What If Accessibility Built Better Learning For Everyone

WEBVTT
00:00:29.459 --> 00:00:30.019
Hello, everyone.
00:00:30.019 --> 00:00:31.460
Thank you for tuning in.
00:00:31.460 --> 00:00:38.899
My name is Jill Anderson, and I'm a senior instructional technologist with Delta, and I'm also a member of Delta's accessibility team.
00:00:38.899 --> 00:00:41.619
I'm excited to introduce our two guests today.
00:00:41.619 --> 00:00:48.659
We'll be talking about their work on improving the accessibility and usability of some video resources for physics undergraduates.
00:00:48.659 --> 00:00:52.579
Now, this project was a part of Delta's course accessibility grant.
00:00:52.579 --> 00:01:04.579
Instructors who apply for and are accepted for this grant receive $2,000 and a semester's worth of support and consultations to tackle a more challenging accessibility issue they've encountered in a course.
00:01:04.579 --> 00:01:08.340
And you can find a link to the information about this grant in the show notes.
00:01:08.340 --> 00:01:09.700
Our first guest is Dr.
00:01:09.700 --> 00:01:10.579
Stephen Lane.
00:01:10.579 --> 00:01:18.420
Stephen teaches in the physics department here at NC State and coordinates the algebra-based physics 211 and 212 courses.
00:01:18.420 --> 00:01:25.859
His background is in experimental and applied optics with excursions into nanoscience and biochemical applications.
00:01:25.859 --> 00:01:33.219
And then we're also joined by Kylie Masser a third-year student majoring in biology with a minor in microbiology.
00:01:33.219 --> 00:01:36.739
And Kylie serves as the club president of Earth, Wind, and Yarn.
00:01:36.739 --> 00:01:38.179
And she was hired by Dr.
00:01:38.179 --> 00:01:40.819
Lane to work on the project we're talking about today.
00:01:40.819 --> 00:01:43.459
So, Stephen and Kylie, welcome to the podcast.
00:01:44.099 --> 00:01:45.299
Thank you for having us.
00:01:46.500 --> 00:01:48.019
All right, so let's start with you, Stephen.
00:01:48.019 --> 00:01:54.739
Can you just start by telling us about the physics course that you teach and the specific challenge that drove you to pursue this project?
00:01:55.859 --> 00:02:04.339
So the course that uh the courses that I teach and coordinate are physics 211 and 212, which are the algebra-based physics courses.
00:02:04.339 --> 00:02:12.979
And these are taught for students that are in science typically, but not physics and not engineering.
00:02:12.979 --> 00:02:20.740
So we have a very wide range of backgrounds, environmental science, biological sciences, you name it.
00:02:20.740 --> 00:02:29.140
And we've been working on making this course more accessible, or starting to work on making the course more accessible.
00:02:29.140 --> 00:02:34.659
And portions of this course are in videos.
00:02:34.659 --> 00:02:36.340
It's slightly flipped.
00:02:36.340 --> 00:02:38.900
So we have pre-class videos that students watch.
00:02:38.900 --> 00:02:46.419
So there are recorded lectures, recorded PowerPoints with voiceover, and there are recorded worked problems.
00:02:46.419 --> 00:03:04.500
And some of these are quite challenging to turn into something that is accessible, particularly worked problems, because you have something where somebody's writing on a piece of paper, narrating what they're doing, sketching diagrams, using math, and so forth.
00:03:04.500 --> 00:03:14.659
And so we pursued this Delta grant to try and figure out a nice way to efficiently make these more accessible, to convert these in a good way.
00:03:17.139 --> 00:03:23.460
Can you describe a little bit about what you felt like were the inaccessible aspects of those videos?
00:03:24.740 --> 00:03:24.979
Sure.
00:03:24.979 --> 00:03:27.699
So the the videos they range in quality.
00:03:27.699 --> 00:03:31.699
Some are 720p, some are 360 and so forth.
00:03:31.699 --> 00:03:34.819
So you don't have infinite resolution.
00:03:34.819 --> 00:03:44.259
And there's potential challenges for somebody who has who needs something more accessible that maybe you can't see clearly what's going on.
00:03:44.259 --> 00:03:54.659
Or even if you can see, maybe you can't hear what's going on, because there's a continuous flow of visual presentation and audio presentation that goes along with it.
00:03:54.659 --> 00:03:58.979
It's as if you were sitting across from somebody who's sketching on a piece of paper.
00:03:58.979 --> 00:04:04.500
And it's a very conversational way to learn, and it's very good in that in that sense.
