Interactive 3D-Printed Models for Students with Visual Impairments | Accessibility NYC Meetup Recap

STACY: Hello, everybody! My name is Stacy Smith. And on behalf of the organizers of accessibility NYC, I want to welcome you to tonight's meetup. I want to thank Thoughtbot, the place you are tonight, Equal Entry, Adobe, and Level Access.

And I want to welcome you on the live stream right now. Hello to you all. Many thanks to Joly MacFie from the Internet Society, as well as Mirabai Knight from White Coat Captioning, who is making all of the lovely live captions that we see right now. I invite everybody here that's in the room with us to stick around after the presentation, to mingle, meet everybody that's here, and perhaps have some one on one questions with our presenter.

But without any further ado, we'd like to hand it over to Professor Shiri Azenkot.

SHIRI: Thank you! Hi, everyone. It's nice to be here. Glad we got everything set up.

So my name is Shiri. I'm from Cornell Tech, the new Cornell campus on Roosevelt Island. And some of my students are here tonight, and they can also answer questions about the work that we do. A few words about me.

I've been doing accessibility as my main research area since I started my PhD back in 2009 at the University of Washington. I was working on technology to make mobile devices more accessible to blind people, and then I moved to New York for my job here. And since then, I've been doing several branches of research, but all of them focused on how we can use advances in artificial intelligence to empower people with visual impairment.

So both people who are blind and people with low vision. And I've had some fantastic students along the way. So much of this work is mostly their hard work and effort. Okay.

So today I'm gonna talk about one of the larger projects that we have worked on in my lab. We've been creating these funny things. So these are 3D printed models that you can interact with. So as you explore them, they tell you what you're touching.

All right. So before I go any further, I just want to thank Lei Chiu, who is a graduate student of mine, who has really masterminded this effort. He's not here today, but most of this is his brainchild. So I was there to advise and support.

And Holly Lawson is a professor at Oregon State University, and she's the coordinator of a program who teaches teachers for the visually impaired. So she's a collaborator of ours, whose expertise is in special education. Specifically for students with visual impairments. And then we have many other students who have worked on this project, and actually this slide is not up to date, because Ricardo, who is also here, is also contributing to this project now.

All right. So throughout our education, we are surrounded and inundated with visual content. So there are a few images here on the slide that I'm sure will look familiar to you all. There's an image of the solar system, a map, and a DNA strand.

And to people who have visual impairments, all of this visual content is not accessible. So this is a huge problem. And especially in stem fields. In science technology, engineering, and math.

So what do we do about this? Well, recently, 3D printers have become much more available. So that presented a very exciting opportunity to create customized models to teach concepts to students with visual impairments that were typically communicated through visuals. So, for example, on the screen is a model of...

I actually don't know what this is. I think it's some kind of bone structure? That's kind of weird. To use a more mundane example, you could imagine a strand of DNA being presented as a 3D model to help someone who can't see the visual what that structure is like.

It's something that's very difficult to explain. And that's the case with a lot of information out there. So another example is a map. Right?

So here on the left we have a map -- a pretty simple map of the world. And the continents are marked in different colors. So you can see where each continent is. There's also a label to identify each continent.

So now we can print a tactile version of this. So on the right there's a 3D printed globe, and this 3D printed globe has tactile delineations of all of the continents. So now if you can't see the continents, how they're shaped, how they relate to one another, you can use the globe, you can feel it, and you can kind of tell how the continents are shaped and how they relate to one another.

But there's still some major limitations here. So what's one possible limitation? Mountains? So is that a limitation for the visual or for the globe?

For the tactile? So let's think... If I'm trying to take a visual graphic like the map here, and I'm creating a tactile globe, what information am I not capturing in the tactile version? Yeah.

The labels. So I can feel the delineations of the continents in the globe, but I don't know what they are. What else? Yeah, color.

Yeah. So again, I can feel the delineations of the continents, but the colors really tell me which continent is which. It breaks them up. So it's not just colors and labels.

There's a lot of auxiliary information in educational visual materials that are constantly being used throughout our lifetimes. Beyond just our formal education. So in addition to labels, for diagrams and colors, there's also captions. Right?

Something as basic as also a title. To an image. So how do we... What do we do?

