Human-Robot Interactions and Social Robotics with Professor Marynel Vasquez

Who speaks in this episode?

• Marinel Vasquez (Guest)

What is discussed, and when

1. 0:00 – 5:00: Marinel Vasquez, HRI, Robotics Institute
2. 15:00 – 20:00: Roomba
3. 20:00 – 25:00: Shuddter

Notable quotes from "Human-Robot Interactions and Social Robotics with Professor Marynel Vasquez"

• Interactive AI with research spanning fields like computer vision, machine learning and robotics. Before Yel, she was a post-doal researcher at CMU's robotics institute and she…

  about Marinel Vasquez · at 0:58

• I was very passionate about. Now I'll tell you from a very personal point of view. One of the things that I really like about hri is that it's not really just about the…

  about HRI · at 3:19

• I'll give you one example, right? I would say Roombas as a robot have been extremely successful. Yeah. How many people have a Roomba in their home? Lots and lots of people.…

  about Roomba · at 18:18

• Much as possible in whatever you're making. we have a robot at Yale that is called Shuddter. Okay. this robot is really a very general platform for studying human robot interactions in semi-public environments.

  about Shuddter · at 23:44

Who and what is mentioned in this episode?

People

• Marinel Vasquez — Person · Main guest · 1 mention · first at 0:38
  Primary interview subject; discusses her research agenda in human-robot interaction, multi-party social settings, and deploying robots in real-world environments.

Companies & orgs

• Robotics Institute — Company · Affiliation · 1 mention · first at 1:06
  Part of Carnegie Mellon University where Professor Vasquez conducted her post-doctoral research.
  Interactive AI with research spanning fields like computer vision, machine learning and robotics. Before Yel, she was a post-doal researcher at CMU's robotics institute and she…

Products & tech

• Roomba — Product · Case study · 2 mentions · first at 18:23
  Cited as an example of a successful robot that has been integrated into daily life, demonstrating clear value and acceptance.
• Shuddter — Product · Research project · 1 mention · first at 23:46
  A robot platform at Yale used for studying human robot interactions in public environments.

Concepts & topics

• HRI — Concept · Core topic · 1 mention · first at 0:45
  Human Robot Interaction is the central research focus of Professor Vasquez, exploring how autonomous systems can engage with people.

Full transcript with timestamps

0:05 Hello everyone. Welcome to the triple AI podcast series, Generations and Dialogues, bridging perspectives in AI. Through these series, we aim to represent diverse perspectives across generations, fostering rich conversations about the everchanging landscape of AI. Generations and dialogue aims to feature unique outlooks from AI experts, practitioners, enthusiasts of all backgrounds. And each episode will cover important topics ranging from delving deeper into the history of AI to exploring the future ethical implications of its newest technologies. My name is Ella and I'll be the host for today. And today we're honored to have Professor Marinel Vasquez and assistant professor of computer science at Yale University where she leads the interactive machines group. Her research focuses on human robot interaction or hri exploring how autonomous systems can engage with people in multi-party settings, social

0:54 perception and collaboration. She has been recognized for her work in socially interactive AI with research spanning fields like computer vision, machine learning and robotics. Before Yel, she was a post-doal researcher at CMU's robotics institute and she holds a PhD from CMU as well. She's also actively involved in the HRI and robotics communities serving on program communities for top conferences like HRI, RSS, IGCA, IGCAI, and Roman. Thank you so much, Professor Vasquez, for being here with us today. Thank you so much for the invitation. Amazing. So, let's just dive into questioning then.