00:04:04.500 --> 00:04:08.900
But there are lots of avenues where you would lose something completely.
00:04:08.900 --> 00:04:16.500
Or even if you can't see very clearly, maybe you can't tell what the math is, what's being indicated in the diagram, and so forth.
00:04:16.500 --> 00:04:24.019
So there are a few moving pieces that all need to be converted to a format where somebody can understand what's going on.
00:04:24.740 --> 00:04:26.980
Oh, yeah, that does sound like a really interesting challenge.
00:04:26.980 --> 00:04:32.740
Is this a really common way for sort of STEM instruction to happen in your experience?
00:04:32.740 --> 00:04:35.780
These sort of work through problems that are handwritten.
00:04:35.780 --> 00:04:37.939
Is that a common method of instruction?
00:04:38.340 --> 00:04:48.259
Well, I'm not sure about in terms of flipping, like having these videos and so forth, but my experience in physics quite a bit is yes.
00:04:48.259 --> 00:05:03.860
A lot of physics instruction in my experience has been somebody writes on a board, a chalkboard, a whiteboard, what have you, it describes what they're doing because there's usually a lot of math, and that's hard to format nicely.
00:05:03.860 --> 00:05:13.779
And there's usually a lot of trying to figure out what the physical application is, so whether that's light, sound, you know, electricity, whatever it is.
00:05:13.779 --> 00:05:19.219
There's there's some attempt to diagram the real world and match this math onto it.
00:05:19.219 --> 00:05:29.620
So I would say it's fairly common, even if it's not done explicitly when in a video for a course, it's probably how the physics is taught in the classroom.
00:05:30.500 --> 00:05:34.740
So, Kylie, you're an undergraduate in biology and biological sciences.
00:05:34.740 --> 00:05:36.420
So, how did you get involved in this project?
00:05:36.740 --> 00:05:51.540
When I first saw this listing, I was immediately drawn to it because I'm someone that I don't strictly need accessibility resources, things like captioning, things like clear videos, but it is something that I find incredibly helpful.
00:05:51.540 --> 00:06:03.459
And because I have that side-by-side where I can see the accessibility resources versus the video that they're based off of, I can immediately tell when something's off, when things like the captioning isn't quite right.
00:06:03.459 --> 00:06:05.860
And I find that super frustrating.
00:06:05.860 --> 00:06:12.579
And I can only imagine that for someone that does really need these resources, it has to be even more frustrating for them.
00:06:12.579 --> 00:06:26.740
So having the opportunity to fix these problems and make something that not only the people that desperately need these resources can use, but also pretty much everyone in the class can use, I found that to be a really exciting opportunity.
00:06:27.620 --> 00:06:33.060
So it sounds like you had taken the particular courses before that these videos were being used for.
00:06:34.259 --> 00:06:37.699
Yes, I had just finished my previous semester.
00:06:37.699 --> 00:06:39.379
I took physics to eleven.
00:06:39.379 --> 00:06:44.659
And while I was working on this project, I was actually in the follow-up course, Physics 212.
00:06:45.219 --> 00:06:45.540
Okay.
00:06:45.540 --> 00:06:47.860
And so you said something about when you saw the listing.
00:06:47.860 --> 00:06:52.420
So it sounds like there was a job posting for some for a student to help work on this kind of project.
00:06:53.139 --> 00:06:56.019
Yeah, it was after I'd finished Physics 211.
00:06:56.019 --> 00:06:56.420
Dr.
00:06:56.420 --> 00:07:00.180
Lane had put out a listing for three different positions.
00:07:00.180 --> 00:07:07.060
I don't recall what the first two were, but one of them was this position making the accessibility or the accessible transcripts.
00:07:07.060 --> 00:07:12.819
And because I just finished physics 211, I was more or less qualified.
00:07:13.860 --> 00:07:19.540
Haven't you done any kind of accessibility type work before, or is this a completely new area for you?
00:07:19.939 --> 00:07:21.540
This is something completely new for me.
00:07:21.540 --> 00:07:25.939
It's something I've always been interested in, but it's not something I'd ever tackled before.
00:07:26.339 --> 00:07:27.459
All right, great.
00:07:27.860 --> 00:07:42.100
So, Steven, going back to sort of the goal of the project, you wanted to make these videos more accessible, particularly these problem walkthroughs, understanding that there was visual content and there was audio content, and all of that needed to be really available to everyone.
00:07:42.100 --> 00:07:53.060
At the outset of the grant, when our team and you and Kylie started working, what did you envision as the final sort of more accessible product that you might end up with?
