#595: The Neuroscience of VR Training with STRIVR’s Michael Casale

michael-casale-2017STRIVR started as a VR training platform for elite athletes for college and professional quarterbacks, but they’ve been recently expanding into corporate training for Walmart. Over 200 Walmart Academy location will soon be equipped with virtual reality training for managers and employees to improve social skills but also get prepared for large-scale events like Black Friday.

I talked with STRIVR’s Chief Science Officer Dr Michael Casale at the VR Strategy Conference in San Francisco. He describes the neuroscience of why VR is such a compelling training platform including the embodied cognition insights into being able to be immersed within the context, and to simulate the process of making choices and taking action. The depth of learning is so much more rich in VR, and it’s a more holistic approach for learning that is also opening up new epistemological methods for objective measures of engagement that will hopefully reveal a deeper ground truth of how effective of a transfer learning processes they’ve developed. Casale found that engagement is a key indicator, which can help them find patterns of reliability and validity in other factors like how much someone moves their head and what people have been looking at.


STRIVR has also been exploring the implications of an embodied cognition insights. For example, what are the implications of performing the physical act of smiling is the cause for a change your mood? Perhaps focusing on resultant behaviors through embodying the actions directly is what leads to changes in attitudes and cognition, rather than the other way around. Another open question is how to model and measure social behaviors, and that’s something that STRIVR co-founder Jeremy Bailenson has been researching at the Virtual Human Interaction Lab that he founded at Stanford in 2003. There are many signs that one of the VR killer apps that drives adoption in the enterprise will be training, and STRIVR’s platform is pushing the edge of the best practices of showing and measuring what’s possible.

This is a listener supported podcast, considering making a donation to the Voices of VR Podcast Patreon

Music: Fatality

Rough Transcript

[00:00:05.452] Kent Bye: The Voices of VR Podcast. Hello, my name is Kent Bye, and welcome to The Voices of VR Podcast. I think training and virtual reality is one of the killer apps. It's actually one of the things that was driving the innovation of virtual reality since the inception, the very beginning. And then when it comes to using virtual reality technologies to be able to train astronauts and flight simulators and military situations, it's something that has been a consistent thread throughout all the history of virtual reality. And one of the trends that I've seen at the last couple of VR conferences, whether it's at Oculus Connect 4 or the VR Strategy Conference or the Immersive Technology Conference, is that there's this move towards the enterprise and starting to discover the specific context and applications where virtual reality training is going to make a big impact. So Stryver is a VR training company that's been focusing on elite athletes, specifically quarterbacks, both in college football as well as in the NFL. And so they have all these different techniques to train all the different plays and teach the spatial memory and Essentially, they're able to tie together the process of being in the context, have your emotions engaged, and being able to make choices and take action. All those things put together is how we learn. And they actually had some people from different corporate enterprises, including Walmart, to see their system. And they were like, you know, actually, this is very applicable to what we need to be able to do to be able to train our employees how to have these different interpersonal and social skills. So Stryver's been expanding out from sports and moving into the enterprise. And so I had a chance to catch up with Dr. Michael Casale. He's the chief science officer there. And I had all sorts of questions about what is it about virtual reality that makes it such a compelling training platform? And how are they able to actually measure and track how well they're succeeding? So we'll be covering all that and more on today's episode of the Ways of the VR podcast. So this interview with Michael happened on Thursday, October 26, 2017 at the Virtual Reality Strategy Conference in San Francisco, California. So with that, let's go ahead and dive right in.

[00:02:13.174] Michael Casale: So I'm Michael Casale, I'm the Chief Science Officer at Stryber. We're an immersive training company using things like virtual reality to train anything from sports athletes to enterprise employees.

[00:02:28.670] Kent Bye: Great, yeah, so I think the last time that we talked was at the Experiential Technology Conference 2016, and that you were, at that point, maybe a year or two in from using a lot of your training system to train quarterbacks in football. And now it sounds like, you know, over the last year or so, you're getting into expanding into other types of training and corporate training maybe you could sort of talk about the evolution of like how the football thing is going with training quarterbacks and elite athletes and then how that sort of spun out and expanded into corporate training.

[00:02:59.814] Michael Casale: Great question and you know on the surface they seem like two pretty disparate areas two different kind of skill sets but as we're learning and as I kind of suspected there's a lot of overlap so You know, in training athletes, it's often critical to get the right kinds of what I would call perceptual fidelity and even effective fidelity. So a lot of the things that you get out of what we're calling immersion is the fact that you're seeing the same thing in the real world, you're reacting to it in the same way. And that's kind of a seamless process. Traditionally, you would get trained kind of in parts and in pieces. Here, we're kind of giving you the whole thing. And that's going to be more translatable. And as we're finding out, and as other studies have borne out, that's more translatable to the real world. Well, similar thing in the enterprise space. So when you think about training your employees, even if it's for something less cool than like a quarterback throwing for a touchdown, it's still important that, you know, you're kind of getting the same perceptual experience so that you know how to react. The decisions that we make and what we know about how the brain actually learns is that you get perceptual information and the sensory information comes in, you're mapping that in your brain, and then you're making a decision, you're mapping that in your brain, and those two things get connected. And if you're getting kind of connections between things that aren't represented in the real world, you might learn in that local training space, but that learning isn't necessarily going to apply. So that basic principle will apply to almost any of those types of learnings, whether it is in the football space or basketball space, or whether it's actually in the enterprise training space.

