Transcript 5 | Building the Dog Internet with Ilyena Hirskyj-Douglas

Kate Armstrong

Okay, so I'll do a small introduction. So everyone is on board with our session today, and then I guess we'll give it the lecture for about 30 minutes and then we'll have an open Q and A. So hope everybody's got some burning questions ready after we have the talk today. So this talk forms a part of the Interspecies Conversations lecture series. So this lecture series is an online and regular workshop series that gives the opportunity to researchers and scientists and professors from the field of animal communication the chance to kind of present their work and then have a discussion around the work that they've been developing. So we're really interested, as Interspecies Internet to advance and accelerate this understanding of the different forms and diversities of communication between species. So if anyone does have any thoughts on great presentations which would form part of this series, you're welcome to contact us. And I hope through this way we can bring some new discoveries and new research to the forefront. So today that's what we're going to be doing. So I'm very pleased that Ilyena is here to give us a talk about building the dog Internet. So she's a lecturer at the University of Glasgow and she'll discuss her work around speculating on the dog to dog Internet and researching on building a dog to human video caller. So before we start, I'll just ask that everybody keeps your mic off during the lecture so that we can hear the full presentation clearly. And then any questions that you have, you can put them into the chat or you can also, of course, ask them at the end just by raising your hand and we'll get through those in a moderated fashion. So, yeah, I think that's where we need to start. So I pass over the mic and then I'll speak to you all after when we do our moderated discussion. Over to our presentation.

Ilyena Hirskyj-Douglas

Hi. So thank you for that lovely introduction and I'm super excited here to be today to talk about my research area and all the stuff I've been working on. So I thought I would start like a bit just by introducing myself, like Kate mentioned, you know, I'm a new lecturer at the University of Glasgow. Before this, I did my postdoc in Finland in Aalto in Helsinki. I also worked with the zoo there and did like various other things. But originally I did do my PhD and a lot of my work was focused on how do we build these sort of dog technology and so how do we develop screen devices for dogs. I really got into this sort of work a little bit by accident, I was doing my master's degree on computer science, and I got my lovely black Labrador, who you can see here, who's traveled just around with me, called Zach, And I noticed that he would be watching tv, and I was building computer products at the time. And so I thought it would be really fun to sort of combine these two things together. And this is really sort of how I really got into developing dog tech and into this sort of world of everything together. 

So I thought today's talk would be like about 25, 30 minutes long, leaving time for questions. And I've sort of structured it into like, these three parts. So just beginning, sort of what is like the animal computer and dog computer interaction fields all together, going into my work where I develop screen devices for dogs, and then at last, going into this sort of dog Internet. So the research field that I tend to build in is called Animal Computer Interaction. And what this is you can call like ACI for short. It in large, looks at how animals can use and interact with computers. And this is largely the sort of the same study that we do in humans already, where we look at how we can build interactive computer devices for them. But I would say it's actually on a larger scale. So we can look at cows, for instance, or goats, or dogs, as I look at we're going to talk about today, or monkeys. But all of this sort of involves animals and how we can design and build interactive computers for them in some form. And so this research inherently is situated really at sort of the intersection of computer science, behavioral science, design and media studies, and sort of all these several different fields. And so in many ways, it's very highly interdisciplinary. But in essence, when I talk about animal computer interaction or dog computer interaction, what I'm talking about is how we can build computers that really look at how animals interact with the world, then build computers around this sort of everyday interactions, how we can sort of iteratively design computers with animals. And in some ways, this helps us uncover really what we know about animals. And so I'm going to go over a few examples in this talk just to really highlight this out a bit more. But what I just wanted to highlight here is really that no matter what animals we sort of design and build technologies for, we use sort of very similar methods actually across this. And a lot of learning can be taken across various different animals. 

But today I'm just going to be talking a lot about dog computer interaction. And in my view, dog computer interaction has historically been mostly in research. I don't think it's such a new thing really to look at how dogs look at computers. But I think it's much more of a newer thing to look at how we can design and build interactive computers, especially for dogs. But they have been using technology for a long time. We get lots of different technology nowadays such as GPS collar trackers, we get lots of wearables and movement trackers such as the fitbark. We get these bark to speech apps and things like this, facial recognition for dogs, tracking eye movement, other emotion detectors. And then especially for the consumer market, we have a lot more of this dog technology coming in that gives them things like treats or you know, allows pet owners to remotely video or voice call into their dogs. And of course, technology is still widely used in research with dogs such as with screens. Here you have in the bottom right where they're looking at sort of facial recognition and dogs with screens. And then on the bottom left here, this one here is a really lovely piece of technology designed by Georgia Tech where they looked at search and rescue dogs and how does a dog alert when it's found someone that's looking for. So it does this through bit providing the best. And so increasingly dogs are using technologies as all these examples show, really through both in the workplace to support us, such as the search and rescue dogs use at home, such as all this pet tech. And a lot about this research purpose, such as here where they're doing facial recognition with screens to really help us learn more about them. But historically a lot of these technology used with dogs is mostly screen based interfaces. It's very easy with screens to map this one to one challenge. So where a dog touch on a screen is what it means. And these have a very long, long history in dog cognitive studies. But there's also a lot of new technologies that are being developed for dogs that really change, I would argue the way that dogs use computers. And because traditionally to use screens, dogs have to be trained to touch screens or use screens. And I think it's like a very humanistic way. 