00:07:53.620 --> 00:08:02.740
So, what we started out looking for was a way to sort of make sure that the accessibility box was ticked.
00:08:02.740 --> 00:08:11.219
And we had a lot of math in these videos, a lot of talking, and a lot of diagramming that went along with it.
00:08:11.219 --> 00:08:22.180
And what we'd envisioned at some point was converting this all to a text document, very similar to what you might imagine for a transcript or a closed caption or something like that.
00:08:22.180 --> 00:08:27.300
If you were to watch a movie, uh Netflix sometimes has description of the action that's going on.
00:08:27.300 --> 00:08:29.219
Something like this.
00:08:29.219 --> 00:08:39.779
And we we weren't necessarily going to use AI, but it was something that had come up, and we thought, you know, let's mess around with it.
00:08:39.779 --> 00:08:49.060
And I think Kylie played with a few different LLMs and got different results out of each of them and said there's something here we can we can leverage it.
00:08:49.060 --> 00:08:52.820
And I said, all right, let's let's run with it and see what we get.
00:08:52.820 --> 00:09:02.580
And so we ended up moving towards a a document that was, I would say, a tremendous amount of text, just uh intractable amount of text.
00:09:03.220 --> 00:09:06.019
So was that document that was an intractable amount of text?
00:09:06.019 --> 00:09:09.939
It sounds like that's not what you thought would be a useful end product.
00:09:09.939 --> 00:09:11.860
Is that is that fair to say?
00:09:12.820 --> 00:09:14.100
Yeah, I think that's correct.
00:09:14.100 --> 00:09:25.379
So there was a moment when we sat down and looked at this document and said, Well, we've got all the words that the person explaining in the video has.
00:09:25.379 --> 00:09:28.899
We we we've got their speech converted to text.
00:09:28.899 --> 00:09:42.259
We've got descriptions of the images here, we've got descriptions of the math, we've got descriptions of what their pen is doing as they point to different equations and so on and so forth.
00:09:42.259 --> 00:09:47.620
And it's just a real wall of information that's very difficult to get anything out of.
00:09:47.620 --> 00:09:51.299
And I think part of that is uh the different medium.
00:09:51.299 --> 00:09:58.659
So when you have a video, you sort of watch it beginning to end and you can have something that's sort of conversational.
00:09:58.659 --> 00:10:05.220
It's a very personable way to learn, and it's a very personable way to interact with somebody if you're there in person with them.
00:10:05.220 --> 00:10:15.460
But it doesn't move well into a document format where you can jump around a lot and where things can be shown in a different manner.
00:10:15.460 --> 00:10:39.539
And uh we made a decision to abandon our rigid fidelity to the video in favor of having something that was useful to people in and of itself, instead of having a video and a document that was uh, you know, uh accessible, I suppose, but kind of garbage.
00:10:39.539 --> 00:10:46.100
Having a video that's pretty good and a document that stands on its own and is also pretty good.
00:10:46.100 --> 00:10:52.740
And you could get to a point where you could say, you know what, maybe somebody wants to learn from the document because it fits what they're doing.
00:10:52.740 --> 00:10:58.259
They want to jump to a particular place in the document as opposed to watching the videos sequentially.
00:10:58.259 --> 00:11:07.939
So abandoning that rigid fidelity to the video, I think allowed us to produce something that was more useful all around.
00:11:08.580 --> 00:11:19.299
That kind of reminds me of what Kylie said at the beginning when she said she doesn't necessarily need the captions, but she kind of enjoys or finds benefit from using them.
00:11:19.299 --> 00:11:38.500
And so it reminds me of sort of this universal design principle where you're going beyond just sort of ticking that box as you described it and making another resource that conveys the concept and conveys the problem-solving strategies and process in another way, but that anybody can use.
00:11:39.220 --> 00:11:39.860
Absolutely.
00:11:39.860 --> 00:11:53.059
You know, uh, I think there's been a rise in people using captions to watch videos, you know, perhaps because they can watch it in a noisier environment, whether it's a movie, YouTube, what have you.
00:11:53.059 --> 00:11:59.379
And I think there is a real opportunity here to make something that's useful to a lot of different people.
00:11:59.379 --> 00:12:05.460
Maybe it helps to have higher fidelity visual representation.
00:12:05.460 --> 00:12:09.939
Maybe you actually use a screen reader and you need it to be processable that way.
00:12:09.939 --> 00:12:13.860
Maybe you can't hear so well and you want certain things in caption.
00:12:13.860 --> 00:12:22.419
So by producing a really good quality resource, like you say, universal design and learning, you can make it useful to a lot of different people.