[00:04:26.470] Kent Bye: It's interesting and Amy from Accenture was giving a talk yesterday and showed the cone of learning and how at the highest level you have like text and abstraction and then as you go down with audio and video and then sort of interactive experiences and then Basically the more immersed you are the more higher fidelity it is of an experience and that I guess the more likely it is for you to remember it. From I guess the science perspective how do you kind of think about that or quantify that sort of as the higher levels of fidelity of experience why is that better for retention?

[00:04:59.586] Michael Casale: So that's just like basically depth of learning, or I would simply call depth of learning. So going back to kind of what we know about the neuroscience of learning, you're basically hitting on a lot of different elements of that learning environment. So if you take her example and you think about text, those are just simple words that are printed in a, not that they're always simple, but they're simple words printed on a page that are going to evoke some area of the, say, the visual cortex. And you're going to learn to map that area of visual cortex to a decision. Okay, now you incorporate sound. So now you have the visual cortex activated, you have the auditory cortex activated, and you're going to map that decision. Okay, now you're going to bring in emotion and kind of affective quality. Well, now you're going to map all that stuff. So basically what you're doing is you're creating a more robust experience so that when you see that thing in the real world, that perceptual thing that you're supposed to react to in a certain way, you're more likely to react to it because you're having all of these things kind of activated instead of just one kind of part of that system of that learning system activated. So if it's just the visual area that's being activated, that's much less likely to evoke that same kind of like natural response in the real world. than if you're getting the compendium of the sensory information through affect, through auditory, through touch, through all these things. So there's a good story why those slides that she showed actually translate and kind of coincide well with what we know about the neuroscience of learning.

[00:06:20.595] Kent Bye: Interesting, yeah. And just from what I've looked at from creating an experiential design framework for VR and then looking at AI, there's like this similar kind of mapping. And so I see it as like these different dimensions. So at the highest level you have cognitive intelligence or mental presence. There's also a social presence and social intelligence that sort of maps between VR and AI. And then sort of going down I guess you have then manipulative intelligence or the ability to express agency and take action on decisions that you may have made. So you may be in your mind making a decision and taking action on those. And then you have the kind of an emotional layer which I guess it starts to get into more and more into the subconscious or unconscious processing of like there's an emotional presence but also emotional intelligence that you have. And at the bottom layer is like the most rich sensory experience. It's all of our senses that are coming in. And so it's our sense of embodied presence and environmental presence, but also our kind of perceptual intelligence. And so to me, like VR kind of activates all of those things. And I don't know, from the neuroscience, if you would have another way of saying that.

[00:07:24.753] Michael Casale: No, no, exactly. And just, yeah, kind of what I was alluding to before, that's exactly right. what we would kind of like colloquially call the richness of the experience is exactly what you just described. It's kind of like the concert of all these different areas producing that behavior. And so you're kind of getting a better type of learning because of the depth of learning, if that makes sense. That's at a very high level. And again, there's a lot of studies both in the behavioral literature and the neuroscience literature to back that up.

[00:07:51.185] Kent Bye: Great. And so it sounds like that you're starting to make headway into the enterprise market. And so maybe you could share a little bit about what you've been able to do in terms of the types of training and the stuff that your system can do.

[00:08:04.418] Michael Casale: Yeah, it's pretty varied, which is pretty cool. So when people first see this, especially like in the sports training, it was a pretty big leap for that first person. And this person was actually from Walmart that saw the things that we were doing in sports. And they decided that, hey, this is a lot like what I have to do to train my folks to be better at their jobs. And again, it's the repetition. It's the perceptual fidelity, it's being able to make decisions naturalistically and in real time. All that stuff is going to lead to a better training experience. But even all that aside, just engaging people in a different way about their learning. So currently, I don't know how much you know about the traditional learning methods, but oftentimes they're in classroom settings, they're observational, so they're very passive, you don't get to make responses, you don't get to manipulate that information as you're learning, that active learning that we know is better. None of that stuff really occurs, because it's not efficient. I'm sure people would love to do it that way. Well, answer VR. VR is able to give you all of that, and then some in terms of the engagement. And so knowing that we have a better training experience for basically the same cost, or even sometimes cheaper, because it's less costly to put people in VR in that same environment than bringing them to the store individually and systematically, that takes up too much time. And again, it makes folks also feel invested. You mentioned agency. That's a powerful thing. Being able to be in an experience and knowing that you're acting and you're kind of affecting that environment, your decisions are creating the way that that environment is going to shake out, you can kind of learn better how your, basically, decision-making is actually going to matter. And I think for a lot of these folks who are going through this training, they don't have agency. Again, they're kind of anonymized in these rooms. Putting them in that experience, knowing that they're actually kind of like the masters of their fate, at least for that little training experience, is a really powerful thing. So these organizations are recognizing that VR can be a way to do that. Now, VR is not always the solution, right? There's going to be things that are actually better in the 2D world or just not possible in VR. We're starting to broach a lot of not just the rote training, but more of kind of the complex people, social interaction training. And that's where I think VR is going to make the most impact, albeit it's got the most challenges associated with it. But the fact of the matter is if you can start to make headway even in changing behavior about how people interact with one another, about being sensitive to things that are hard to describe in words, nevertheless they're taught that way, that's going to be a much more powerful and we think effective experience. And so a lot of these companies are focused on making just better people, I guess, in general, for lack of a better word. As Orwellian as that sounds sometimes, there are certain things that we know, being a better listener, being a better social kind of interactor with your peers. being able to have difficult conversations with them in a more effective way. I mean, these are all things that happen at work, and we know things like retention, employee churn, those are affected by these kind of micro events socially that happen every day. So if we can start to kind of make headway in facilitating and kind of enhancing performance in that domain, that's really, I think, going to be a really special thing. Again, it's got the most challenges associated with it.