However, dogs, the way they behave and the way that they interact ordinarily in the world is much different, right? It's much more tactile. They exhibit all these wonderful behaviors. They dig, they chew, they sniff. Here they're playing with a balloon. They have all this wonderful foray of behaviors. And this is really where I see sort of the future of dog computer going really with these technologies that fit into their behavior, that sort of fit into the way that they already interact with the world. And so what I try and do with my work. And where I really see the future of dog tech going is really if we just sort of threw out sort of what we see as computers for humans to look like and really started again from the dog's perspective. And it becomes this large question then of really what do computers look like that would really react to how they behave in the world already? You know, what would they do with these computers? And how can a dog really control a computer if you build it around the everyday behaviors? This is the larger questions that I think animal computing widely tackles that I try and tackle also through all this novel technology that I looked at. 

Why do I think this is important really for the future of dog tech? I guess this is the key takeaways for this presentation. My research and other research in the animal computer interaction field has really shown that making these tailor made computers, computers for animals so bespokely for their needs and what they require, really offer things to animals which they don't normally have, such as autonomy over different aspects of their lives. They can control different things such as the light, the heat, the visual displays, or various different aspects of their environment. And this choice and control in animal welfare has been massively documented to improve an animal's lives and overall welfare indices. And so I really think if these computers are going to be successful and really improve a dog's life, they also need to be safe, they need to be functional, they need to have some sort of meaning to the dog. And so I really think that by looking at how a dog can use computers and making these more functional and more fit into how they behave with the world, I think we can make computers that have better this user experience are more enjoyable for dogs over long term and that can really have a benefit towards their life. And this is really sort of my overall goal and where I see the future of dog computing going for now. And so to help make this more concrete, what I thought I would do is sort of go over like two key core aspects of my work that I mentioned in the beginning. So the first one is to look at sort of dogs and screens. So as I mentioned, dogs have been using screens for a long time in cognitive science and animal behavior research. And actually this is how I also got into making dog tech was I noticed that my dog was watching TV with me in the morning when we were drinking coffee or just relaxing. And I started to think, wouldn't it be really cool if he could control the screen himself? And so I started to do some research into this and I found a lot of the ways that we track dogs with screens is we traditionally train them to rest their heads. You see here these sort of images or we, you know,  you need them in a very still position to do this very finite tracking. And I thought that a dog doesn't really have an option here of looking away. And I don't really want to train my dog to use this technology. I very strongly believe the technology shouldn't instead already react to his normal behaviors. And so I started to think that then what would screen look like for a dog if we're going to change this technology over? And so one way this could have been done is maybe through a TD remote, right? So we just. Here's one that was developed by WAG, a dog food company brand. So it might have grips for dog paws. And here you see it's like this yellow and blue designed to fit in with the dog's vision. And in many ways you could argue that this is made for a dog. But I don't think this is really, you still have to train a dog to press these buttons. A dog doesn't really go around it's life pressing buttons as it does in this. And so I think this is really just computer for a human sort of sized up for a dog. And this is still very much a human idea of what a TV remote would look like. And so I thought, okay, how do I go back to recognizing his normal behaviors and how my dog would watch tv, you know, is just through his gaze. But I didn't want to train him. So I thought maybe I start with the computer side of things. And I look at how can we build an image recognition algorithm that would recognize my dog watching the screen automatically. And so I tried some human ones you see here. Usually these boxes are around things like your eyes and your mouth and wasn't able to recognize him at all. And so then what I did was I used MATLAB's box of features, bag of features even, to build like my own one here. And this was this very gross approximation whether he was looking left, right or straight. So this is my dog I did it on. And what I found was we can get these sort of high numbers, so we can, we can get fairly accurate recognition, but it takes a long time. And if you think about the way that a dog's faces are, they're all such various different shapes, various different colors. And so it's extremely hard actually to track back in Freeform. And this was Also back in 2014, as you see my paper on the bottom there, where there wasn't that much actually out around tracking dogs as much as there is now. And so then what I did was I thought, okay, so we can track him looking like center right or left. And so I thought, okay, so what can we use this for? Because we can't get this finite detail. And then I thought, okay, maybe instead of tracking, we can just recognize when a dog's looking at the screen instead. And so what I did was I built this system that was… We built it so it was able to recognize sort of the dog looking between these three different screens. And then we change the content around the different screens to see if the dogs would follow the content on the screens or whether they would just look at the same screen. So, in essence, whether this would be something worth doing, you know, if we could, dogs could have a prolonged gaze at the screen. And the end goal here was really just to see this, whether they could use this method to select from screens. I thought this would allow them more of this freedom and control. And what I did was I cycled, you see here, all these different videos and sort of 20 second iterations between these different cycles and to see if they had a favorite screen.

And because I wanted to do this in a sort of a friendly way to animals, what I actually. Well, to dogs, what I actually did was I just, just put the dog in the study room with me and, or with their, you know, their owner. And I allowed the dogs to just behave however they wanted to behave. What the main outcome was with this. My method that I developed was successful at gathering data. We found that dogs tended to not have preference between the screens, but they had these very low attention times. Here you can see it's typically under one second. This was only done on a few dogs. This was more of a method study. So I can't apply this across to all dogs. What we also found was that they tend to actually spend most of the time watching nothing, not watching any of the screens. Indeed.