How do we create 3D printed models that can also represent all of this auxiliary information? So this is the problem that we set out to solve. And our vision was creating an interactive model that, as you touched it, as you explored it by touch, it would tell you what you were touching. So, for example, if I'm touching this airplane model, and I touched the wing, the model would somehow detect that I'm touching the wing, and it would tell me: You're touching the wing.

So let me show you what this looks like. Do we have sound?

Southeast by South America and the Caribbean Sea. North America. South America.

More.

North America up. South America down.

SHIRI: All right. So this is actually Lei here, the student working on the project, and he's touching a printed globe model, like the one on the previous slide, and as he's touching it, something mysterious, which I will explain later, is telling him what he's touching. And he can even ask for more information. So he says "more" when he wants to hear more information, and the system tells him that...

Gives him more information about what he's touching. So this was our vision. This is what we wanted to create. Some kind of interactive experience where you can get auxiliary information like this, that responds to your touch when you're exploring a 3D printed model.

All right. So once we developed this vision, the question was: How do we actually make this happen? And there are different ways of sensing your touch on a 3D printed model. So we weren't the first people to investigate this type of technology.

So one method that people have worked on was adding conductive filament to a 3D print. So filament is the material that's used to print a model, a tactile model. This is plastic right here. But there's also a certain type of filament that's conductive.

That can essentially, with some electronics, sense when you're touching it. But the problem with that approach is that you would need a more sophisticated 3D printer, because you would need to print conductive filament in addition to this plastic filament here that we've used. And you would also need to embed some kind of electronics in the model. So there's a computer in here, essentially, that can sense when you're touching and respond to your touch.

By using speech. Like what you saw in the video. So that approach is pretty complicated. And for us -- yeah!

And expensive! Exactly. And it requires specialized materials. So who is going to be able to use this?

People who have access to all those things. And we really wanted to design something that would work for anyone. So we've been seeing that 3D printers are becoming more available in community maker spaces, in schools, in libraries. And we want a teacher or a student or a parent or a friend to be able to go into one of these spaces and to create one of these models.

And we want the barrier of creation to be as low as possible. So we didn't want to follow this approach with conductive filament, because it would require specialized materials and assembly, and probably a lot more expertise. So we nixed that idea. Okay.

So then another approach is actually what we developed originally. So our first stab at this problem was to use acoustics. So we added little extensions to the model, and there's a picture here on the right of the slide that shows one of these little extensions. They're these little components that you can add onto the 3D print that you can kind of flick.

Like a guitar string. You can almost strum it. And it creates a sound. And we designed a set of these that have different sounds.

Different pitches. So then, once we have a model that has these little components attached to different points in the model, and each one generates a different sound, now we can create an application on a phone, an iOS application, that will detect these very specific sounds, and also be able to figure out which sound is which. So now you just need to associate a label with each one of these sounds. And you have an interactive model.

So that was an interesting idea. And it got... We developed the project. We did some user studies, and it got pretty positive feedback, but it also had some limitations.

So... Any ideas as to what those limitations may have been? Yeah. It couldn't work in noisy environments.

Well, not very well, anyhow.

How many of those different sounds did you have?

SHIRI: So we prototyped ten of them. You could prototype more in theory. Right. It's very breakable.

In fact, they did break. So we would carry around this box that we would use to demo. And we always needed extras, in case they would break. Yeah.

In theory, you can create an Android application too. But yeah, you do need a phone, yeah. So... Any other ideas?

There's one more limitation that... At least I can think of right now. What about the size? Yeah.

So they were about a centimeter, I want to say, long. But I think you're getting at something, which is that they changed the texture of the model. Right? They would actually interfere with the tactile components.

So it could be difficult to add those -- we called them tickers -- to add those tickers to different points in the model, depending on what the topology of the model was, and it actually changed the topology too. So that wasn't ideal. We didn't want something to interfere with what we were trying to convey, which is the shape of the concept. Okay.

So... Onto our next attempt. So we decided to use a camera. And all of you are carrying cameras on you right now, I'm willing to bet.

What are they? What are they attached to? Yeah, to your phones. So we wanted to design something that would leverage the camera on a phone or on a computer, but ultimately really just a phone.

Or an iPad. A device that we knew that most people have, and especially teachers. So we wanted this to be an educational tool, first and foremost, and we know that teachers used iPads already with their students. Teachers for the visually impaired, that is.

So we developed two applications. The first one is called Markit, and this is an application where you can select regions on the model. So, for example, I can select the wing. And I can add labels to it.