00:12:22.419 --> 00:12:29.460
It's not specifically and only useful for, say, somebody who can't see or somebody who's deaf.
00:12:29.860 --> 00:12:39.139
Kylie, so you are the person who really kind of did a lot of the hard work of sort of manhandling all this content and figuring out how to make it into something accessible.
00:12:39.139 --> 00:12:39.620
Dr.
00:12:39.620 --> 00:12:49.139
Lane mentioned that there was spoken word and that there were some figures and images, and that there were some math equations as well.
00:12:49.139 --> 00:12:53.620
What did you do with the captions to make sure that what was being said was helpful?
00:12:54.019 --> 00:13:04.980
Yeah, the first step for the captioning was pulling the auto-generated captions straight from YouTube, which YouTube does a really great job actually of making pretty good auto-generated captions.
00:13:04.980 --> 00:13:08.259
There are some things where it's going to mishear words.
00:13:08.259 --> 00:13:17.379
At one point, it had misheard something as happy 2001 and like happy New Year's, which obviously not what's being said in a physics video.
00:13:17.379 --> 00:13:23.220
There are times when it hears the like scratching of the pencil on the paper as laughter.
00:13:23.220 --> 00:13:27.700
And so it makes it sound like there's a really funny joke in the middle of this video that's not actually there.
00:13:27.700 --> 00:13:42.659
So the first step is just going through and fixing up those captions, making sure that everything is accurate to what's being said and adding in some punctuation, which isn't present in the original captioning, but does make it a lot easier to read through the captions and figure out what's going on.
00:13:42.659 --> 00:13:57.059
And then once those captions are fixed, they're accurate and they're timed correctly, I can pull those straight from YouTube and start going through and editing them so that as you're sitting there and reading them without listening to the video, you're still getting all of the right information that you need.
00:13:57.059 --> 00:13:59.779
And it's easy to read without seeing that video.
00:13:59.779 --> 00:14:07.059
Because a lot of the times in these videos, there are parts where he's saying this has to equal this, and things are being circled on the screen.
00:14:07.059 --> 00:14:10.019
And the whole time, none of it's really being explained.
00:14:10.019 --> 00:14:13.620
So if you can't see what's being circled, then you have no idea what's going on.
00:14:13.620 --> 00:14:25.460
So part of it is going through and making sure that all of these very vague parts are clarified more in the text versus needing the diagrams and whatnot to understand what's going on.
00:14:25.860 --> 00:14:27.860
So I have a question about the captions then.
00:14:27.860 --> 00:14:34.899
It sounds like there were, you know, these in these particular videos, there were math equations being written and described.
00:14:34.899 --> 00:14:50.820
And I'm assuming that then the transcript, you know, if it said four minus two is two, it wrote out the word minus and wrote out the word equals, which is not really our standard way of reading or looking at and interpreting math, right?
00:14:51.220 --> 00:14:51.539
Yeah.
00:14:51.539 --> 00:15:11.620
A lot of the ways that I got around that was anytime that the captioning would be describing an equation, I would just remove a lot of that and just put the equation in the document because we do have things like MathML, where a screen reader can look at this and read out exactly what's supposed to be said in this equation.
00:15:11.620 --> 00:15:20.100
And it's a lot easier to read the equation versus trying to parse the two minus two equals zero that written out in text.
00:15:20.100 --> 00:15:29.460
So it's just a lot easier to have it written out exactly as it would be written out on a piece of paper in a way that is both visually easy to read and accessible to a screen reader.
00:15:29.779 --> 00:15:31.379
Yeah, so you mentioned MathML.
00:15:31.379 --> 00:15:34.820
And for those listeners who aren't familiar, that's a markup language.
00:15:34.820 --> 00:15:40.659
It's sort of like HTML, but it has some other special tags and coding that describes mathematical notation.
00:15:40.659 --> 00:15:52.580
And it's used to put mathematical notation on the web, for instance, so that a screen reader can read it accurately and someone can sort of navigate around in it and sort of work with it a little bit.
00:15:52.580 --> 00:15:56.659
But since we started talking about those equations, let's talk about that a little bit.
00:15:56.659 --> 00:15:59.939
What was your sort of workflow for these equations?
00:15:59.939 --> 00:16:08.259
Did you start with just the audio, or how did you go about making those equations into something accessible?
00:16:08.580 --> 00:16:13.059
Yeah, Jill, you had actually just given me some pointers on how to improve my process.
00:16:13.059 --> 00:16:18.019
So this is something that we're continuously working on as we figure out new better ways to do it.