1:28 So, the first question is actually pretty broad. It's what exactly inspired you to actually choose your research direction to explore these kinds of human robot interactions? So that's a great question. You know, some people say you don't choose problems, but the problems choose you. And so in a way, I feel like my my experience sort of happened that way. So uh let me backtrack and and tell you how I got into this. Um I did my undergrad in Venezuela in South America. And at the time I got really interested in robotics because there was an internal robotics uh class that was using Legos to you know have people solve challenges with with you know uh sort of simple mobile robots made out of Legos that could solve different tasks. Um this is you know very similar to what you see these days high school students building up with with uh these type of tools and technologies. Um, and from there I ended

2:29 up basically engaging myself with the AI lab and doing a lot of different things including a a final project figuring out how to do computer vision from a small scale uh robotic helicopter. Now in Venezuela we didn't know about human robot interaction much at least not the way I would say people would think about it in the United States. Yeah. And lucky for me, when I was finishing my undergrad, I decided to apply to graduate schools and I was admitted to the master's programs in robotics at Carnegie Melon University and Carnegie Melon well turns out it had a long history of work in human robot interaction and kind of serendipitous right like encounter of oh my advisor likes human robot interaction. There's a class in human robot interaction which I really enjoyed and and that just naturally sort of became something that I was very passionate about. Now I'll tell you from a a very personal point of

3:24 view. One of the things that I really like about hri is that it's not really just about the technology, right? It's about how do you use technology for people? How do you make this work out so that people can benefit in in some useful way from having this technology at hand? And it happens right with human robot interaction. You have an embodied AI agent which um in a way has it its its own pecularities. it provides particular affordances that you don't get to see with other types of technologies. But ultimately, right, it's it's this idea of well, it's not just about the technology. It's actually kind of uh bridging that gap to people that got me really interested in it.

4:04 That's really fascinating. And I also feel like the fact that it's honestly like it's really exciting for me as a college student, right? being able to listen to how like in addition to your personal reasons for obviously like how interesting interdisciplinary AI and humans work. It's like honestly just taking classes and then really realizing that this is a direction that you want to continue exploring. That's super fun. All right. And then how would you say your perspective has kind of evolved over time? Yeah. So, I would say not so long ago, but long time ago, um, a lot of the things that I was thinking about were questions of this sort of, well, if I have this robot here do this, what would happen? How would people react to this? How would people perceive the robot? How would this change the dynamic of of that social interaction? M and the more I studied this type of question, the more I got interested

5:01 into, wait, how do I make that robot do what I want it to do? Because it turns out making robots is really hard. And so um a lot of the shift that I'd say I've seen in in my work from my personal perspective is that while we're still very interested in testing out technologies with people, running research studies and and you know checking what happens when you put that robot in in that particular context, um a lot of the questions that we're really focusing on now have to do with how do we make this robot work on its own without having to have a human behind the curtain operating and and you know commanding what is going to happen here and what is going to happen over there.

5:42 Um, and you know, this has been kind of uh out of out of our own experience, right? It's like, oh, let's try to have this robot do this. And it's like, oh, so hard. And so then that motivated us to work on kind of like the more fundamental uh sort of uh open algorithmic type questions that that you may need to solve. That makes a lot of sense. It's like in addition for the robot being able to listen to human commands, it's like how can the robot itself be able to like learn enough from its own personal experience to be able to act on its own? Okay. Amazing. And how would you say that like future generations obviously like everything has been progressing so quickly. How would you envision that human generations of robots or future generations um would be able to contribute to like human well-being?

6:27 Like what do you envision for the future? So let me let me clarify that when I think about a robot, I think about an agent that can change the physical state of the world in some way. Okay, it because I care about human robot interactions, it will also change the social uh sort of environment, right? Because as it's interacting with people, it's also changing that. Um, but this makes a difference with other types of AI agents that perhaps may leave confined to the digital world for instance. Um, and so thinking about in particular these uh agents that would change the state of of real world in a physical way, in a social way, I think there's lots of opportunities to one help people that have some um some task that they need to finish, right? And and this is obvious today, right? You see robots trying to do things like delivery for moving an item from one place to another. You see robots that are trying to, you know,

7:29 move boxes in in a warehouse from one place to another. Um, you see, you know, all sorts of the physical task. I think this is kind of like the most immediate impact that we're going to see. And I think what's going to change is that while these robots typically have been confined to spaces where there's just the robot or there's very clear rules of you know the robot is over here the human is over there we don't like mix together. Yeah. I think these robots are going to become more and more capable of navigating the social world such that they're going to be basically working closer more alongside people in the way that was really hard before both because of safety but you know because of of challenges that may come up in understanding what they should do in those cases. Yeah. Um and so that's like the most obvious to me though the most exciting things are things that perhaps are are not so obvious right away. Um,