And so this was kind of a nice contribution because if you had trained a dog to rest its head, you wouldn't necessarily have had this sort of outcome. And so it was really evidence to me that if we start looking at methods of dogs to use computers differently, then this can be effective and we can get very different results. And this also started to highlight that dogs don't typically look at screens for long periods of time. And so we started to think about, maybe I need to think about this a bit differently. And how does a dog, you know, activate a screen. And so from this I thought the next stage would be so this, the, the content was cycling automatically itself. And so I thought the next stage, because they didn't tend to have a preference being able to follow content between screens, I thought this is not so good for dogs then to use this as a method. And so what I did after this was I thought, how can a dog control the screen themselves? So back to this original idea of it. And what I did was I built this system for dogs called Doggy Vision. And what it does is it recognizes when a dog is stood in front of the screen and then it will automatically play randomized. I did it just on randomized videos. This device was a little bit harder than what I necessarily thought to build. We have things like dogs do crazy zoomy runs, they wag their tail in front of all the senses. And so there was lots of these small sort of problems that we face when I was trying to build all this dog tech. And the system would, in essence, it would film whenever a dog was interacting with it and save this online and to, to like sort of spreadsheet styles. So then I could see exactly what the dog was doing. And what I did was I did this with two dogs. In the end, I ended up putting this in two dogs homes for two weeks. So I left the screen and they also had like sound as well as the videos off for two weeks. And then I turned it on for two weeks. So I sort of had this baseline of how their dogs typically use the space and then how it was affected by this sort of screen device. And what I ended up finding was this system worked and dogs, both of the dogs involved in the study did use the screen. But it didn't really seem that the dogs learned how to interact with this screen. So just by stepping close towards it was like, I think about like 120cm in front of the screen, this boundary box, because their behavior didn't really change much between the screen being on and the screen being off. Again, like my three screen study, I found that dogs typically use a system in these very short bursts. So typically they were under 3 seconds all of their interactions. So once again, this really tells us and highlights that how dogs interact with screens is very different. But it's still sort of a larger question of what does the future of dog screens look like. And so I started then to also look at how often dogs are paying attention to the screen with this data. So I found that you see here like above 50%, it wasn't getting really above this on the best of days. And so then there's no real trend towards their looking behaviors. It sort of happened a bit randomly. And it isn't really clear why dogs increased or decreased their looking behavior. It wasn't towards any video, it wasn't towards any sort of day. And so it's still sort of open question really, of why do. Why do dogs look at screens? What attracts the dogs towards the screen and sort of the future of screens in this. But overall, what we did learn from this study was we can build technology so a dog is able to activate the screen itself. And we. I learned a lot, really, as a developer of computers for dogs here, such things. You know, I mentioned the wagging tail, the sensors at the body height, for instance. I made the technology so they wouldn't trigger when they slept in the space. There was lots of different sort of optimizing things that was learned from this study. But I did find that this idea of proximity did work well. But there needs to be more structured learning to how a dog can figure this one out. And then when I started to think about this further, and this is sort of like the closing bit on screenshots, what I started to think about was maybe it's because I'm editing how the dog interacts with the screen. I'm editing singular or multiple screens, but maybe it's the screen itself. And this is really, I think, where the future of my work and screens will be going. Because a dog's vision is typically much wider. They have this, like, smaller focus area. And so then I start to think that maybe one screen isn't enough. And I'm still thinking in this humanized way, this where we just have a singular screen with humans, we don't tend to. And we have like more now AR and VR immersive events. But I don't really want to strap anything onto my dog. And so I think the future of screens is really something that looks very different from the screens that we're seeing at the moment. And so then I'll move on to, like, the second sort of topic that I'm interested in as well as this sort of screen technology for dogs is really sort of how can dogs use the Internet? And I got really interested into this technology because I was building all these screen technologies for my dog. And I was, you know, doing a lot of these sort of novel programs that I never published. And I would just write about on my blog, all of this like, dog tag and where I can do these different things. And I then got a job in Helsinki. So I was in England and I sort of moved across to Finland and I thought, oh, you know, whenever I was ringing my dad, my dog was going a little bit crazy. He purrs a little bit like a cat. And I started to think, wouldn't it be wonderful if the same way that I miss my family and I was ringing my family, if somehow the dogs could connect over the Internet. And so this is also very relevant. It became more relevant. I moved into Helsinki in 2017/18, as we had, we've had during COVID lockdowns, many dogs now, you know, are left alone for long periods and not typically being alone for long periods. And so I think really here the Internet has a lot that it could offer dogs. And in the same way that it's been sort of scaled up, it started out in very specialist for humans and then we've increasingly got it into. It's now ubiquitous, you know, inside everything we do. I think the same thing that could be happened for dogs. And so no one had really looked at sort of at this point what two animals, what two dogs