[00:11:03.241] Kent Bye: I'm curious to ask you, because you come from a neuroscience and learning background. It's from an academic perspective where things are very separated in a certain way. And it sounds like VR, in some ways, is this holistic medium that's able to combine things together in a new way that transcends the reductionistic way that you may have previously been looking at it. But it offers you, as a scientist, new ways to tie it all together.

[00:11:30.380] Michael Casale: Yeah, that's a great point. I hadn't actually thought about it in the way that you just articulated it. But yeah, for sure, the academic world, almost by nature, and because it kind of has to be, is reductionistic. To be able to systematically understand and test hypotheses, you kind of need to modularize it to some extent. To your point about then being able to kind of bring it all together in these kind of more naturalistic environments in VR, there's a lot of power in that too. So our co-founder is actually a guy named Jeremy Bailenson who runs the Stanford Virtual Human Interaction Lab and for years he's been trying to understand social behavior through the medium of VR because we know in the lab, to your exact point, even with social behavior you're still kind of taking this reductionistic approach, right? Well what about this social phenomenon? The VR environment allows you to manipulate all those variables efficiently and effectively so you don't have to actually replicate that over years studies. You can do it over a series of weeks, you know, in virtual reality by just being able to manipulate the CG environment and seeing how people react. It's an extremely powerful tool just for social behavior, but I'm sure for all kind of other fields of study.

[00:12:33.281] Kent Bye: Well, I think one of the other things that I've seen emerging is right now you have a split between the objective data that you can measure, and then there's the direct first-person perspective of the person going through that experience. And so what I see is this thing called neurophenomenology, which is to be able to look at the correlates of the neurology that are happening, but you're actually trying to correlate someone's first-person perspective with the third-person sort of objective data. Is neurophenomenology something that you've been looking into in terms of how do you combine somebody's first-person perspective with the data that you're able to capture?

[00:13:06.890] Michael Casale: That's a really interesting point and I think if you really want to get more certainty around it, looking kind of objectively at kind of what the brain is producing in terms of its kind of like physiology and then how that correlates with these subjective experiences is the way to go. I know Jeremy's kind of done some of that work. At least with more basic cognition, like learning and memory. But things like, did they enjoy it? Were they engaged? I think that's ultimately the way to go. Obviously the problem with a lot of these neuroscience techniques, while they're non-invasive and they're getting better all the time, it's still not trivial to put people through a scanner, an fMRI scanner, or even put an EEG. instead of electrodes on their heads. That said, I still think there's a lot of power in doing just the behavioral correlate. So while it's probably more desirable to have that neural correlate, looking at just their objective behavior, how they respond and how they reacted, while it's a little bit more of a gross measure than the neuroscience provides us, it's still a good objective measure. So we can actually look and see, hey, maybe some of these behavioral measures that we're collecting that are objective, their performance on a task, how they actually interact with the environment, which is very objective, which is something that's very unique to VR, by the way. Seeing that a learner is sitting in a classroom is often the assumption that they're learning, but now we get to see individual differences and like, hey, are they more engaged? Are they exploring the environment more? Are they interacting with the environment more? Those are really objective behavioral measures that I think are sufficient enough to tell you something about their subjective experience. So we are looking at correlating how they're performing in the VR space behaviorally and looking at their subjective qualitative experience as kind of a validation of like, hey, yeah, they were more engaged. And of course we know a more engaged learner is going to be a better learner ultimately.

[00:14:46.457] Kent Bye: Yeah, I think part of the challenge is that because there's like a level of cognitive awareness and then maybe like even the level of your behaviors and the way that your action may be starting to get below the level of conscious awareness and so you're making behaviors, you're having emotions, and you're having sensory input, but yet all of that is happening at an unconscious dimension that you may not be able to actually articulate. And so maybe some people made decisions before they're able to kind of make a story about why they're doing something. And so I just see that, you know, as a research scientist looking at this and looking at these human behaviors, you have the challenge of being able to try to quantify this sort of holistic experience and try to figure out what's going on and that, you know, just looking at the challenges with presence research is that, you know, the fact that somebody's reporting about something after the fact, it's sort of like they have to recount something that's a part of their sensory experience, but then whenever you have to articulate something into words, there's some sort of information loss that happens there, and so they've sort of found that. those surveys aren't necessarily always useful in the sense that they are giving you a number but you can't add those numbers across multiple people to say okay generally this is the experience of presence because everybody is going to have a different way of kind of quantifying their degree of presence. And so it seems like there's larger issues as you as a research scientist of like trying to take all these kind of mix of objective and subjective quantitative and qualitative experiences and then try to like measure like how do you know what you're doing is working.

