← Harvard Data Science Review Podcast
Harvard Data Science Review Podcast
Digital Twins and Virtual Twins: What Are They and What Do They Do for Humans?
00:38:17
Loading summary
Transcript28 lines
- [00:01]
Liberty Vittert Capito
Hello and welcome to the Harvard Data Science Review Podcast. I'm Liberty Vittert Capito, the feature editor of the Harvard Data Science Review, and I am joined by our editor in chief, Shali Ng. Today's episode dives into one of the most powerful ideas shaping how we understand product development, manufacturing systems and society. Digital twins. From aviation to city design, to transforming how organizations build and operate complex systems, our health system digital twins are redefining what's possible when data models and reality converge. We're joined by two exceptional guests who sit at the intersection of data, geography and advanced simulation. Rachel Franklin, executive director of the center for Geographic Analysis, brings deep expertise in space, spatial data, urban analytics, and how place based data can be used to model complex human environmental systems. And Patrick Johnson, head of corporate research and sciences at Dassault Systems, leads cutting edge research on virtual twins, computational science, and the future of large scale simulation across industries. Together, we'll explore what digital twins really are, how data science powers them, and why they matter, from cities and infrastructure to sustainability and decision making at global scale. All right, so Rachel, I'm going to start with you to lay our groundwork here. What is a digital twin? What does that even mean?
- [01:42]
Rachel Franklin
Okay, that's a big question. So I'll see what I can do and then maybe Patrick can jump in if he thinks I'm getting it wrong, because I'm definitely coming at this from a social scientist perspective. And digital twins, they're used everywhere. They're used in health, they're used in engineering, manufacturing. They're definitely used in like city systems where I come from. And so what I'm more familiar with would be almost the exact metaphor of a twin, right? An exact replica of a system that we work with. So if you take a jet engine, for example, or a heart, you could imagine having an artificial heart that does all of the things a heart does, but is not actually a heart. Right? And in that case, it becomes super useful because you can try new things, right? You can operate, you can try medication, you can poke at the heart, but you're not actually hurting a real human heart in the case of a social system. So working in cities, we think about digital twins of cities as being replicas of these apparatuses that put cities together, right? And so this is things like stoplights, roads, movement of people, electricity grids. And in these cases, what we need in order for it to be an actual digital twin would be something that approaches as closely as possible the actual city system. So all of the pieces, all of the perspectives that you need for a city, which requires models, but it also requires lots and lots of data. And then the additional ingredient that we would expect to have on top would be something that's approaching live data, real time data. So it's not that you take all of your data in a cross section and then you look at all of the relationships you're dealing with, these sort of constant feeds of information that are coming at you from all of these different systems. And when that happens, then you start to talking about the need for big compute. Right? Because you can't hold an entire city in memory and watch it work and then also poke at it and change things, like close a road and see what happens. Imagine a hurricane hitting and seeing what happens. That requires an awful lot of computing infrastructure. But from a social science perspective, from a city perspective, it's that real time data and it's the systems approach that I think makes us consider something a digital twin.
- [03:56]
Liberty Vittert Capito
Thank you, Richard. Then I'm going to ask Patrick, because you and I, we had this conversation about the difference between digital twin and the virtual twin. And I remember you explained to me, I thought it was very refreshing. So I want to invite you to talk about your view of digital twin and the virtual twin.