8:20 and so for instance, I think there's a lot of potential for using robots in education. There's a number of people, like I'm I'm making this idea, there's a number of people that have studied this for a while. And why I think robots can be very useful is because that embodiment can actually be kind of um a magical uh way to engage people into the learning task. Yeah. Which sometimes doesn't happen with other types of mediums. Um and you know you can imagine right now there's lots of things that we want people to learn. There's, you know, perhaps many situations where we may want to reskill people um because technology is changing, because the economy is changing and for the most part, we don't really have enough teachers to to do this both because of economic reasons, right? Like right now at least being a teacher doesn't pay as much as other jobs and that makes it really hard for people to take this job and even

9:17 then it's a very difficult job very demanding and so I could imagine having robots are not necessarily replacing teachers but are serving as um I would say you know supporters to help enrich the educational experience and help support the learning goals that a teacher may have for a particular student. And so I think, you know, there's lots of opportunities here and I can imagine, you know, robots teaching people lots of things that maybe we don't think about too much. Things like how to navigate an interview for a job.

9:48 Yes. Things like how do you um perhaps become a very good communicator? You know, these are things that in a way are very fundamental skills that if we can do this well, we're probably going to be very successful. But how many learning opportunities have you had to really have someone dedicated to teaching you that is probably not so common, right? And so I feel like robots can help fill that kind of gap. Definitely. And going back to your point on education, I actually have a friend that I met here um at Stanford who throughout her entirety of high school, she didn't have a single teacher. She was taught by AI. It was a very small high school, but I feel like she was just telling me it's incredibly productive because they're able to really like like be able to pinpoint your exact problem and then customize like teaching you according to that way.

10:41 So, I feel like it's really interesting to see how like AI is able to accelerate honestly even like improve the lives of teachers to be honest like whether it be like grading tests even, right? It's just all of these like small little tasks could be automated. That's amazing. All right. And I guess going on a little bit of a different direction. Um, if there was one research outcome or innovation from your career that you could highlight, what would you highlight? Um, so I would say one particular research direction that we've been pushing forward for a while now is this idea that in human robot interaction, we should sort of make uh a first class citizen the relational structure of the data. And I know this sounds very technical, but I will explain. Yes. So, um, a lot of work in human robot interaction has focused on one-on-one interactions. I have a robot, I have a person, and how do I make my robot work

11:37 for that person? Yes. And when you do that, from a computational standpoint, for the most part, all you have to worry about is encode the information that you need of that person in in your computer in a way that you can use it to, you know, make your robot do whatever it needs to do, right? Now, when you start having multiple people, well, you might imagine, oh, well, the same the same applies. Just encode the thing for this person, the thing for this person, the thing for this person, and you're done, right? But it's actually not as simple as that because different people can influence each other. M and the relationship between how they're interacting with each other and with the robot actually matters a lot. Yeah. And so computationally if you think about it it's not like oh I just have a list of things probably you want to represent things as a graph because there's lots of relationships between these different

12:39 agents. That makes sense. Now, this is something that people have been doing for a long time in fields like multi- aent systems, but to my surprise, we haven't really done that much of that in human robot interaction, which seems strange. Yeah, very interesting, right? Because again, we have these different agents that are interacting together and there's all these relational information that matters. Why not encode information these graph representations and then try to leverage algorithms that take advantage of that representation that in a way um kind of have an inductive bias to to make sense of that data from the beginning. Yeah. And so that's really um a lot of what we've been doing lately.