[00:16:13.038] Michael Casale: That's a great point. So there's a lot of things to unpack in what you just said and they're all very important and nuanced points. And so the first thing you mentioned about kind of the lack of basically what we would call reliability in the subjective data, meaning that people are using the, like if you give them a scale to rate themselves on engagement, they're going to have different associations with what those numbers mean or even their definitions. I mean, you can't just ask someone if they were engaged or not, cause that's going to mean different things to different people. And then beyond that, if you try to get objective about it, typically looking at engagement means looking at the behavioral effort or output. And that means different things for different tasks. So there's really not a great unified way to be able to say from exactly to your point, one person or another, that when I see a two on an engagement survey, that that means the same thing for everyone, which really kind of renders it not very useful at the end of the day. Again, that's why VR has the power to kind of provide a level of insight and objectivity into things like engagement. That all said, the third issue, which is, I think, kind of what you're at least alluding to, is how do you really get ground truth for these objective metrics? It's difficult, and I don't know, I don't know that anybody's really solved that problem yet. What you can hope to do is see, like, so, of course there's gonna be noise and variability for all the reasons that we just discussed around, like, the subjective measures of engagement. The best thing you can do, and this is typically kind of in behavioral sciences, when you have these kind of, like, tough times getting ground truth, or tough times figuring out if there's just one measure. There's typically not. So ideally, you have several measures, or a handful of them, all pointing in the same direction. And then you can feel a bit more confident. So it's kind of this converging evidence. As a research scientist, I never would look at one metric and say, even if you think you demonstrated reliability, validity, and even some association with some sort of behavior, I would never take that on its own. There's always going to be ways to pick that apart. you give me six and they all start to say the same thing and converge on the same conclusion, I feel a lot more confident about that. So for us it's really going to be a challenge of like finding other ways to kind of measure that ground truth and hopefully those all agree with each other and then we can find if there's associations between those measures and these objective measures of engagement. Ultimately, if we're able to kind of validate that so that we have an objective measure that we know is a great indicator of something like engagement, for example, or like learning, then we can just use the objective measure because we know we have a lot more faith that that's reproducible, that's going to be the same thing. I actually wrote a paper on this. We have it published on our website. There's a bigger, more technical version coming out soon, but it's taking our existing data from, I think, thousands of users at this point, real-world users, going through one of our experiences and trying to measure their level of engagement with the experience. We did provide some validation, though, and the way we did that was with another behavioral measure. So not just asking them whether they were engaged, but if they quit the experience or not. So we allowed them to be able to say, no, I don't like this experience, or like, yeah, sure, I want to keep going. Those that made it to the end, were able to basically show more exploratory behavior. They showed kind of more volume of head movements. And again, these are things we can track with VR that you typically can't otherwise. So that, to me, gives me a lot more faith that that measure of engagement in that example is going to be indicative of some behavior that we care about, which is, I don't want to do this anymore, or like, yeah, I want to keep doing it.

[00:19:31.459] Kent Bye: It reminds me of both YouTube and Netflix moving from like a five-star system to like either a thumbs up thumbs down because rather than trying to average out the stars across many people it's more of a binary like did you like it or not and that's a little bit easier to combine across multiple people and it sounds like this sort of equivalent binary is like did you shut it off or did you stop and that is a maybe a better signal than sort of having them try to from a scale of one to seven rate their level of engagement.

[00:19:58.517] Michael Casale: Yeah, and I think it's something we care about. I mean, I can't speak for the big production companies, I'm sure they have great, huge research teams on this, but if a movie got three and a half stars versus four stars, I don't know what that means. Maybe they do, but it seems like that's kind of a difference that, to your exact point, could be attributed just to noise and how people are using the surveys. For us, it's a meaningful construct. Hey, they want to do this experience or not. As an organization who's interested in getting people to use VR, like, yeah, I actually care about that. So being able to tie it back to like something not only behavioral and objective, but also like something that means something to me. and that logically makes sense, something I care about, is critical. So we were fortunately able to get that. And the fact of the matter is, we created an experience where a decent amount of the people didn't want to do it. If we found that everybody wanted to do it, then it's not very useful because then you can't really make that dissociation. But for whatever reason, they didn't like the VR, they didn't like the headset, it could have been the experience, whatever it was, we were able to kind of show that those quote-unquote quitters, which are about 20% of the people who went through the experience, showed a lower level of head movement. And that's, I think, a meaningful way to look at that behavior.

[00:21:10.102] Kent Bye: Since you are starting to look at social dynamics and training people to essentially become better people in relation to other people in social situations, to me it's really difficult to put any numbers on that type of thing. It goes back to this philosophy of science, which is much more like transcendental idealism or phenomenology of like, what was your direct experience of this interaction? So asking people, again, there's all those other issues and challenges of having them sort of try to put words to something that was an experience. But I'm curious, how do you start to even wrap your mind around that process of like, I'm trying to train you to be a better person. How do I know if I'm succeeding or not?