00:16:18.019 --> 00:16:26.340
But currently, the process that I had created was to take just a screenshot from the video where it doesn't even have to be that great of quality.
00:16:26.340 --> 00:16:32.659
I've had good results from just like a 360p screenshot, which I could barely read.
00:16:32.659 --> 00:16:50.019
And then I give that to Gemini and I say, turn this math in the screenshot into LaTeX, which is another markup language that's commonly used to write physics equations, but it isn't as accessible to screen readers as MathML is.
00:16:50.019 --> 00:16:55.700
And then I would take this LaTeX and I would give it to a tool called Equatio.
00:16:55.700 --> 00:17:00.580
And with Equatio, I can say, turn this LaTeX into MathML.
00:17:00.580 --> 00:17:09.299
And Equatio makes it really easy to go in and not edit not only the LaTeX code, but also the math directly.
00:17:09.299 --> 00:17:15.700
So this is something that's accessible to someone who has no idea how to write this code, which I certainly don't.
00:17:15.700 --> 00:17:23.380
It makes it very easy to just go in and do exactly what you need to without having to learn all of the right tags that you would have to use, all of the right coding lines.
00:17:23.380 --> 00:17:38.900
And then once you have the equation that you need, you can export that as MathML and put it right into Word, which Word is actually able to natively render MathML into the proper write-up that anyone can look at and read and still be accessible.
00:17:39.300 --> 00:17:40.500
So you mentioned LaTeX.
00:17:40.660 --> 00:17:48.660
Stephen, can you talk a little bit about how LaTeX is used in physics, like sort of what that is for those of us who may be new to STEM content?
00:17:49.700 --> 00:17:50.100
Sure.
00:17:50.100 --> 00:18:00.820
So LaTeX is very commonly used and almost a cultural thing in a lot of physics and computing science, which historically have quite a bit of overlap.
00:18:00.820 --> 00:18:05.860
And it's a very, very powerful, it's not a word processor, it's a typesetting language.
00:18:05.860 --> 00:18:12.260
And it lets you show complicated equations and things that are very challenging to show up in Word.
00:18:12.260 --> 00:18:19.060
So if you wanted to put a great big integral into Word, it would take you a long time.
00:18:19.060 --> 00:18:21.940
And in LaTeX, you can do it very, very powerfully.
00:18:21.940 --> 00:18:26.500
And LaTeX also gives you a lot of control over how things are set up.
00:18:26.500 --> 00:18:38.980
And so it's not really a surprise that when Gemini analyzes these physics problems, it produces equations that are in LaTeX because it's widely known as a very portable thing.
00:18:38.980 --> 00:18:41.940
Unfortunately, it's not accessible.
00:18:41.940 --> 00:18:48.820
And so this is one of the things we were running into and where it was very useful to be able to talk to DELTA about this.
00:18:48.820 --> 00:18:54.660
Because we were hitting a point where we knew LaTeX wasn't accessible, but we didn't have other good options.
00:18:54.660 --> 00:19:02.820
And we're starting to say, all right, do we do we just say this is as good as it gets and run with it, or do we try and try and solve it further?
00:19:02.820 --> 00:19:04.340
Do we try and find plugins?
00:19:04.340 --> 00:19:06.660
You start to hit a bit of a dead end.
00:19:06.980 --> 00:19:14.420
Yeah, and so I want to mention that Equatio, the tool that Kylie mentioned that she used, NC State has an enterprise license to this tool.
00:19:14.420 --> 00:19:22.260
It can visually scan math to produce accessible versions, provides editing and conversion tools between different math formats.
00:19:22.260 --> 00:19:36.100
And recently, Kylie and I have been talking about streamlining the process even further, where she may be able to go straight from her screenshot, if it's a little bit grainy, try out to see whether it can be read directly by Equatio or not.
00:19:36.100 --> 00:19:40.900
But Microsoft Word and Microsoft PowerPoint equation editors do a pretty good job now.
00:19:40.900 --> 00:19:46.900
If you just enter in LaTeX, they will render it accessibly as an equation object on a screen.
00:19:46.900 --> 00:19:53.620
So all of these technologies are really, really getting better by the day, and we're getting more and more of them.
00:19:53.620 --> 00:20:00.900
So it's really cool to see how you were able to use these and leverage them for this work that you're doing.
00:20:00.900 --> 00:20:03.940
And last highly, just those those images and figures.
00:20:03.940 --> 00:20:05.380
I know that that was something else that Dr.
00:20:05.380 --> 00:20:07.220
Lane mentioned being in these videos.
00:20:07.220 --> 00:20:08.100
What did you do with those?