interacting over the Internet would look like. So I started to write papers about this and sort of speculate on the future of this dog Internet. So we used sort of very much drawing and very much like blue sky thinking. And so we had things like this one over here is where a dog is running on a treadmill with its friend. Because dogs use so much of these referential gazes when running. So they could run inside the home and still run with a friend. And we had one like here where one dog, they had two balls and they were connected inside the home. And so when one dog threw a ball, it would throw the ball in the other home. So it allowed them to sort of play remote fetch almost with each other. We had another one here that was a bit more light hearted where two dogs are sat at home and they both decide they want to meet up. So a laser comes on their collar and then it points them both towards a local dog bar which has like a gate inside that only lets friendly dogs or dogs that get on on the inside. So it allows dogs to remotely meet up. And then we have another one here where they had these sort of tug of war rope style toys and they would recognize how each dog played with his tug of war and sort of mimic the style back. So you could have this sort of big huge virtual tug of war between multiple dogs. And so you can see we sort of started imagining the Internet and very much already how a dog behaves with the world and how this could really fit in with them. What I mentioned at the beginning, and we had these sort of like lovely discussions around it, and what we sort of found was sort of the key finding here was we were often repurposing current dog. So whether it was we had that balls or we had the rope interface. And this was because a lot of dogs already know how to use these interfaces and they already know what they're for. So in terms of their affordance and how a dog could use them, this would be something very easy to do. We often spoke about a lot of recognition behavior. So using machine learning to automatically recognize a dog was doing something such as some people recognize they wanted tail wags for liking things, or, you know, drop a toy inside a basket, start the interaction, and then recognize that the toy was there. And this was often spoke about a lot as well in this idea of choice. So that a dog has really a choice to take part in the. The technology interaction or not. So it's just not automatically sort of turned on. And then we also were discussing these screens, I mean, discussing these designs. We also spoke a long time talking about the fact that technology here is really a mediator between two dogs. And so there is really a need for both dogs in the interaction to somehow understand that they're both interacting with technology in some form. And we speak about this a lot in like, human computer interaction, as in the gulfs of interaction. And, you know, and what does it mean to understand technology? And you can raise this sort of as a philosophy question, I guess. And, you know, how does a dog know a dog is playing with another dog? You know, they have this asynchronous interactions and how can we get a dog's intention of playing towards another dog? And I think Steve North summed this up very nice when he said all animals are on this sort of continuum of awareness. And, you know, I believe the dog's also possibly on this. But hand, do we know really what they understand when using an Internet is such a large question here. And so I believe we can capture it in some sense, as in, we can see what they're doing with Internet devices. But in this case, what a dog would expect to get from another dog in a computer system and its intentions of using another system is still very much unknown. And how we would capture this, especially when we think about it, is the meaning and experience of what a dog. Meaning and experience of what a dog has when using technology is very much through the human. So either through the owner or dog behavior as acknowledgement. So we have to recognize that only behaviors that we see and that we can recognize as behaviors, the ones that we are really noting down. And so I think the dog Internet is still like a little bit of a way off. But this doesn't mean that I don't think it's totally not possible. And so what I did was I took a little bit of a step back in this because I had such these huge questions of how do we know a dog's intentions and how do we build something even that is for dogs performances. And so what I started to do was I started to build this device for my dog. So Zach, that allowed him to video call me. So the very basics, I guess, of what we use the Internet for, this video call, it would have been lovely to do like a smell device. But this technology just doesn't exist. Or something more towards how they interact with the world. And what I did to build this was I originally started out with looking at how he uses the world. So we gave him like new toys. So this is the ball toy here. He's a Labrador, so he carries around toys and fetches toys a lot. This is him playing with a Winnie the Pooh. He loves sticks and he chews sticks. The problem is sticks sort of wear out over time. So they're not so good to put technology in or augment in this way. Or maybe this is kind of the point of it. He tends to lick and groom all these toys that he has, such as this cuddly soft toys. So these aren't necessarily good because these have different behaviors for him. But these ball toys, he doesn't chew balls. He actually just ends up carrying them around his house. Well, our house. And so I decided to look at then different sorts of balls. So I gave him like a tennis ball. He associated this way too much with fetch, the sort of softer ball. But he ended up mostly sort of using him. A sort of favorite toy was this softer ball. And so what I did was I started to build technology that would fit inside of this ball. And I put it inside a tennis ball so it was chew proof, even though he wasn't really known for chewing. But it enabled the ball to maintain its sort of squishy factors which and sort of shape, which is, you know, part of the reason he originally chose the ball. And what this technology did was anytime movement was detected and we did a lot of testing and trials, so it was a significant enough movement, it would video call me. So he had another computer set up in my house. So it Would video call me, and then when I video called him, he could also answer the video.