[00:21:51.293] Michael Casale: It's a great question. I don't know that anybody's able to answer that right now. There's some great social science minds that have really dedicated themselves to understanding this issue. That said, it's really difficult. I think it's very context-dependent at this point. It's really difficult to get systematic measures that say these three measures will apply to any social situation. Like I said, it's pretty context-dependent. But we did things are starting to emerge so a lot of kind of the work in the nonverbal space So the kind of way that you react to someone based on how they react to you So averting your eye gaze when they try to make eye contact Kind of sitting in a kind of slouched posture then kind of an attentive forward engaged posture There's some systematic findings now that apply in a lot of different social situations that you can start to use. So, again, YVR is great. It's the ability to capture that in a high-fidelity way and in a way that is used with a real-world, reproducible, kind of translatable social experience. that allows us to capture this data and start to make assertions about, we know this isn't great. I would never go say like, hey, you should get this kind of training or you're a six and not a seven. Like, we're far from that. I don't know if we'll ever actually get there. But the fact of the matter is that's a level of insight that is objective that you can use as a guide. And I think that's what all this is right now. These are kind of guides. to be able to say, how is someone doing in training situation X, whether it's social behavior, whether it's identifying mistakes in the workplace, whatever the context is, and then say, hey, OK, this is where an area we know we could use more work on. And so it is relatively qualitative. But again, it's a level of insight we didn't know. Back to the example, when you have a bunch of learners in a room, you just assume that because they're in the room getting the same information, they're all learning at the same rate, they're all taking it in. they've all learned, right? But then you see the application of that learning in the real world and there's going to be a lot of variance. Understanding kind of who's better, who's not, where supplemental training can take place, that's huge and that's something that we don't really get from current training or even current diagnostics right now.

[00:23:54.225] Kent Bye: It sounds like with VR you're able to start to at least put people through a standardized training regimen. Maybe there's some AI and interactivity at some point in order to really dial into the different interaction sort of branches that you may be doing. But just in terms of evaluation, it seems like Is it possible for somebody to be kind of an expert at body language as well as kind of their own depth of presence of them being really present, having interactions with other people? And it's actually, from a phenomenological perspective, more about that expert's experience of this person and them being able to actually dial in and say, actually, this is what you're doing wrong. And it sort of gets into more of a human interaction at that point after they've gone through kind of like the standardized VR training.

[00:24:35.908] Michael Casale: That's a great point and I don't have a great answer for that. I would say though that when I think about using these non-verbals as kind of diagnostic features, I don't also think of them as like the things to train, right? Although that mapping is typically like, you know, one-to-one. I think what's more like we know that maybe this conversation isn't going well because you're kind of like not making eye contact, you're clearly like your body language when I'm talking is you know kind of pulling away or whatever the the phenomenon is. I don't think that subsequent training then is to say well fix those behaviors and you'll fix the conversation. Maybe, but I think it's more that okay we understand that you're not good at paying attention, you're not good at attentive listening. Ideally and what I think is going to happen is that body language is a result of you not being attentive. So being more attentive will necessarily just kind of engage the rest of the nonverbal. So by the nature of like the higher level social thing that you're getting better at or that you're correcting, the kind of nonverbal stuff will follow. So I look at those features as more diagnostic and less a thing to actually train and become expert on, I guess. But who knows? Like, you know, the whole world of embodied cognition says, well, maybe that's not true. Maybe If you're not smiling that much, then I tell you to smile more, you'll actually be happy, and that's more the causal mechanism. There's a lot of interesting research going on in that space, so I wouldn't claim to know what the necessary subsequent training would be. What we do know, from a diagnostic perspective, what makes a good conversation, what makes a bad conversation, just based on your nonverbal.

[00:26:05.894] Kent Bye: And are you looking at any artificial intelligence interactive type of scenarios? Because I imagine that once you get dialed down into these social situations, it really becomes more of a interactive experience that has like many more branches.

[00:26:17.978] Michael Casale: Yes, that's really hard. And you mentioned branching and that's about as AI as we're getting right now. We can accomplish a lot with this kind of intelligent branching right now. And again, these are kind of baby steps into this space, baby steps that are really helpful and create an awareness that wasn't there before. But there are going to be limitations, and there are for everyone, these limitations about the ability to interact successfully with kind of like a virtual avatar, for example, so you can have that meaningful, realistic conversation. But again, because we can control the branching a lot better, and we kind of constrain the nature of the conversation or even the social situation, we're going to be able to provide a lot more of a robust experience, albeit limited in the sense that it's not like a free form, like you can talk about anything with this person, but we don't want you to. We want you to interview them, or we want you to have this difficult conversation about firing Linda. So these kind of very specific examples. But obviously, there's a lot of really smart folks working on the AI problem in general. I know down at USC they've done a lot of interesting work at the ICT there with being able to provide virtual therapy where a therapist is able to kind of read your sentiments, read the words, and kind of use that information and provide some sort of intelligent real-time decision that's not like broken up. The decision tree kind of thing inherently kind of gets you in a little bit of a broken up dialogue, which isn't ideal but still useful. But the next level things are kind of emerging from those research labs. I think it's coming. I also know how difficult those problems are. But that's the holy grail.

[00:27:47.960] Kent Bye: Well, and just looking at some of the experiences at both the USCICT, but also University of Central Florida, they have these systems that are kind of like driven by the Wizard of Oz. So there's actually somebody behind the scenes listening and kind of driving the experience. And depending on what they're saying, since the natural language processing isn't really actually at the level perhaps in some of these scenarios, but at least there's like a standard branching that they can help guide and interact. Is that something that you're working on in terms of like having somebody having to drive the experience as a Wizard of Oz?