00:20:08.100 --> 00:20:13.060
Did you leave them in the final document or rely on the descriptions?
00:20:13.620 --> 00:20:18.420
In the original versions of the document back in the early days of the project, there weren't any images.
00:20:18.420 --> 00:20:20.420
It was just all image descriptions.
00:20:20.420 --> 00:20:36.500
And we found that as we moved further away from the transcript being a supplemental document to being a standalone document, that it was much more helpful to have all of these images in the document for people who are able to take a look at and see the images.
00:20:36.500 --> 00:20:46.820
So what I do for these images is a lot of the videos being so grainy and just being handwritten are a little difficult to see what the diagrams are trying to say.
00:20:46.820 --> 00:20:52.100
Even for someone with perfect vision, it can be a little difficult to parse.
00:20:52.100 --> 00:21:05.380
So rather than going into Photoshop and trying to make these diagrams from hand again, which would take probably hours per video, you can just give the image to Gemini and say, give me a clearer version of this image.
00:21:05.380 --> 00:21:14.420
And for the most part, it does an incredible job of like 30 seconds generating a really nice, very crisp image that is much easier to parse.
00:21:14.420 --> 00:21:20.580
Every once in a while, you have to do some tweaks, tell it to change this letter to this letter because it misread it.
00:21:20.580 --> 00:21:22.660
But for the most part, it does a great job.
00:21:22.660 --> 00:21:27.780
There have been some difficulties along the way trying to just learn the different quirks of Gemini.
00:21:27.780 --> 00:21:39.700
I found that the more images you give it, kind of the more polluted its understanding of the images becomes, or it starts to integrate pieces of past images into the image it's now trying to generate.
00:21:39.700 --> 00:21:45.860
So every once in a while, I just have to restart the chat and give it just a fresh blank slate.
00:21:45.860 --> 00:21:53.300
And then it starts picking right back up where it's able to perfectly generate that image with just a few tweaks to get it where it needs to be.
00:21:53.300 --> 00:22:03.060
So it's something that takes a little bit of understanding of where Gemini can go wrong, a little bit of playing with it to understand what the best way to go about generating these images is.
00:22:03.060 --> 00:22:16.260
But once you kind of have those understandings of those quirks, it becomes a very fast, very powerful process that saves probably the most time in this process is just being saved by using Gemini to generate these images.
00:22:17.460 --> 00:22:34.980
It sounds like your process included maybe a lot of sort of left turns, right turns, a couple of dead ends, turn around and try something else, experimentation, you know, not necessarily sort of, I know there's a Process for this, I just kind of like plug and chug my way through it, but a lot of iterating on different problems.
00:22:34.980 --> 00:22:38.500
So I think that's really great that you were able to do that.
00:22:38.500 --> 00:22:47.140
And I think as an undergraduate, being able to help in this project, I'm sure that that was super valuable, time saving for you to be able to use a lot of these tools.
00:22:47.140 --> 00:22:55.140
Stephen, what were the most valuable and unexpected benefits of having Kylie as an undergraduate student work on this particular project?
00:22:56.100 --> 00:22:56.340
Right.
00:22:56.340 --> 00:23:02.660
So there's always a challenge when you take a student, especially if it's a shorter project.
00:23:02.660 --> 00:23:11.700
There'll be a certain amount of time to get somebody up to speed, and then they start producing things, things that are really useful and move the project forward.
00:23:11.700 --> 00:23:16.020
And there's always the question: is it going to be worth it having an undergraduate?
00:23:16.020 --> 00:23:18.340
And I think in this case, it was.
00:23:18.340 --> 00:23:25.700
One thing that Kylie has alluded to is trying all these different things, being creative.
00:23:25.700 --> 00:23:34.740
And so as an undergraduate, Kylie has a different perspective on the challenges and on the technologies that I have.
00:23:34.740 --> 00:23:36.980
I have a lot of other things to get done.
00:23:36.980 --> 00:23:40.100
And so I'm very, all right, what's what's a pathway through this?
00:23:40.100 --> 00:23:42.340
Let's take it, let's let's do it.
00:23:42.340 --> 00:23:49.220
I also learn new things a little more slowly, perhaps, and AI is a little newer of a thing.
00:23:49.220 --> 00:24:00.580
Whereas Kylie has this approach where she could say, I'm going to try this, I'm going to try that, I'm going to throw a whole bunch of things at the wall, see what sticks, I'm going to do this and see if it works or if it doesn't.
00:24:00.580 --> 00:24:05.700
And from the beginning of the project, that was something that was obviously helpful.