13:21 This is uh kind of the core idea of my career NSF uh grant that I have and I think it's very important because both in terms of making sense that we can generalize right we can extract information from this type of perhaps an interaction with two people or three people and then see what happens with one person or four people trying to generalize to to this new context I think this relational structure is very helpful but then also I think it can help from a sample efficiency point of view because we're taking advantage of kind of the underlying structure of the data rather than fitting into some other structure that in a way doesn't quite fit what the problem is. Like that makes sense. It's like nowadays especially for robots and honestly in AI in general, right? Like data is so valuable like if when we're trying to customize one person's experience, we can utilize other people's while obviously like

14:10 protecting privacy and everything. It's like that would really allow for us to efficiently customize like certain experiences. That makes a lot of sense. And do you think that there was a key moment where there was would you say this was a significant change in your research direction or what were you working on before um you kind of had a revelation that this was maybe the way to go? Yeah. So actually I was working on um multi-party human robot interactions but not from the point of view of how do you model computationally this these interactions and again because I was shifting right from oh what happens if I have my robot do this and that to how do I make my rob more autonomous it sort of felt like a very natural transition to start focusing on on how do we leverage these type of relational models for hri um as a kind of a way forward from a technology point of view. That makes sense. Okay. Amazing. Thank you so much.

15:08 And then uh the next question is a little interesting. I was thinking about how nowadays we have so many unfortunate like horror movies which really depict robotics in a pretty like some robots in a really negative sense. What do you think is some of like the most um problematic kind of perceptions that some people might or especially young researchers or the younger generation might have regarding like social robotics? Like how are we able to address those kinds of challenges? H So can you give me one example? Yeah, maybe that can help ground. Um, I'm forgetting the name of the movie, but I remember there was there are a few movies nowadays where it's like they depict robots as like dolls, right? And then the idea is like this doll is initially trying to improve like your human life and then as everything progresses, they take like your instructions too literally, for example, and then they end up becoming this like

16:10 unstoppable force which kind of creates a lot of chaos in people's lives. I feel like that might be one of the reasons why some people without like as much exposure necessarily to how AI really works might feel very like scared for like what it really means to have robots incorporated into our so like daily lives. Um is there a way that we're able to maybe address this or what are your thoughts on um how or like what kind of misunderstandings might they be having? Yeah, I follow. So I think there are definitely many challenges in introducing new technologies in whatever human environment. Um and this is not new right this happens with any new technology that comes up. I think first from the design point of view of of making that machine is very important that as much as possible we think about how we can take input from the target users to make that robot and you know there there are methodological um approaches to do this

17:20 things like uh participatory design of technologies is one way to do that. Um, and I think this is very important because it's not just sort of magically, oh, this robot shows up here, but it's more like there's a whole process that goes from making the robot, then introducing the robot and in a way selling it, pitching it to that target uh set of users that you know it's not like a one thing is it's actually a collection of activities that all matter to make sure that things go well. And so I would say from the beginning it's very important to make sure that uh in a way you take that input and you're not just assuming that you are the one that knows everything because probably the way that you see the world is very different than how the the ultimate user may see the world and what they may care about. Um and then the next thing that I would say that is very important is make very clear what is the value of

18:16 having that technology there. And so I'll give you one example, right? I would say Roombas as a robot have been extremely successful. Yeah. How many people have a Roomba in their home? Lots and lots of people. Right. Exactly. And you know, if anything, it's perhaps one of the most successful robots that I can think of today where people may not even think about it as a robot because it's so part of their daily life, right? Um and why how how does this happen? Well, first there was a very clear use case, right? Oh, I'm I'm going to be there to help clean your floor. Yeah. Right. And so I think that's very easy for people to to understand, to manage, to evaluate, to to judge. Um and then second, I would say, you know, if you start looking around different uses of Roombas, people then went crazy and started doing all sorts of things with the Roombas, like putting their cat on the Roomba, making all sort of funny

19:07 videos. And so like these Roombas in a way kind of became part of part of their life, right? Yeah. Um and so I think you know when you think about integrating this technology you have to see well okay what are my assumptions are my assumptions correct and then having a very clear message as to how this is going to benefit people and if for whatever reason things are not working out the way that they should then you know pivoting backtracking changing updating that system so that hopefully it does work the way people would want right like sometimes there are things that may happen that you don't expect and you have to change the way that you were looking at problem, but it's only with with sort of that communication channel being open that you can realize these things early on. That makes sense.