[00:28:17.300] Michael Casale: Yeah, we've explored that and we've certainly, you know, in our kind of first foray, it's on the table. We do find obviously severe limitations to the scale of that. That's the obvious one. There is a company, an interesting company called Mersion, and I don't know if they solve this problem, but they do basically a really sophisticated kind of wizard of housing where they're able to use a small bank of people in a kind of remote centralized way and then being able to produce a lot of the real time adaptive reactions using humans, but humans who can substitute for any given avatar. So you want to make, basically these people can wear the face of an infinite number of people depending on how you create the VR experience. There's a bank of them, I don't know how many, 20, 100, that's something I don't know. But they're able to then, again, reproduce hypothetically in infinite number of people. The problem is it's still kind of a one-to-one thing where you can't have one person who can be ten different avatars at once. You can have them systematically become ten or a hundred avatars. So I don't know if that's a solution. It's working for them, I think, to some extent, and it's an interesting Gambit, and so we're keeping an eye out. We're not wedded yet to any of these techniques at this point, but it's certainly something we're exploring. Obviously, if you can have something that's less adaptive and even foregoes the Wizard of Oz, but is able to be put in a corporate setting that the individuals at the corporation or the enterprise setting can use on their own as an adjunct to current training, and it's powerful enough to create that awareness and to create some of that kind of like better, effective way to train these social concepts, that might be enough for us. So then while we lose a lot of the value of maybe the Wizard of Oz, we gain in scale and the ability to deploy and touch a lot greater number of people than, say, what Mergin is able to do now. But we don't know. I mean, these are things that are a little bit TBD, but those are kind of the options in front of us.

[00:30:09.897] Kent Bye: Yeah, to me I've been really excited about this concept of embodied cognition because to me it kind of represents a big philosophical shift in terms of moving towards these more holistic ways of thinking about how we think and beyond just our brains but with our entire bodies and actually also the environment. I'm curious to hear from you, like, you know, how do you think of the concept of embodied cognition and how are you specifically applying it at Stryver?

[00:30:31.845] Michael Casale: Yeah, so we've just broached this. I've started working with some of our folks, especially in like these social settings. And the conversation we just had about, you know, is it really just changing the result behaviors that will then change the attitudes and the cognition? Or is it really changing the attitudes and cognition that will change the result behaviors? That's kind of the crux of embodied cognition is like actually changing the physical thing that you're doing will result in the change in the cognition that the causal mechanism The thing that you mentioned, the kind of unconscious behavior, that's happened. Our consciousness is actually then just kind of like coming up with a quick story about why we did something or how we did something. So like you forcing yourself to smile will then make you happy and then you'll maybe make up a story about why you're happy. But it was actually the physical act of you just smiling that actually created that. And that's kind of the crux of embodied cognition. I think there's a lot of power for that in VR, and we're not necessarily applying it in any of our things directly as an explicit kind of construct, but if you think about a learner sitting there passively listening to someone versus a learner then actively engaged, that physical movement that's being facilitated by the VR experience actually may lead to a better learning. Again, it's the physical thing that's changing the cognitive state that's at the heart of embodied cognition. And our ability to do that, or anybody's ability to do that, frankly, with VR is something that I think we're doing even implicitly. But making that more explicit part of our experiences is probably something we're going to explore in the near future.

[00:31:59.559] Kent Bye: I just wanted to follow up on that point around the subconscious sort of making decision in your mind sort of after the fact. I think a number of people have looked at that and said, oh, we don't have free will because we've already kind of made the decision. And so I'm curious to hear your perspective on that. I personally believe that we have desire and will and that we're able to actually exert our will. And maybe it's a matter of we have to cultivate new habits in our body through a holistic practice of repetition and that we're able to actually learn in that way. But in terms of learning and agency, What do you think about these people that look at some of that data and they come to the conclusion that we don't have free will?

[00:32:33.316] Michael Casale: Well, this is an interesting, very deep, and sometimes dangerous conversation. I don't know. I don't know is my short answer. But I think as a practitioner of neuroscience, which I was more deeply at a previous point in my life, You do kind of have this, at least, tacit assumption that if you think about the hardware that's controlling behavior, which is the central nervous system, and maybe even you think about all the peripheral nervous system stuff, again, kind of with that body cognition phenomenon, but just say that our nervous system is really the thing that's ultimately in control of our cognitive state. Well, those set of interactions physiologically are pretty well defined at this point. Just like a heart beats, there's neurochemical interactions that happen. Cells fire, cells respond, they get chemicals, there's action potentials. All these things happen in a certain pattern that produce the observable behavior. In the case of the heart, it's a heart pumping. In the case of the brain, it's a memory, it's a thought, it's an emotion. And so, hypothetically, if you can reduce back all of these cognitive states that are really hard to touch and really hard to feel to the physiology exclusively, I think that the strong implication there is, like, well, where does that leave free will? Or at least if there is free will, where does it exist? Maybe it's something outside of the physiology, which I think, you know, again, being a scientist, you have to be open to all these hypotheses, and at least if they're testable. And maybe we'll get to the point where the techniques and the engineering and the ability to detect kind of like these physiological states are sophisticated enough that we'll find out something else about how free will actually works. Because it certainly feels that way. And some people think it's a byproduct of conscious behavior. I think that's a valid argument. I think it needs to be tested. So I guess my short answer is I don't know, but it's something worth exploring.