Wing. And then the second application is called Talkit. And with Talkit, you can interact with the model. That's been augmented with Markit.

So I'm gonna go through now and explain how both of these applications work. Okay. So we're gonna start with Markit, because the first thing you need to do is actually add these annotations to the model. Okay.

The process goes something like this. So the first thing you need to do is find a 3D printed model. You can also design one, if you wanted to. But there are many, many that are available online already.

And then the next thing you would do is add annotations to the model with Markit. Then you would save that as one interactive model. Then you could print the model, and you can learn from it with Talkit. That's the overall process.

All right. So Markit is an add-on to a 3D modeling software application called Blender. Does anyone have experience with 3D modeling applications? Okay.

A little bit. Yeah. So not many people do. And I'll come back to that later.

But basically... There's a software that you install on your computer, and then it allows you to manipulate the 3D model. So the first thing that you do here is you make a slight modification to the model. So we didn't want to make too many modifications.

Like I was saying with tickers, we didn't want to change the shape of the concept. That's what we want to communicate here. Right? But you do need to add this little scaffold.

And that's this thing right here that I'm showing on the plane model. I'll pass those around later. Okay. And the next thing you would do is select the regions on the model and give them text labels.

So on the screen, there's an example of a tank model, and there's already the scaffold that we added previously, over on the top left of the model. And then we color coded the different regions that we've selected to add annotations to. So like I said, this is a tank. So we've selected and labeled the gun turret, the main gun, the road wheel, and the spaced armor.

All right. So that's basically all you need to do. So now you have... You print it, and you have an interactive model that you can use with Talkit.

So let's take a look at how Talkit works now. So Talkit is an iOS application. So we developed it so that it can work on... Actually, we developed also a Macbook application.

So it can work on multiple devices. On a phone, on an iPad, on a Macbook. So you -- first thing you need to do is you need to load the correct model, so the application knows what model you're holding up. So I would load the plane over here.

Then, as you're exploring the model, as you're feeling it, if you want to retrieve information about a certain region, you would tap and hold your finger. That's all you need to do. I need a third hand here. Okay.

And then you could also use speech commands, like "more", if you want to get more information about the item that you were just touching. So... Pretty simple. But how does this all work?

So I'm gonna get a little technical here, and tell you about how we made this happen. So we're using computer vision algorithms. So like I was saying, we're using the camera on the device to figure out how the model is positioned, and to also detect your finger, and figure out where on the model your finger is. And this is all done with this little thing right here.

Okay. So if you think back to my description of Markit, I mentioned we made one little modification to the model, and that was adding this scaffold. So this scaffold is used for this funny little cube, which we call a tracker. So you can just slide this on, and you can print the actual tags here with just a regular printer.

3D printer. And of course, this is... There we go. All right.

It's a lot harder to do when 40 people are staring at you. So... Okay. So now we put this on.

And we have a computer vision algorithm that we use, that can detect these tags right here. And not only can they detect, but they can also figure out the orientation, the angle, and based on that, and what we know about the model itself, we have the digital information, the digital model, because we have that information already. We can figure out where you're touching, how the model is positioned, and where your finger is touching on the model.

And there's one little thing that we need to do. So... We also need to detect your finger. And we do that with a little trick.

We add one of these red stickers to your finger. So... Very simple. And again, I think we bought a thousand of these for $5.

So these are all materials that, once you have a 3D printer, everything else is very easy to get. And once that's all set up, you'll see that the application knows exactly what the annotated regions are, and it will respond to your touch. Okay. So we designed this application.

But now it's really important for us for this to actually be used in practice. So we wanted to know exactly how this could be incorporated in an educational context. And we still have many open questions. So we have our special education collaborator, Holly, but we're not teachers of visually impaired K through 12 students.

I have to remind myself that I'm still a teacher of some kind, but a very different kind. So we wanted to know how these could best be used in a classroom, in an educational setting where a teacher for the visually impaired is interacting with her students and teaching them about certain concepts. So just to give you some background, typically students with visual impairments will have a special education teacher, a TVI, teacher for the visually impaired, who's specifically trained to teach students with visual impairments.

And a lot of the times, these teachers, TVIs, they meet with their students in one on one or small group sessions, and they review concepts that were taught in class. So, for example, if there was a science lesson, and there were a lot of visual materials that were given in class, then the TVI would review them with the student later on, and give them some adapted materials to make the content more accessible, and make sure that the student understood.