Vint Cerf

Looks like we need the video call ball to get Ilyena back.

Kate Armstrong

Yeah, let's just. Let's see Ilyena.

Kate Armstrong

Okay. Let's see if we can. Oh, we're right on a. Right on a good moment.

Ilyena Hirskyj-Douglas

Can you hear me? I don't know if my Internet dropped.

Kate Armstrong

Oh, there we go.

Ilyena Hirskyj-Douglas

Yeah, I just heard, like, someone say video call, and I tried to answer, and it, like, cut my Internet. I'll just go back to sharing my screen.

Kate Armstrong

Awesome. Thank you. Yeah, we were right on a really good moment. We're all wondering what's gonna happen.

Ilyena Hirskyj-Douglas

Yeah, so I use the system sort of with my dog over many months, and it was a video call, so it's would call from the laptop, but the ball was the thing that would trigger the call. And what I sort of found was he would ring me daily, sort of often at the same sort of time. I don't really know why this is. You know, he could have just liked the ball, but equally, if he didn't like the videos, he could have just abandoned the ball altogether. Occasionally he did approach the screen. He did also bring me things every now and then, such as he had this, like, pink. Pink puzzle feeder that you put, like, treats inside that he would often bring me. And what happened actually, was he started to ring me a lot near the end of the study. And so often I was in sort of the underground. I was busy somewhere. And so I would just always try and answer the calls and tell him that I was doing other things. And so at some point, because I thought, oh, he's, you know, we're just having the same conversations, I started to flip the camera around. I started to show him buskers on the street. I showed him my office. I showed him, you know, all around, sort of my friends, where I was, what I was eating. And I noticed, in my opinion, that he had much more interest in these sorts of calls. And so I thought, this is, like, really some. Maybe I've been a bit, like, centered that he just wants to call me. And maybe this might be more about expanding his world out there. And so why this work was really a case study in itself, as in, you know, I built this really for my dog to be able to use. These findings really have usage across different areas. And I think what we ended up finding was that dogs do use technology differently from day to day in this. And again, we don't really know what a dog's intention was around this. We demoed the technology to him initially. So I had someone in my house demo the ball and saw that it rang me a couple of times. But I think this wasn't enough really to establish a connection that this ball was necessarily interactive. And I think this is possibly one of the hardest steps, I would say, in making interfaces for animals is really in finding something that they understand how to use, but then not making it so so much like their current environment that it's a differentation across. And so it's this sort of battle and balancing, I guess. And I think also that sometimes I would call these interactions he had with me accidental because he would like nudge it with his butt or he would lie on top of it, or he would just be running around the house with the ball. What I started to realize after the study was maybe this is just how he wishes to use video call interface. Right? And we really don't know how animals intend to use computers. So we have to really remain open minded here as to what we define as intention with animals, as what we define as interaction looks like. And really how animals are going to behave with Internet devices, because he arguably could have totally avoided this device altogether. And so I think that perhaps, maybe new terminology is needed when we talk about the animal Internet, really beyond what we see as this humanistic world and humanistic expectations. And also part of this process was really exploring sort of how can, you know, these Internet and Internet of things devices really support dogs and human relationships. So at the start, when I introduced this technology, in the first call, he did do like a whining noise and I quickly shut off the technology. But then every call afterwards, he never did this in there. But actually towards the end of the technology usage and why the study actually stopped was he didn't ring me as he typically would, and I was like, oh, where is my dog? Is he okay? You know, this is such a typical ordinary interaction. And so this caused me some anxiety towards the end. And so there's a real need here that when we develop, you know, Internet connected devices that we need to look at, how do we balance this and how do we make sure that both of these users are okay? And so because I looked at a lot of designing technologies for dogs, I think one of the key challenges here is that we don't understand so much about how dogs use computers. And part of this process is really sort of this slow unraveling of all the requirements of what a dog needs. And part of this is also acknowledging that we are essentially humans building computers for other species. And so we might not know what they need. And while a lot of this technology is aimed to empower and give animals choices, this empowerment and choice implies in some way that an animal understands what they're doing or understands what's going on in the technology. And it could be argued here that if they don't understand what's going on, then it's not really so empowering for them. And so I think we still need to heavily guard and sort of make space for animals being okay with the technology and sort of here I did it. So I was responsible for hanging up the video call, for instance, in case, as I mentioned, when he whined. But importantly, what I really wanted to highlight from this study is really that dogs can use computers differently and they can use the Internet and they can control video systems, they can control these screen systems. And that we really on the precipice of looking at how can we design like computer systems for dogs who have these very different cognitive and biological needs. And how do we rethink sort of the Internet devices for animals? And I think here we're really forced to see, you know, dogs do use computer systems so differently and we should really start opening up into these new possibilities of really what the dog Internet could look like when using Internet systems. And so in summary, I sort of given it that I really my work and what I'm really interested in and where I see the future of all this going is really these bespoke systems built for dogs that give them these different ideas of choice, explore the different ideas of agency. A lot of new methods will be needed, sort of how do we measure animals interactions with computers and how do we do this in a friendly and compassionate way. And then I also am a strong believer that I think like the Internet is coming for all these animal devices. And I hope that this future technologies can really empower animals and that the Internet in many ways can give us the same benefit to animals. So you can find out more about my work on sort of my website. So ilyena.com I also do a lot of tweeting or you're more than welcome to email me. I also build systems for monkeys, so please don't get confused if you look at me up and find a lot of cute photos of sort of small monkeys. But thank you so much for listening today. You know, it was a real pleasure to talk here and I'm thankful for being invited and I'd love to answer sort of any questions you may have.

[Q&A]

Kate Armstrong

Awesome, thank you so much, Ilyena. I think we already have some questions in the queue, which is good because I'm not surprised because it was such a. Interesting and I think really. Yeah. Thought provoking talk, especially because we're here to talk about the Interspecies Internet. And so I think a lot of questions and thoughts come up. So we already have. Steve has a question. So if you want to go ahead, Steve, let's start.

Steve Crocker

Thank you. You've made my day. This is delightful work and I have to just applaud the immense amount of energy that you put into it. Lots and lots of questions come to mind. But the one that comes to top of mind is one of the most expressive parts of a dog is the tail. And you only mention the tail in passing. Has there been any work, have you or others done any work on trying to instrument the tail or give the dog the ability to express itself using tail functions? And I have no idea what the history is. I'm sure there's a lot of history on how precise those actions are, whether they're measurable, whether they're repeatable, what the range of motion is and so forth. Thank you.

Ilyena Hirskyj-Douglas

Yeah, so I haven't looked at tail itself. I know people that have looked at it as a form of sort of liking. So whether it likes this. But I wouldn't know the work itself. Yoram, I don't know if you know this work as well because you're sort of in this sort of closest research field. But I've seen it used that as in terms of user experience to measure the tail. But I haven't looked at it myself. It's always very hard. Right. Because I'm not sure if there's a unified agreement on. Is all tail wagging the same across species, Is all tail wagging the same across different spaces or different contexts or different cultures? Right. And so you open up this huge like I feel like always Pandora box when you open look into all of this. But no, I haven't looked at it myself. Thank you for your question.

Kate Armstrong

Yoram, do you have any thoughts on it? I mean.

Yoram

Because I've been tangentially involved in the field but I haven't seen anything on sort of tail interactions. I think it's actually quite interesting because it's obviously a way of. Of communication but most of the, most of the stuff that I've seen has either been screen based because it's a way to interact or various forms of tactility. So trying to get whatever animal it is touch and do something and using sort of facial recognition or movement recognition to do So I haven't seen anybody do anything with sort of tail wagging or trying to interpret that. But that's, that's an interesting, an interesting, interesting point.