[00:34:23.091] Kent Bye: Yeah, yeah, it's it's something that I have a visceral direct experience But I don't know if I'm being tricked and it's an illusion. And so I've been having these discussions with people and I I don't know either Yeah, it's a great conversation.

[00:34:32.873] Michael Casale: And obviously it's something I've been thinking about for a very long time. There's some great reading I don't know if you ever read like David Chalmers books or like even some of like Steven Pinker's books or Dawkins would kind of like start to Not directly broach these subjects, but certainly have implications for that matter

[00:34:49.301] Kent Bye: For Stryver, what are some of the biggest open questions that are driving your work forward?

[00:34:54.735] Michael Casale: I think just kind of getting like a strong connection between what we're doing on the training front and how that's affecting real world behavior. If you think about the ultimate behaviors we think we're trying to affect, Stryver is a part of that, but it's a part of it, right? It's not the whole entire thing. So when you look at someone's ultimate behavior that you're hoping to change, whether it's becoming a better person that interacts with people socially, whether it's you're better at doing your job, if it's like you're faster at loading the FedEx or whatever the application is, that's affected by a lot of things, right? So striver training is going to help, but if you're like a FedEx worker and you're loading boxes and you happen to be really tired or you're going through a divorce, like those are things that are going to affect your job performance. Ultimately, quarterback, it's the same thing. Your offensive line decided not to block that day. Well, striver's not helping with that. Striver's just working on your kind of like, kind of cognitive components of your performance. And so it's really being able to ferret out how striver performance does affect that ultimate, kind of like, real-world performance that everybody cares about. While I think there's some fundamental limitations in being able to do that, as I just articulated, because it's made up of so many other things that happen in that person's life. There are some really cool things we can start to do to assess real-world performance just within VR. Again, if you buy this assumption, and I think that's kind of a lot of what our work is predicated on, that we're providing you with that real-world environment, and we can control that environment, to your point about kind of the research potential of VR, We should be able to know how you're going to perform if everything else is kind of like well and situated and controlled for. VR should be a good way to analyze how our training is affecting performance. So not only providing training in VR, but also assessing them in these real world naturalistic ways. And again, something you couldn't do before, but always better is the real world thing. So we hope to get there, but we also know how difficult that can be.

[00:36:46.623] Kent Bye: Great. And finally, what do you think is kind of the ultimate potential of virtual reality and what it might be able to enable?

[00:36:54.189] Michael Casale: Well, yeah. I mean, I saw a pretty cool thing yesterday. I don't know if you saw the Neurable guys, but that's the path, right, is to be able to basically interact seamlessly in a virtual environment as you would in a real-world environment, but you get to control the virtual environment in any way you want. to be able to kind of have a situation where everything in the real world and more is possible in the virtual world. And, you know, again, you can think of some really sinister applications for that. You can think of some really benign applications for that. So, you know, at Striver specifically, I think we're generating so much interesting data about how we're changing behavior just with our small foray into the space. how VR can really affect that, and what are the parameters that you should be building VR to really effectively reproduce that change in behavior from one application to the next. But yeah, for you ultimately, I don't think it's gonna be like the headset, but what we're calling virtual realities is the idea that you can perfectly simulate this world, and that you can start to manipulate it in ways that are hopefully desirable, at least for most of the people in society, not just the hands of a few. Awesome. Well, thank you so much. Yeah.