And then the teacher also teaches extra content, extra subjects, that have to do with blindness or low vision. So we wanted to know: We've developed this system. And we wanted it to be used in context. So how do we design these interactive models?

How do we design the 3D models themselves? So that they're accessible to the students? How do we design the annotations so that they make sense when you're trying to teach certain concepts? And did we do a good job?

We also just want to evaluate the system that we have. And like any system, it can always be improved. So we wanted to know: What should we do to move forward? So we had all these questions.

And we decided to conduct a study to answer them. So... What we did was two workshops with teachers of the visually impaired. And the backstory behind this is that I was asked to give the keynote at a conference for teachers of the visually impaired.

So I said: Great! Can I also get some research done while I'm there? So there were two conferences that we went to. One here in New York State.

And one in the Pacific Northwest. And so we did these two workshops with TVIs, to try and get some ideas about how to best design interactive models. So we'll go through some of the details here. So like I said, there were two workshops, and we had 16 TVIs.

There were only three men there. So it's a very different scene than technology. Which is what I'm used to, and probably many of you too. So that was in the first workshop.

In the second workshop, we had 19 TVIs. We had 5 males and 14 females. And in advance, we printed a bunch of models for the teachers. So this was just based on our general knowledge and with the help of our collaborator, Holly, what might be a useful model.

The globe model I was showing as an example, we had an elephant model, we had a brain model. So we were thinking... What concepts might be useful and relevant to teachers? To K-12 TVIs?

And then for our actual session, we started out with an introduction, we told the teachers about... We gave them an overview about how 3D printing works, talked to them about the interactive system, and then we asked them to get into groups and to choose the 3D models that we printed, and to figure out how they might make them interactive. So this was all conceptual, because we didn't actually give them the software to do it.

But we gave them a bunch of arts and crafts materials. So we gave them play dough, we gave them post its, poster board for them to write on. So conceptually, their task was to design interactive materials and to also give a rationale. Why did they want these materials to be designed the way they were?

And we also had them try out our application. It was running on a desktop at the time. So you can see from this photo here that some of the TVIs were pretty pleased, and some of them were a little confused. All right.

So what did we find? So it was really fascinating. At the end of the hour and a half, we had the TVIs, each group came to the front, and they presented their interactive models. And they had added a bunch of content to them.

So they used play dough and they had sticky notes and stickers on the models, and they explained all of their choices and all of the annotations that they would add. They also explained how they would modify the actual models, and it was really interesting. So I'll highlight here a few of the comments that they made. So one of the things that they said was: Well, it's great that you guys have added audio labels.

You can add text. But a lot of my students have vision. So they also wanted to have visuals, not just audio. For their low vision students.

And even in terms of audio, they wanted not just verbal descriptions or verbal labels. They wanted to be able to add sound effects too. So... This is a model on the screen of a heart here.

And they also had comments like... Well, we want the model to feel more like the actual object. We want to be able to change the size. We want to be able to give the students some sense of scale.

And they also wanted to be able to show animations. So, for example, if it's a model of a heart, they wanted to add an animation of the blood flowing through the heart. So again, this is really interesting. And I think this gets to something a little deeper, which is that a lot of researchers and people working on accessibility forget that most students with visual impairments -- most people with visual impairments -- have functional vision.

They have vision. And they're not fully blind. So it's a very small minority of people with visual impairments that have little to no vision. All right.

So based on all of this feedback that we collected, we iterated on our application. On Talkit. And we incorporated all of the feedback that the teachers had. So now the application is able to not just provide audio, verbal descriptions, but also sound effects.

It also gives visual information. So visual highlights, to show the difference between the different annotated labels. And it can even show animations. All right.

So that was all very interesting. But again, that study, like any study, had its limitations. So we had asked the TVIs to use models that we had printed. And as I was saying earlier, they had a lot of preferences for how they wanted the physical models to be.

So they said they wanted the models to be a different size, or a different texture. So their comments went far beyond just adding the annotations. It had to do with the whole package. The physical model too.

And also, all of the design work that they did was very conceptual. So they didn't actually have the experience of interacting with the models that they designed. They only wrote down what annotations they wanted. They mocked up the model as best as they could, but they weren't able to experience it, or -- even better -- to use it with their students.

So we did another study. And in this study, we worked with three TVIs. So just three in this case. But we worked with them over the course of six weeks.