Ilyena Hirskyj-Douglas

I think it was Clara Mancini’s PhD student who looked at tail wagging as a form of liking. So I thought someone said tail wagging is pretty common across species. From my like small knowledge about this and feel free to correct me if you're wrong. If I'm wrong, it isn't always unified the same wag. Right. They have like different sorts of things within that.

Kate Armstrong

Interesting. If anybody has any, anything on that, we can put it in the slack because I think that would be a very interesting, interesting thread to follow up. Con, did you want to ask a question?

Con Slobodchikoff

Yes, I'm really interested. Does anybody have any information on the resolution ability of dog eyes with respect to screens? In other words, can they see fine SC differences in the screen or are they more attracted to motion? And how does that relate to the flicker resolution of the screen?

Ilyena Hirskyj-Douglas

So no one's looked at it in finite scale. Right. So it's in like how small? I've not seen any papers look at how small. I know I've seen papers that look at. A lot of dogs are able to recognize things on screens that have a high enough flicker rate. Right. So these can't be the older sorts of screens. People have often contact me and ask what do dogs like to watch? And I'm like oh, you know, this is like such a side part of my work. It's more about the technology side of things. So dogs. I know there's research on dogs able to recognize other dogs faces, dog ables recognize human faces upside down and dogs able to recognize like if they're shown an image of the toy to go get a toy. But I don't know to what granuality they're able to recognize. And a lot of this is just taken from the biology of the eye, what we think dogs can sort of recognize on screens beyond them being able to recognize these things.

Kate Armstrong

Vint, did you want to.

Vint Cerf

Sorry, I wasn't sure in which order this was coming. So this is fascinating. Of course I'm always interested in anything that makes the Internet more useful for any, anyone including dogs. One I put in the chat a question about Fluent Pet, which is a dog signaling system that we've discussed in earlier meetings. Dogs get to select buttons to push and the owners can record whatever sound that button produces. So that's how dogs can express themselves. And apparently they learn which buttons you know or I want to play, I want a food, I want to go out button could obviously activate a screen and make a call.

Ilyena Hirskyj-Douglas

Yeah, I could, but I don't see a dog's pressing a button ordinarily in their sort of everyday life. And so I'm really interested in moving beyond how we. We press buttons, right? We're trained to press buttons. We learn to press buttons and sort of new era of technology for humans are very much, you know, going off on normal, everyday interactions. They root screens recognize, for instance, my iPhone when I look at it, to unlock. We have all this lovely VR/AR technology that recognizes our normal everyday interactions. And so when I think about building technology for dogs, I think, okay, how can I build technology that fits in with their normal and everyday interactions and behaviors that really fit with them? And what I don't want to do is I also worry a lot about this training bias. Right? So if we train an animal to do something, how do we ever really know it's doing it for other reasons? We get this clever hands approach. We enter into so many different forays here. That how I've always wanted to build technology has always been from a dog's perspective as much as possible that I could get into.

Vint Cerf

Well, I'm going to make the argument that dogs are intelligent enough to learn how to do things and they do this on their own in many cases. So I'm not sure that I reject the idea of the button pushing as somehow weird. Dogs can learn to do things and they are using that to express agency, so I wouldn't reject that. And the fact that they could use this in order to make the calls that you've been recommending or trying out seems like it's a good direction to test. So anyway, we'll see what happens. It looks like in the chat there are several paths going in that direction.

Ilyena Hirskyj-Douglas

Yeah, I mean, if it works for some dogs and it brings us something in their life, then why not? But I just think the future of technology really needs to look very different.

Kate Armstrong

Mark.

Mark Graham

Great, thank you. This is fascinating and I see all the conversation here about fluent pet and Vint comments there taken to heart, I'm curious and I apologize I've not been following your research or maybe you've covered this already, but I'm just wondering if you've shared any techniques or tools that would encourage other people to do the kind of work that you're doing with animals and maybe do it in a, I don't know, kind of an open, sharing, collaborative kind of fashion. So that kind of, you know, Crowdsource this effort a little bit. And I'm especially intrigued by the notion of, you know, the multiple animals coming together and I know, choosing to do something together, which, you know, I don't know this. You can be doing something together but not know you're doing something together. So I'm not really quite sure how to tease that out, but maybe through more experimentation with a lot more data collection, once again, that could be open and shared and accessible might be a good route to pursue.

Ilyena Hirskyj-Douglas

Yeah, yeah, I agree. I think this would be super nice in the future. I mean, I try and do all my software open access, I write blogs about how to use all this software, how to use my tech, you know, why I made these choices and code on why things that don't work. But I guess this is such like a new thing in, in my mind to sort of rethrow out everything and just start to build it again that I really hope that this sort of just grows and that more people look at this dog tech in a different way. But it would be wonderful to have this sort of collective around it. I totally agree.

Kate Armstrong

Yeah. Maybe one day it could be merged with some open hardware movements. So we can also have this development of different kinds of tech for different dogs as well. Because I think one of this really great point that you made is that different dogs use technology differently. So this is also interesting to think about dogs as individuals as well.

Ilyena Hirskyj-Douglas

Yeah. There's also like a big argument that personality recently has affected it. So people argue that, you know, how dogs use technology is also to do with the personality as well. So be interesting to see that.

Vint Cerf

So it's Vint, again, just. Have you got to the point where there's been another dog on the other side as opposed to a human?