[00:38:06.088] Kent Bye: Thank you. So that was dr. Michael Casale. He's the chief science officer of Stryver So I have a number of different takeaways about this interview is that first of all the overwhelming takeaway that I have from this is just that VR is this ultimate training platform that is able to allow you to enter into the context and be able to tie together all the different systems in a way that actually is it's better aligned to how we actually learn. And I think the principles of embodied cognition is kind of inverting the causal loop that we tend to think about when we think about how we actually learn about things. Usually we just think about, okay, we're going to get the concept, we're going to learn it, we're going to hear about it, and then we're going to actually do it. But actually that may be switched in the way that you may actually have to be performing and doing the actions. And once you do the actions, then it's that process of doing that actions that you really start to put it into your body and you actually learn it. It's that metaphor of, if you're smiling, then you may not be happy, but then it may be the physical act of smiling that causes you to be happy. The way that Michael phrased it is, is changing the resultant behaviors the thing that actually changes the attitudes and the cognition, or is it changing attitude and cognition that changes the resultant behaviors? you're kind of starting with the result behaviors. And then what emerges is that your attitudes and cognition is changing. And I think that the takeaway here is that the body is a holistic system and you are able to put somebody into an environment where you're able to make choices and take action, have all of their emotions engaged. And the more depth of experience that you're able to create within a virtual reality experience, then the more likely is that whatever you're teaching them is going to have higher retention. They're just going to have it in their body. And it's going to be deeper memories because they're just going to have the experience. The problem, of course, is being able to actually measure and prove that. And I think that that's where a lot of those open questions are that I think Michael is on the leading edge of trying to figure a lot of that stuff out. It was interesting to hear that in terms of engagement is one of the biggest predictors in terms of how you're going to actually learn something. And rather than doing a survey, because our subjectivity is very unreliable, we don't really know why we're doing things, we have all sorts of habits and behaviors and emotions and our perceptual systems, all these things are below our conscious level of awareness that we're kind of like these zombies, in some sense, we're going around and we're doing things that we don't actually have a logical or rational reason for, we're just kind of almost acting out of habit. And so taking surveys after the fact and trying to have people describe their direct experience has been traditionally pretty unreliable in terms of, you know, translating that qualitative phenomenological direct experience into numbers that can generalize out into some sort of objective data. And I think that's where neurophenomenology is trying to, in some ways, close that gap. I think it's a big open question of how can you actually validate someone's direct experience in the moment and also have these third-party objective data that are correlating to that and being able to actually have many different indicators. And that was the thing that I heard from Michael, is that in terms of trying to determine a causal loop, then he's looking at reliability, validity, and an association with specific behaviors, but there's always ways to pick it apart, and you actually need many different indicators to be able to see what is actually happening. So one of the most predictive indicators that they found was whether or not someone stops and says, you know, I'm done. And that's like a binary decision that they're saying, OK, well, at some point, they kind of checked out. They weren't engaged. And then they're able, from that point, look at all the other behaviors, like what is the volume of degree to which that they're actually looking around and participating within the experience? And so then they're able to differentiate between the people who quit and people who didn't quit and to be able to determine some of those other higher order ways of seeing, OK, what are the types of behavior that are really going to predict whether or not this is super effective or not? But at the end of the day, it's still the problem of when you're dealing with human behavior, how do you actually measure how you are improving? And this is largely an open question, as far as I can tell from this discussion from Michael. And that's what a lot of the social research scientists like Jeremy Bailenson at the Virtual Human Research Lab at Stanford University, he's in the process of using virtual reality to be able to control all these different social dimensions and be able to see how people behave in these virtual environments. And then perhaps do a level of social science research that is able to increase the amount of rigor and insight that we have about the dynamics of human behavior. in that evaluating how people are doing within different social situations are extremely context dependent. But they're able to come up with some things like either body language as a measure of social success, or someone's averting eye gaze, or if they're slouched, or if they're attentive and leading forward, all these sort of subtle cues that you get in terms of how people are actually performing in these environments, then they're starting to come up with some of those higher order ways of being able to teach and judge different social skills. So I have a number of different interviews that I've done over the last couple of months about training and training in the enterprise that I'll be diving in as well. But I think, you know, this interview with Michael is kind of really diving deep into the science and the neuroscience of everything. And I think there's a lot of progress that's being made, but also still a lot of open questions for how all this is going to play out. But I see that VR training is going to be one of those areas where there's going to be a huge uptake in terms of VR and immersive technologies. And I think specifically the standalone VR headsets I think are going to have a specific use case when it comes to training an enterprise. I'm going to talk to different people about some of the challenges of using VR in the enterprise, one of them is deployment of the Gear VR or the Daydream being three to off and not quite as immersive as they need to be able to do some of these training situations with six degree of freedom, positional tracking of the head, as well as the hand track controllers, and that the high end PCs and all the other wires and logistics of that tend to also be a huge barrier as well, depending on the different industries. But to be able to mass deploy some of these standalone VR headsets, I think a lot of people in the enterprise market are excited to see what's possible when it comes to training, when it becomes a little bit easier to distribute some of this technology. So finally, the last thought, just the idea about free will. It's an open question. It's a philosophical question. It's a deep question that I think that nobody really has a clear answer to. One person that I really enjoyed talking about this is Seth Lloyd, who has this article and video on YouTube about a Turing test for free will. And he's making the argument that free will is kind of like the halting problem. the halting problem being this, you know, kind of theoretical computer science problem saying that, like, as any computer program, if you ask it whether or not a program as it runs, is it going to halt or not? And if you kind of self-referentially feed that program within itself, then you have this sort of paradox where it can't even actually answer that question as to whether or not it's going to have enough computer resources or if it's going to be able to halt or not. And so free will could be modeled in terms of a halting problem. just not knowing and be able to predict what behavior is going to happen. Our mind is kind of like that in that there's so much of our actions that are coming out of behaviors. But I think at the same time, what Dr. Khasali was saying is that consciousness is kind of like the wildcard here in terms of like, some people are theorizing that maybe that free will is contained within that phenomenological or qualia of a direct experience of being able to exert will, even though they're not able to find any physical correlates to that action. Maybe it's coming from some other sort of non-local conscious field. And that, at this point, doesn't have any testable data to be able to falsify that theory. So at this point, it's still kind of an open philosophical question. So that's all that I have for today. I just wanted to thank you for listening to the Voices of VR podcast. And if you'd like to help out this podcast, there's a few things that you can do. First of all, just spread the word. Send out this podcast on social media, send it to people. If there's anyone that you think will enjoy this episode, send it to them. And also consider becoming a member to the Patreon. I rely upon your gracious donations to continue to bring you this type of coverage. So please do become a member today at patreon.com slash voices of VR. Thanks for listening.

More from this show