And throughout the six weeks, we co-designed, we collaboratively designed, models from the ground up that would suit the students, suit the needs of their students. The student that they currently had. So the way this worked is that my student, Lei, would meet with each of the TVIs, and they didn't have any background in 3D modeling skills. But together, they brainstormed different model ideas.

So the TVIs, for example, would say: Well, I'm teaching my student right now about flowers. So I want to design a model of a flower. And then together they would look online for possible models. The teachers would contribute sketches, or images that they found online, and together, Lei would use something that -- a 3D model that he found, and adapt it, because he had a lot of experience with this by now, and adapt it to suit their needs.

And as they were doing that, they also designed the annotations. So now there was a lot more flexibility in terms of what kinds of auxiliary information you could add to models. So in addition to just verbal information, you had different kinds of audio, and also visual. So together, they also designed the auxiliary information.

And there were three TVIs, and we created three models. So I've brought them here. So the first model is this plane model. This next model is a volcano.

And then the final model added another element. So this is a map of Washington State. And it's actually a puzzle. So the different regions...

Are puzzle pieces, and you can fit them into the model. So it was really important for the teachers to allow the students to take the model apart. And to actually feel the different components. Because it would give them a better sense of how they relate to one another.

So let me show you a short video of how these worked with the different annotations. And then at the end, there would also be time to try a demo.

The volcano model has four interactive elements and features an animation of a volcano eruption.

Crater. Main vent. Secondary vent. Magma chamber.

The magma builds in the magma chamber. It travels through the main vent and the smaller secondary vents and explodes from the crater.

The plane model has seven interactive elements, and three of them have audio effects.

Underside. Horizontal stabilizer. Vertical stabilizer. Wing.

Body. Jet engine. (propeller roaring)

This is your captain speaking. We're flying at an altitude of 37,000 feet...

The map model is a puzzle that has five interactive elements, and each of them has audio effects.

Columbia plateau. Okanogan Highlands. Cascade Mountains. Western Lowland.

Coast.

The teachers used these models in classrooms with their students.

More.

Magma chamber. Magma chamber is a hot bubbling furnace that's formed by a collection of magma underneath the volcano.

SHIRI: So the teachers did use the models with their students that they designed the models for. And we got some good initial feedback. I was actually surprised that the students were able to work with the models fairly independently. And they seemed to really enjoy it.

So moving forward, we want to do more rigorous tests to see how much the models actually help the students learn these concepts. And to use a baseline, like a tactile graphic or a model that's not interactive. So we're going to do that in future work. So just to conclude, this has been really interesting and exciting work, and I think that we learned a lot from it.

So if you think back to the beginning, we had this very narrow vision of what adding labels to a model would be. And what we found is that we really want to give students and teachers as much flexibility as possible in terms of the kinds of multimedia information that can be added. I think it's also really important to lower the barrier of creation. So right now, we've done that a lot with the Talkit application.

But with Markit, with the application I showed you in the beginning, where you mark up the models, where you add the annotations, there's still an incredibly high learning curve there. So I think only one person here said that they had some experience with a 3D modeling application. And those 3D modeling applications are not easy for beginners. So it's completely unrealistic to expect the teacher, who's completely overloaded with tons of work and students, to also become a 3D modeler, in all her copious amounts of spare time.

So the big question now is: How do you create accessible design tools? And it's not just for the teachers who -- some of them are excited. Some of them, quite a few actually have visual impairments -- but for anyone to create these models. It's really important to allow people with visual impairments as well to create these models and to create any sort of new type of genre of media that we're developing.

So that's our next major challenge. Creating fully accessible 3D modeling tools for people with all visual abilities. I forgot to turn that on silence. Okay.

So... Actually, she had really good timing, because the talk just ended. Okay. So now I'm happy to take any questions.

I also have demos here. So everyone can come up and try the demo. I'm sorry?

We have a mic for questions. So I can run around the room. And we have a question.

I came very interested in the 3D printing. But it does occur to me that there's 3D scanning. So maybe there are interactive educational toys that I could scan and get a model on my computer of, and then do this process? Couldn't I?

SHIRI: Yeah. Yeah. Definitely. It's actually...

This talk is less about 3D printing and more about augmented reality. Right? It's a bit of a bait and switch.

Now, what are the components, exactly? Besides a 3D printer? You have the software on the computer.