19:49 That makes a lot of sense. And would you say this has a lot to do with like not like as you were saying like not just focusing on like one specific field or like one specific aspect of robotics but being able to incorporate like interdisciplinary collaboration across different fields and then really taking into account what exactly it is that humans really need and what the purpose of um or what the impact or applications the robot that I'm creating would have. Amazing. Okay. And I guess kind of following up on that then, um, what do what would you say would be like ethical considerations that are most challenging to integrate into robots as we further um, increase their user case and human experiences.

20:34 So I think one of the biggest challenges that we have is an open challenge um, that is very interesting to think about is how do we recognize when errors happen? Interesting. And I don't mean like obvious drastic errors that we know can happen, you know, like an autonomous car crashing. This is obvious, right? We know we don't want that. That's already we've been we want to avoid it from the very beginning. Yeah. Um but I think there are um challenges that are unknown that can happen as you are sort of deploying this technology in the real world and being able to somehow recognize that something went wrong would be a very important thing to do in terms of making the systems in a very ethical way. Right? And if you think about it, right, like we humans, we navigate our world being able to do that to some extent. We may not be 100% accurate, right? But sometimes I say something and I see how

21:37 people react and I'm like, whoops, did I say something that I shouldn't have, you know? I see. I see. That makes sense. Yeah. And and it's not something that I I had already kind of preconceived that that was going to be like an error or something that I shouldn't have said or anything like that, but it just happened that I have a mechanism that I in sort of in my mind I can use to at least trigger an alert of wait what maybe something went wrong. Um and so I think this this is actually a very interesting problem. We're we're right now uh doing a lot of research trying to leverage implicit human signals, implicit communicative signals, things like, you know, it can be expressions, but it can also be perhaps different actions that a human takes um that communicate some kind of information. Um and trying to use that in order to make robots a little bit more clever about understanding what's happening in their

22:28 interactions. And you know, ultimately one cool thing that we could do is potentially try to recognize when an error occurs. Um, exactly. That makes sense. It's like almost having like an automatic feedback loop, right? Like being able to not just have the very obvious errors, but even like the small ones when we're interacting with humans, like and then being able to learn like, oh, what kind of reactions are like considered good or like um desirable versus which ones may lead for us to like question maybe what the robot did or said. Okay. Amazing. That makes a lot of sense. Yeah. And importantly, right, it may be that the robot is wrong in making a prediction, but at least it knows that probably it's a good idea to check Yeah. if I'm making a mistake, right? Like Yeah. Yeah. Yeah. That makes a lot of sense. Okay. Amazing. Um, and I guess kind of adding on to that then, how do you think we can bridge the gap

23:21 between like research advancements and then public acceptance and usage of these kinds of robots that you're making? Maybe if you wanted to talk about like a specific robot like um that you're currently working on or um anything along the lines of that that would be amazing. Yeah. So I think like I mentioned before right part of of this process in a way involves thinking about the users and trying to engage them as much as possible in whatever you're making. Um we have a robot at Yale that is called Shuddter. Okay. Um this robot is really a very general platform for studying human robot interactions in semi-public public public environments.

23:59 So we use this for instance to think about problems like can we build models that predict human intent um or for instance you know trying to think about whether we can uh enable the robot to recognize when something goes wrong interaction kind of like I was thinking before and um putting this robot on a public space in a way has meant you know there has been a lot of discussions on how do we do this and how do we make it work and I think you know from thinking about the um kind of ethical deployment data collection point of view. We've had to uh sort of you know discuss how we're doing what data we're keeping private what data we're releasing with people at at Yale that regulate uh these type of activities. We have also, you know, had to talk with the people that that manage spaces. You know, is this a good place to put the robot? Why do we put it over there? Would this be, you know, a problem for any activity that is

24:50 happening nearby? And we've also had to think about, you know, how the robot operates in its environment, right? Like, is this a place where it can be making sounds or is it going to disturb a class that is happening nearby? I see. And you may think, you know, these are all very perhaps not even not very research focused questions, right? They're more of a process question. Yeah. But you kind of have to engage with all these stakeholders that in a way are going to help you succeed, right? Yeah. Um and there has been a lot of insights that originally we didn't think about. Um even things like, you know, where we're putting the cameras in relation to other cameras in the environment and things like that just so that we're not violating people's privacies and things like that. Um and so yeah, this is just one example from from one of the projects we have. Yeah, I feel like as you're mentioning it now,

