Do you see educators using specific

UDL review tools such as these in the future? Yes. No, the Ludia tool is a splendid thing. It was developed just at the end of last year in late  2023 and what the researchers, whom Richard is referring to, what they did was they took a chat  bot model and customized it only on a data set of universal design for learning documentation, so  if you're putting in “Hey, I have this challenge, or here's this barrier, or my students are having  these kinds of challenges,“ Ludia will say, “Hey,

this sounds like this is the barrier. This sounds  like applying this particular checkpoint from the universal design for learning guidelines applies.”  And I use the word checkpoint here because it's still on version 2.0 of the UDL guidelines,  they're working right now to update that to 3.0. I can see that one of the have and have not barriers  that is likely to be lowered in the near future is that we're in the adoption curve with artificial  intelligence, where the tools are still designed

for expert users. They're still designed for the  coders, the instructional designers. They're not designed for everyday folks. Here I'm going to  hold up my mobile device. Think about every app on your phone. All of those apps started out as  something that only people with specific skills and knowledge could use, and they've now been  designed into things that everyday people can do

and use and participate with. So that's going to  happen with artificial intelligence tools as well, and that have/have not divide right now, the  people with all the money and the specific business cases they get to train their AI on  sort of higher quality or narrower niche kinds of data sets. And Ludia is a wonderful step  in that direction because it's open source and everyone can use it. So good plug for that. Thanks for your great questions. I know we

didn't get to all of them, but we got to  the majority of the ones that came in. So thank you. We always wrap up by asking: “What's  next?” and we'll start with Liz for this one. Okay, I actually want to get some clarification  on this question, because I don't know if “what’s next” means for me or for this topic. We always leave it on our podcast to be very open.

It can be like, I'm gonna go eat lunch.. …or I'm going to Disney World… Yeah, or it can be the big existential  question, so it's really up to you, Liz, or it could be both/and… either/or. My particular flavor of neurodivergence is autism, so I just need, like, some clear expectations.  So what's next? I think my relationship to AI

continues to evolve. I think when it first came on  the scene, I was curious, then I felt overwhelmed, and so I just started saying, “but the  environment” to avoid thinking about it very much. Now I'm kind of in this uneasy frenemy  sort of relationship with generative AI, where I'm sort of like, “okay, I'm curious, but also kind of  skeptical.” So in terms of what's next, I'm just gonna say our relationship continues to evolve,  and I'm not quite sure where it might end up.

And from my perspective, I also feel like we're  gonna see some big things happen. My big takeaway that I would love for you to think about is  remembering that underneath AI are humans. The intelligence of AI comes from humans. And  I think there's some real value in solving big world problems by the combination of all of  our human intelligence together. There's some major issues that we've not been able to solve in  medicine and other areas that I think we're gonna

see AI actually give us some solutions for.  So keep an eye out for that. And I think it's dependent on us. It's not AI that's solving it.  It's our ideas joining together to find really big world solutions that we can't find unless  everybody puts their information in. And if the question is, “what's next?” we  will soon see a diffusion of innovation

shift around AI use in UDL practices. We  will move from training people how to craft effective prompts to how to use AI tools  to shortcut already expert processes. Well, thank you all for joining us.  Thank you to our experts on the panel, for providing us with your responses, and  thanks to everyone who's attended. [APPLAUSE] Thank you. Safe travels home too. 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.