00:24:05.700 --> 00:24:12.740
Her willingness to just try one of the other AIs and see, is it better, is it worse, and so forth.
00:24:12.740 --> 00:24:25.140
And so there was a lot of creativity that went with it, and perhaps a more natural use of these more modern tools than I would have done at the same speed.
00:24:25.540 --> 00:24:35.220
And then Kylie, turning to you, what did you feel like was sort of the most valuable part of this in terms of your own professional development and this point in your undergraduate career?
00:24:35.860 --> 00:24:47.540
I think that this has given me a lot of skills that even going forward as a biology major going into other science fields, completely unrelated to physics, it's going to help me out a lot.
00:24:47.540 --> 00:25:03.140
I think that the biggest thing was just having this experience in problem solving, in a real world problem, where this started out as a very daunting task where it's just take these videos, make something accessible.
00:25:03.140 --> 00:25:06.740
And I had to really go through and figure out what's the best way to do this.
00:25:06.740 --> 00:25:10.500
I had to bounce off of other people to get ideas of how to go forward.
00:25:10.500 --> 00:25:17.380
And just being able to work through each of these problems, I think is really valuable experience in any career field.
00:25:17.380 --> 00:25:27.540
And I also think that having this better understanding of how to make these accessible resources, I have always been really interested in science communication.
00:25:27.540 --> 00:25:38.740
And being able to produce something that is accessible to the widest amount of people is something that's really important in any science field to be able to communicate these findings to get people interested in your research.
00:25:38.740 --> 00:25:43.380
You have to be able to present it in a way that everyone is able to parse and understand.
00:25:43.380 --> 00:25:46.740
So having this experience is really going to help me going forward.
00:25:47.140 --> 00:25:53.700
And I'd say to add on to that, I think Kylie's interest in this was a key element in the success.
00:25:53.700 --> 00:25:59.460
So when you bring somebody onto a project, you want the right person for the project.
00:25:59.460 --> 00:26:05.380
Kylie, from the beginning, I would say, was almost ideologically motivated with the accessibility stuff.
00:26:05.380 --> 00:26:10.260
So it wasn't a question of getting her excited about it or, oh, look how cool this is.
00:26:10.260 --> 00:26:19.140
It was letting her do what she really wanted to do and getting problems out of the way, finding solutions so that she could do it more efficiently.
00:26:19.140 --> 00:26:26.500
And I think that's that's the real trick in having somebody working on this project is finding the right person for that.
00:26:26.500 --> 00:26:33.220
And we have an advantage with the 211-212 students that we have a lot of people who are motivated.
00:26:33.220 --> 00:26:36.500
And you know, we had we had a lot of applicants for these positions.
00:26:36.500 --> 00:26:41.540
And Kylie's focus on the accessibility stood out from the beginning.
00:26:41.540 --> 00:26:45.060
So I think that's one of the biggest elements of success.
00:26:45.380 --> 00:26:48.740
Well, thank you both for sharing this experience and your work today.
00:26:48.740 --> 00:26:59.860
As we wrap up, I'd love it, Stephen, if you had to sort of coalesce your learnings into your biggest piece of advice for a faculty member who's struggling with working to make their course materials more accessible, what would you tell them?
00:27:00.180 --> 00:27:03.780
I'd say it's it's a challenging task, but it's doable.
00:27:03.780 --> 00:27:11.860
And there might be, I don't know, five things on your list that that need to happen in order for a course to be accessible.
00:27:11.860 --> 00:27:18.740
And maybe two out of those five are just going to be grinding on a problem and until it's done.
00:27:18.740 --> 00:27:22.580
But some of those things, there are tools that will help.
00:27:22.580 --> 00:27:25.140
And I can mostly speak to AI.
00:27:25.140 --> 00:27:28.580
With the transcripts, you know, AI was decent.
00:27:28.580 --> 00:27:32.820
We've had things that can generate transcripts for quite a while.
00:27:32.820 --> 00:27:35.620
The images were a game changer.
00:27:35.620 --> 00:27:40.580
So I'm used to building images, diagrams for papers, and so forth.
00:27:40.580 --> 00:27:44.340
You know, pick your favorite software, and it takes time.
00:27:44.340 --> 00:27:56.500
But with Gemini at least, you can give it a roughly sketched, hand-drawn image, and it will produce something that is actually pretty good quality.
00:27:56.500 --> 00:28:04.420
Sometimes it'll do something really goofy, like it it knows that paper has lines, but it doesn't know that the lines are regularly spaced.
00:28:04.420 --> 00:28:06.500
So it'll do something goofy there.