25:38 it's like these seem like taking into consideration like whether the environment that you're testing is like allowing for noise or not and then being able to ensure that you're respecting the privacy of other people who are in the environment that you're testing your robot. Like these seem like very obvious like problems, but then like you I feel like for me at least as you were talking I was like oh that is something which is like really important for us to consider when we're doing these kinds of experimentations. So I feel like that's something which is really interesting regarding like the this combination of like AI and technology and then being able to implement them in robots and then deploying them in human environments. So that's super interesting. All right. And then yeah let let me add one more thing that I think is is very important um and often drives a lot of the way that we do

26:23 things. So human expectations are critical to the success of these robots or any other AI technology for that matter. Right. Yeah. And so what we have seen in general is that if humans have expectations that are very low Mhm. when they encounter the technology, if it turns out it does a little bit better, people will be like, "Oh, you know, great robot, you know." Yeah. Yeah. Yeah. Yeah. If people have expectations up here and then the robot operates at this level, then people will like suddenly the expectations plummet.

26:58 It's like, "Oh, this robot is terrible." you know, it doesn't do what I was hoping it would do. Um, and this applies to to everything from understanding having a mental model of how the robot works to how the robot looks to to everything that that we're doing with the robot. Um, and so I think this is another very very important aspect that when you think about integrating the robot into into some particular application matters a lot, right? Create the robot in a way that you can set the right expectations so that people are not up here when the robot is actually operating at this level. So that hopefully you know engagement continues rather than oh you know there's this robot in there I want to avoid it now.

27:32 That makes a lot of sense actually. Yeah I feel like no that makes a lot of sense. It's like not just uh in terms of objectively how well the robot is able to do but also like how society is able to accept it or um expectation versus reality and how that like kind of alters its actual usage. Right. So that makes sense. Um yeah. Yeah. Then do you have any advice for the next generation when it comes to being able to balance like technical innovations with robots and then societal needs? Like what is a good way to be able to merge these two together? H So it's interesting because the way the question is phrased in a way sounds like these are opposing uh options, right? Because like you're balancing, right? It's like are you going this way or are you going that way? Yeah. But I don't think they are.

28:18 In fact, I think there's lots of interesting technical questions that arise when you try to have some kind of technology that does good. Yeah. Um and that can have a positive social impact. And so I would say if anything try to think about how you want to help people first. Yeah. And then backtrack and think about what kind of technologies could be good for that. And I guarantee you there's going to be interesting technical challenges in making that happen. This is almost a guarantee. And so I think starting from that direction is a very easy way to to find you know some common place to push forward both ideas. Yeah. Okay. Amazing. That makes sense. And would you say that the role of like what kind of role does diversity kind of play when it comes to designing like social robots for the next generation like not just for like a specific demographic but kind of like for everybody like how can we ensure

29:10 that the robots that we're making are inclusive and diverse. Yeah. So this is very interesting because in a way it poses an impossible challenge. Yeah. What do I mean by this? Right? So imagine you have person A. Person A likes X and I have person B. Person B does not like X. And now you're asking me how do I make a robot that works for everybody if person A and person B do not agree? This seems impossible kind of by definition, right? Yeah. And so what I would say really, well, there's no one robot that works for everybody, but we can make robots that adapt and become really good at helping people in the things that they need and the things that they care about. Okay. And so I think, you know, thinking about adaptation, personalization, that's really the key to making these robots really work in in different situations with different people. Now, getting back to your question of of why diversity matters.