SHIRI: You mean what components are required to get this whole thing working?

To get it all working.

SHIRI: All right. So at the very minimum you need to install an app, Talkit, on an iOS device. And we don't have it on the App Store quite yet. We're getting really close to being able to release that.

And then you would need to print... You would need to have access to a 3D printer or a 3D printing service. And we are going to be publishing some sample models. So to print one of our sample models, you could download them from online, and then you would also need to print these tags with a regular printer, and you would need to get a red sticker.

That's it.

Oh, okay.

SHIRI: And if you wanted to create a model, that's a little more complicated. We can talk about that.

Oh yeah, that is.

So I'm curious as to whether you had any kind of constraints as far as size and complexity of the 3D models that you could use for this.

SHIRI: Yeah, that's a good question. So constraints for size and type? So there's always a challenge with size and camera placement. Because, again, if you're touching the model, but the camera can't see where you're touching, then it's not gonna detect your finger.

So in terms of... So that's in terms of model placement. Right? The model has to be within the view of the camera, and so does your finger.

So it gets a little tricky. So what we did in our study is we suggested that the teachers prop the iPads or the iPhones on stands. Stands that cost like $10, that we bought on Amazon.

And were there any constraints as far as granularity? I saw for the airplane, for instance, you could mock-up the cockpit and the engines, which are pretty small. The engines look less than a centimeter in diameter. And maybe a centimeter long?

SHIRI: Yeah, so that gets a little tricky. So technically, yes. Right? Because if it gets much smaller than your finger, it'll get tricky to detect.

Also, from a usability standpoint, if it gets really small, it'll be hard for someone to locate.

Right. Makes sense.

SHIRI: Yeah.

Questions in the back? Here I come. It's our friend Thomas. Hi, Thomas!

Thomas writes: Awesome and innovative work! Question. Can other languages other than English be used with Talkit/Markit? Can it support Japanese?

Smiley face. He is in Tokyo asking this.

SHIRI: In theory, yes. Right now our first major milestone is just getting the app on the App Store in English.

So I think this is so awesome, what you're doing. I'm curious if you have thought about... Are there any plans, possibly, to take this outside of the realm of education? Like, to adults or people outside of that?

SHIRI: Yeah. I think that there's definitely a lot of applications beyond just this particular one. Which is K-12 science-related content for students with visual impairments. So, for example, I spoke with this woman a few weeks ago, who I need to get back in touch with, but she was working on a program, where students from disadvantaged backgrounds were designing cities.

They would get together in the summer program, and they designed their little cities, and it was a way for them to learn about urban development and economics and all this other stuff. And they were building these models of the cities, and I told her: Well, what if we could add audio annotations? You could point to different objects, different things in the cities, and they could add descriptions to them, and she thought that was an amazing idea.

So I thought that would be another really cool application. And then we've also gotten a lot of feedback from people who do work with museums, that this would be a really cool thing to do in a museum. And there's actually a gentleman who has a company, who is creating interactive tactile displays in museums. But that's a very different setting.

So in a museum, when something is there on a longer term basis, and it's just one object, and people come to it, to the object, then it doesn't have to be low fidelity, 3D printed. It could be something that you invest a lot more resources in. So you could potentially add electronics to it. So that's a slightly different problem.

But I think the general approach is definitely applicable.

So I have a question about the construction. Because these are 3D models, and they're 3D printed. For simpler models, could you have instructions for basically how they could make them out of other materials? If you don't have access to a 3D printer, but you do happen to be really good at woodworking?

It's just making sure the scaffolding is in place?

SHIRI: Yeah, that's something we want to do moving forward. Right now, that would not work, because we're using information about the model itself. So we would just need to figure out how to scan the new -- the gentleman here mentioned 3D scanning. So we would just have to figure out how to scan the object and then add some kind of marker to it, or use more sophisticated computer vision.

So it's all very possible. It's more of an engineering effort than a research effort.

That makes sense. Other questions? All right. Then I'll make my way up to the front of the room so that I can be a little more useful.

Thank you so much, Shiri. That was an awesome talk. (applause) It's really fun stuff, and I'm glad that you brought toys that we can play with. I'm excited.

Also thank you to the Internet Society of New York. Thank you, Mirabai, for captioning tonight, and thank you, Thoughtbot, for hosting us, and Adobe, and Level Access.

SHIRI: Thank you for having me.

Goodnight.