Ilyena Hirskyj-Douglas

Not yet. No, I've not yet built that because I was just. There's so many unknowns already that I didn't want to then decide jump into something where I couldn't really have much takeaways from it. Right. I don't really like to use technology with dogs if there's no real meaning behind it or I can't see a very clear outcome behind this because we inevitably always have this risk, right. That an animal won't like technology or it will, you know, cause some sort of anxiety or other issues. And so I started really with, and I only published like back in November, this paper about sort of me and the dog video calling and are now starting into the sort of research into sort of what do two animals look like on the Internet. And how can we facilitate this? Maybe it starts with dogs. I do lots of work with monkeys. Maybe it starts with monkeys. Like, who knows really where this has gone?

Vint

So the chat is moving quickly towards control, using the Fluent Pet as a control device. It seems to me that dogs are really interesting because I don't know how they recognize each other. The breeds look so different, and yet when you take two dogs on a walk, you know, and they encounter each other, they're barking like crazy. Sometimes when it's meant, you know, yards, hundreds of. Not maybe not hundreds, but, you know, 100 yards away or 100ft away. I don't understand how dogs recognize another dog. And I do wonder whether their reactions differ depending on whether it's a dog in a video versus a dog in real life. Because real life case, there's potential for olfactory recognition as opposed to visual recognition. Although I'm really surprised that if a dog is 100ft away and they're both barking at each other, it's not clear whether they're barking because of an olfactory discovery or because of visual. And I don't know, you know, but it could also be something like a dog.

Ilyena Hirskyj-Douglas

It could also be something beyond our recognition. Right. I tried to touch upon this a little bit in the talk that we can only really recognize the way a dog is using the world in ways that we're able to recognize it. Right. Dog could have a totally different sense than what we know, such as, you know, with birds and infrared and all this was only recently discovered. So I kind of try and keep, like, sorry, I live, like, right by a business street. I try and keep like, sort of a very open mind into how dogs can experience the world and really sort of start on this ground up of looking at how they can build technology from there. I mean, it's a really interesting. We get into like the villa philosopher question, right, of when they said, like, you know, I think it was Ludwig, right, who said, like, if a lion could speak, we wouldn't understand what he was saying. Right. So if a dog could recognize another dog, would we understand what they would necessarily say to each other or what they would do to each other? So it becomes like a bit of a philosophy foray. I think at times.

Kate Armstrong

In the chat we have a question or also a statement from Sasha. I don't know if, Sasha, you want to ask this question, because I think it's really interesting, kind of along the lines of what we were just discussing, discussing between the natural and unnatural dog behaviours. What is learned behaviour. And what is a behavior that is naturally occurring in dogs? Did you want to talk about this?

Sasha

Sure. My robot vacuum was. I've shut it up, though. So I actually have two questions since I have you. One of them was this idea you were speaking about the button pushing as being kind of an unnatural behavior.

Ilyena Hirskyj-Douglas

And I'm thinking maybe we don't know what's natural. Right. For dogs. Dogs is sort of like a subspecies. I meant ordinary behavior. We don't see dogs ordinarily doing this. Yeah.

Sasha

So I guess that was kind of the basis of my question of sort of like, well, how do we know? I mean, dogs constantly develop because they're used to living with us for so, you know, they've been living with us for so, so long. I guess that was kind of my question of like, well, how do we know what's ordinary and what's not ordinary? As one of the things that seems to me, to my mind, ordinary, if you will, about dogs and lots of animals is their ability to adapt and be creative, I guess, about ways to communicate when they, you know. So like a dog that will. I had, I had a dog who would always sit across the coffee table between me and the TV when he wanted to go out. Why? Because I'm watching the TV and he knows he'll get my attention. Now, is that a natural? But, you know, there's an ordinary behavior. Not all dogs are something that he's learned from the situation. And so to my mind, the buttons. And I'll confess, my dog has buttons. So maybe I'm biased, but, but. And my dog picked up on it just watching me press it once. I mean, she pressed one within 30 seconds. She's abnormal. But, but, you know, so, so to my mind, sort of like just the, you know, if the ordinary behaviors, their ability to creatively learn, I think it, I guess to my mind it's sort of hard to tell what, you know, what is ordinary and what is it. So I guess I wonder your thoughts on that. But then my other question is about this idea of screen time. Like, screen time is something that we're all so concerned, you know, people are concerned about screen time with children, we're concerned about our own screen time. We have these apps to prevent us from spending too much time on the screens. So I wonder if that's ever something that you've thought about in using screens as the basis. Is this something that maybe is a concern with screen time with other animals as well?

Ilyena Hirskyj-Douglas

So with what's ordinary, I agree. You get into like a huge, huge down rabbit hole of like what's ordinary when we don't have a species that is like exit you in situ. Right. Well, maybe you have wild dogs as well. Right. It gets into totally crazy rabbit hole. And so when I, when I use the term ordinary here, I'm meaning something that's like untrained, something that I see them doing irregular behavior of this. And I guess because my whole background is in creating new technology, I'm super interested in how can new technology support dogs from this. My dog doesn't use buttons. I think it's typically dogs, when they see a button, wouldn't necessarily know how to use a button. And so I just wanted to build something really that fitted in with an everyday, normal, expressive behavior of things. And I understand that this is very flexible. Right. And maybe if every dog starts using buttons and this will become part of their ordinary behavior. Right. And so it is this flexibility, I think in there a little bit, but it's just not what I do. And the screen time question, yeah, this was something that concerned me a little bit actually more it came up more in the video calls side of things because he called me a lot, my dog, near the end of it. And when I made the screen device for him, it didn't have any time limits on it actually. So he could have watched it for a long time if he wanted to, but he typically didn't and other dogs haven't done before. But if I was going to develop something and all my technology reports, basically whenever it has an interaction, it automatically reports this to me. So I can see if there's a huge amount of screen time happening. And if this happens, then I think I would start putting limits on. I'm always pretty shocked that a lot of technology that we have for dogs in our home has no limits on. So you can effectively give it a treat feeder. And it can get like, you get anxiety behaviors, you can get frustration behaviors with all these treat givers. And I've talked about this for a long time about, you know, owners just giving dogs this technology, walking away, leaving it, and we don't know really what's happening. And it can cause a lot of negative behaviors. And so all my technology is tested with all the animals before I leave it for these long periods, make sure that I don't have any of this negative reaction. And then I look at all the files online and check all the data as it goes along. Because then if I think if I did see this huge amount of interaction, I would then start having to question you know, how much is good for a dog? Are we seeing any negative, anxious behaviors around this? And it's the same with the work I do with monkeys. This is exactly the same. We just monitor the technology. We ask the keepers. So what? You know, whoever looks after the animal is the animal. Okay, here. And this is why when I did the three train study, the animals carer was also in the room with them, so they could actively monitor and make sure the animal's welfare was safe. Because above everything, this is primarily what I'm concerned with.