30:07 Well, this goes back to the idea that I mentioned before that, right? Like I I cannot assume I know everything that matters for person A or person B. Yeah. What are what are the chances that I do very little, right? Imagine that I was making a robot that helps my grandma in the kitchen. I mean, I know my grandma very well, but in no way I see the world like she sees the world, right? And so having diverse opinions, diverse backgrounds is one way that you can in a way increase the chances that you're actually getting closer to the perspective that you need to design the right thing. The other idea I would say is that actually there's a lot of research that shows that having diverse teams um can contribute to being more creative, solving challenges better even for companies to be more successful economically. Um and so you know there's lots of research that supports the idea that bringing together these different

30:59 people not only is useful for perhaps making you know a good robot if you wish but also just generally for the success of of that unit as a team. And so I think there's lots of value there. That makes a lot of sense. Thank you so much. And then I guess our last question um for today is how do you think your current work will influence the next decade of human robot interaction research? Um it's kind of open-ended again but yeah. Thank you so much again for your answers previously. Yeah. So I will mention two things. Okay. The first thing is that through my work, it's not just my work that's happening. It's my students work that is taking place. Yeah. And via my effort of working with them, they're actually growing. They're becoming experts in their own sort of, you know, thesis topic or or particular project that they're working on. Yeah. And I think this is going to have a tremendous impact in 10 years because if you

31:57 imagine them as students, they're going to grow and do other things and there's this like exponential effect that suddenly happens, right? And so I would say that's like number one immediate thing that's going to happen which I particularly like a lot. Um the the second thing that I would say more from a research standpoint um a lot of what we have seen especially in human robot interaction is that we do research in very controlled environments and there's good reasons for this right um if you are bringing someone to a lab where you're controlling for as many things as you can then you can manipulate certain aspects of the interaction to understand very well the effect of very specific variables on the outcomes on how the interaction happens. Right? So for instance, if I bring people to my lab and I always have the person sit over here and the robot over here and I control how the robot looks around and

32:52 everything else is the same, then I can be sure that if there's a behavior change on that person or or that population is probably because of the gaze of the robot, not something else. Right? So I'm I'm basically reducing the the chances of compounds, right? And so this is great from a scientific point of view in terms of understanding very fundamental relationships between variables but is not directly applicable to having a robot in the real world right there right because my lab is is not out there in the real world right is much different and so I think a lot of what we're doing now really the the medium-term goal is try to help us push forward away from the lab so that we can start studying these interactions in more unconstrained environments where hopefully people's behavior will be more natural where hopefully the type of applications we can do would have in a way more tangible impact right because

33:48 it may not just be this um sort of particular task that we want to do in the lab but something that actually maybe the robot would be helping with um and so I think that transition is something that a lot of the kind of fundamental research that we're doing to improve robot social skills is going to help us do right making the robot more robust for more complex social environments so that hopefully we can have them better operate in in real world situations, not just simple lab environments. That makes sense. That makes a lot of sense. I think it kind of goes back to what you were saying in terms of generalizability, right? like being able to ensure that um when we're moving from the research laboratory into the real world, we're able to take into considerations for how each individual might have different preferences and then being able to utilize that kind of data to improve each other's

34:34 experiences. Okay, amazing. So, I do think that is time. I think those are all the questions for today. Thank you so much for your time again, Professor Vasquez. And yeah, for anyone who's watching, we will be sending out or we will have published by this time a form to be able to fill out if you're interested in being a guest. Um, thank you so much again, Professor Vasquez, for your time and we look forward to seeing you guys in the next episode. Bye-bye. Thank you for the invitation. Thank you.

Questions about this episode

Who speaks in "Human-Robot Interactions and Social Robotics with Professor Marynel Vasquez"?

Marinel Vasquez (Guest).

What is discussed in "Human-Robot Interactions and Social Robotics with Professor Marynel Vasquez"?

In order: 0:00 Marinel Vasquez, HRI, Robotics Institute; 15:00 Roomba; 20:00 Shuddter.

What is mentioned in this AAAI episode?

5 people, works and ideas are mentioned, including Marinel Vasquez, HRI, Roomba, Robotics Institute, Shuddter. Each one on this page links to its own page with every episode that mentions it.

How long is "Human-Robot Interactions and Social Robotics with Professor Marynel Vasquez"?

The episode runs 35 min; the transcript is about 6,467 words in 42 timestamped sections.

More from AAAI

• All analyzed episodes
• People mentioned on AAAI
• Topics mentioned on AAAI