00:28:06.500 --> 00:28:11.540
But you get a tremendous time savings just from that element.
00:28:11.540 --> 00:28:23.620
And not everything has that win, but there are things like this where you can say, you know what, I've got this problem, it's going to take hundreds of hours, and then bang, it's uh it's gonna take a tenth of that.
00:28:23.620 --> 00:28:36.980
And so my advice would be to find those things, find those efficiencies and leverage them where you can, because it doesn't work for everything, but it works for some things, and that really helps.
00:28:37.780 --> 00:28:48.100
That and that particular tool that you mentioned sounds like it could be helpful for you not only in making things more accessible, but just in creating a slide or a presentation or something like that.
00:28:48.100 --> 00:28:50.660
Anything that you need, anytime you need to render a figure.
00:28:50.660 --> 00:28:54.820
So that's really interesting how all these tools can kind of help us in so many different ways.
00:28:55.220 --> 00:28:55.780
Absolutely.
00:28:55.780 --> 00:29:08.420
And I think I've used Gemini for tweaking images for an exam, where I don't want to recreate the image from scratch and give it to Gemini and say, all right, sort of a voltage for a capacitor, I want this to be a voltage for an inductor.
00:29:08.420 --> 00:29:13.540
And it has enough to go on that it comes up with something and it's it's pretty good.
00:29:13.540 --> 00:29:20.260
You know, you always want to verify it and make sure that it didn't hallucinate something exotic, but a lot of the time it works.
00:29:20.260 --> 00:29:23.060
It's really efficient for certain problems.
00:29:23.860 --> 00:29:29.140
Excellent. Kylie, what would be your advice to fellow students who might have this kind of opportunity?
00:29:29.140 --> 00:29:36.020
Or just, you know, as they may be thinking about, you mentioned science communication, as they may be thinking about communicating in their own fields.
00:29:36.020 --> 00:29:37.780
What would be your advice to them?
00:29:38.180 --> 00:29:45.460
Yeah, I think that my biggest pieces of advice are number one is just don't be afraid to try something new.
00:29:45.460 --> 00:29:52.020
I think that if I had been too scared to kind of step outside of my wheelhouse, I wouldn't have gotten involved in this project.
00:29:52.020 --> 00:29:55.700
Just the fact that it was in the physics department probably would have scared me off.
00:29:55.700 --> 00:29:58.980
So I think that being able to try something new.
00:29:58.980 --> 00:30:10.020
And when you're working on the project, being able to just kind of throw things at the wall and see what sticks and not being able, not being afraid to just turn out all these different iterations to see what's going to work.
00:30:10.020 --> 00:30:11.460
I think that's really important.
00:30:11.460 --> 00:30:15.300
And then also to not be afraid to be wrong.
00:30:15.300 --> 00:30:22.500
Because there were a lot of things that I did throughout this project where I put so much time into working on these iterations.
00:30:22.500 --> 00:30:25.300
And then it was, well, we could do this better.
00:30:25.300 --> 00:30:29.700
We could do something that's more helpful, make something that's more helpful for everyone.
00:30:29.700 --> 00:30:40.180
And you have to be able to take that in stride and say, yeah, how can we improve this rather than focusing on where you went wrong, focus on how you can improve.
00:30:40.500 --> 00:30:48.660
And I think to add to that, you might see professors who are teaching class and they've prepared very cleanly for it, and it's at a level that they know all the answers.
00:30:48.660 --> 00:30:56.420
But when you actually go into research or solving a problem that people don't know the answer to, things are going wrong all the time.
00:30:56.420 --> 00:30:59.540
And it has it has to suck before it gets good.
00:30:59.540 --> 00:31:03.300
Like you make you make a bad presentation before you have a good presentation.
00:31:03.300 --> 00:31:08.580
You have a you have a bad poster before it's a good poster, it's a bad transcript before it's a good transcript.
00:31:08.580 --> 00:31:15.060
And that's not always evident to an undergraduate, but I think it's the reality of research.
00:31:15.060 --> 00:31:24.580
And one of the things that helps people be successful in research is willing to try it and it goes horribly wrong and you try something else or you figure it out and and you progress.
00:31:24.580 --> 00:31:26.580
And it's a really good life skill.
00:31:26.980 --> 00:31:28.180
Absolutely.
00:31:28.180 --> 00:31:30.660
Thank you both for being our guest today.
00:31:30.660 --> 00:31:39.140
I think a lot of our listeners are gonna gain a lot of good advice and interesting information from your work on this project and you're sharing it with us.
00:31:39.140 --> 00:31:42.180
And to our listeners, tune in next time and take care until then.