Sasha

That's. That's really interesting just to. I don't want to take up too much airspace, but I'm just thinking, you know, so. Yeah, of course, you know, the pressing of a button, not something you would typically see. But I wonder how much we're parsing that out from, say, like, pawing at someone or something like that. You know, like, how much is this just a behavior that's just being moved to something or digging at the ground or. Anyway.

Ilyena Hirskyj-Douglas

Oh, it'd be lovely to have technology that recognized, like, when a dog dug a hole and then did something for it or something like this, you know, this would be excellent. I thought about this once before with balls, because balls like to dig so much. Yeah, this would be really nice. Very cool.

Sasha

Thank you so much for your answers.

Ilyena Hirskyj-Douglas

Thank you.

Kate Armstrong

Con, did you have the question? I think you're still muted.

Ilyena Hirskyj-Douglas

This happens to me a lot.

Con Slobodchikoff

Sorry about that. I found it fascinating that your dog like to look at your office and surroundings and various other things. I wondered if you could include in your experimental design something like a surround sound vision screen, which would give a large, broad range of visual stimuli to the dog, including you. Then if your dog lost interest in you, that the dog might perhaps focus on something else going on and then come back to you and so on. You know, we've been having this discussion of smell versus vision, and from what I know of people who have looked at dog vision, dog vision isn't really that great compared to smell, which is extraordinary. And so maybe if you give the dog some interesting visual stimuli, that the dog might spend more time focused on what you're doing and listening to your voice at the same time. Just a thought.

Ilyena Hirskyj-Douglas

Yeah. Yeah. I mean, this would be super interesting. Right? We start thinking about, like, does a disembodied voice. What role does this play in here? Or, you know, the future. I really think that this dog phone, really, what it showed me was that maybe my dog would like a video caller into sort of a more dog park, into nature, into the city. And where would he like to video call? What would he like to see? Which elements of me or not of me or maybe his friend or other friend do we include into this? It's a super interesting sort of study. Thank you so much for your comment.

Kate Armstrong

There's a question in the chat. In the interest of time, I think we should maybe, you know, wrap it up in the last few questions. But the chat is really taking off lots of different discussions. So I encourage all of you to head to the slack and continue the discussion because I think there we can really also share some insights. But there's an interesting question we have. Jiabao, I don't know if you're still with us and you want to ask the question yourself or I'm happy to pose it,

I don't know. I can't see you in the chat so I might ask it for you, which is do you have any techniques and tips on user research in inverted commas with animals beyond observation?

Ilyena Hirskyj-Douglas

Oh yeah. So I think observation is very valuable, but I think it should also be backed up with quantitative data such as timings of when they use things, how long they use things for, quantity of their behaviors. I always really advocate for people to get like experts involved to really sort of help give more perspective on the behavior as well. So not just like the, you know, the animals carer, whoever they are, can give an perspective. A behaviorist can give a perspective. We can measure their interactions. We can form this against how long they use something for, how frequently as well they use something for. These are also two different things. So often I use a sort of mixture of all of these methods together to really try and get a rounded viewpoint.

Kate Armstrong

Awesome. That's great. So we have that also in the recording which we will put onto the YouTube and we also upload it to Internet Archive. If anybody's interested in checking those, we include the link on our website. So if you're interested to look again at the session or anything or any of the other previous sessions that we've had, you're welcome to. 

In the interest of time, I'm going to kind of wrap it up now and invite you to join our next lecture, which we're very happy that Jennifer Mather, who is also here with us today, will be sharing her insights around some of her octopus research. So it's going to be very exciting. We will meet on the 30th of July, so approximately one month from now at the same time. So we will send out a registration quite soon so everybody can jump on and register. And of course that also goes through our newsletter. So if you don't receive the newsletter please get in touch and we will also give you the sign up for that and of course see you on the Slack. 

Thank you Mark for sharing that on the chat. That's where the conversation kind of continues. So we've had some really great conversation today and also we'll see if we can get the chat out to you all as well. So yeah, I think this has been a fantastic session and thank you Ilyena. I think this has been amazing way to, you know, really get into the depths of what's happening in terms of how we engage dogs through their own sensory behaviors. So thank you very much and of course get in touch with us or also join our next session. So thanks for joining us and enjoying.

Ilyena Hirskyj-Douglas

Thanks for having me for the lovely questions everyone. It's been really wonderful talking about something that I had just super loved.

Kate Armstrong

Awesome. Thanks so much everyone. Bye for now.