- [04:17]
Patrick Johnson
So of course, the first approach of digital twins is really about what digitalization is about, which is a replica. The little difference that we are making in my company, and we prefer the more virtual, is because we don't want to do just a digital photograph or a digital, let's say, a copy of something that is coming from reality. We want to use the power of imagination and the power of possible worlds and want to explore them. This is the semantics that we put behind virtual as opposed to digital. When we have a digital heart of a patient, it's supposed to be the exact replica of the heart. When we have a virtual heart, it can embody not only patient A or patient B, but it can also encompass all kind of pathologies. And we can explore in fact, the realm of all the possibles. And so we basically see two ways, two usages of twins. First one is really the replica, but the second one is really an engine of imagination in order to invent new things that do not already exist in reality. So there's like a virtual circle between going from reality to the virtual, which is the digitization, but then going from the virtual to the potential real, which is here a cornucopia of ideas and imagination. I would say our definition of, of virtual twin, first and foremost, they embody a specific representation of the World. And what goes with representation is always a set of questions we're trying to answer. So there's always some sort of, not a bias, but a view of the world, an observation of the world that we're trying to tackle, or a set of questions we're trying to answer. And to be able to do so, those twins are to be virtual representation in constant relationship with reality, with real time data. We call them VIP R twin, the virtual plus real twin. The twin is a fused representation, let's say an abstract model plus a real sets of data in a continuous discussion, in a continuous dialogue between the two. And we apply that for products, we apply that for organization at large, and we even apply that to business models. The tariffs, for example, has led to virtual twin offerings in our portfolio for supply chains, for all stakeholders and industries to be able to understand exactly how the economical model will evolve with those new conditions and those new economical geopolitical settings. So it's really a representation of the world and a constant observation of that world, feeding the representation with data.
- [07:19]
Liberty Vittert Capito
So I was teaching a class last night, a data analytics class, but to professionals. And one of the questions I asked them was, how many of you guys are using AI in your work? And I don't know, probably about four out of the 20 of them said that they were using it. And all four of them said it was a disaster at work. You know, it wasn't working, it was a mixed bag. They've been told from up above to use it. The only way they really use it, that it's useful is writing some emails that they probably could just do themselves just as fast. So, you know, I think that brings me to the same question. When it comes to digital twins, you know, there's this sort of really glamorous concept that it's going to be used for healthcare, for cities, for countries, for entire populations. But what domains do you see this truly initially benefiting? Not in a way that's, you know, hand wavy and sounds magical, but that we can actually do real change in. Now you want to go first Rachel and then Patrick?
- [08:26]
Rachel Franklin
Sure. I mean, I've got opinions on this, so maybe we can, maybe we can unpeel it like an onion. Because I think the analogy is really good between AI with sort of scare quotes, quotation marks, or digital twins in that you open up that box and it's actually a whole continuum of objects. Right? It's a whole bunch of things. And so with AI, like the question that you asked your students, most people are thinking about generative AI Right. They're not thinking about all the other kinds of artificial intelligence that exist in my area. When we talk about digital twins, one of the most common conversations that we have is like, is that actually a digital twin, or is that just a really kind of complicated model? And sometimes it is, you know, old wine, new bottles, sometimes it's the branding that's necessary in order to either sell something or get your research funded or to persuade people that what you're doing is really important. I think where I land at the moment with digital twins is that on the academic side, there's just a lot of hype, but I think that the hype is unavoidable because that's the way science works. Right? There will be something after digital twins, and there was something before digital twins. Before digital twins in cities, we had smart cities. And actually smart cities wasn't really all that different from what we're now calling digital twins. So you could point to that and say, well, this is a problem. Clearly, if it's hype, there's nothing of value. I think it is the nature of how we work with anything analytical or technical. But I also think that it's the journey and not the destination. We recognize a little bit where we would like to be. Cities that run more efficiently, cities that are more inclusive, cities that are more productive, cities that are healthier. We can see the potential in terms of streams of data, in terms of the systems of systems. It's not perfect, but the ideal sort of the way Patrick was talking about, like the, you know, the sort of the virtual as an ideal, I see that as a destination that we're heading towards. And it's good to have goals. Right. It tells us when our computing infrastructure isn't actually fit for purpose at the moment. It tells us where we're missing key data about certain pieces of the system. It tells us, especially in anything social, like what is it about humans that we're not actually capturing preferences and behavior, so it actually highlights where the gaps are. And then that shows us where, from a researcher standpoint, where we may want to invest more in terms of learning how to manage these things. So I do tend to roll my eyes a little bit when something gets badged on, like AI. Definitely I feel that way, and I feel that way about digital twins, but I think we should have a moment of grace for ourselves and accept that that's just the way things are. And actually people are trying. And the challenge is an interesting and fun and worthwhile one.
- [11:21]
Patrick Johnson
I think a good way to complement what Rachel just said is just to look at the industrial footprint. Previously in the industry people were talking about digital mockups. And in fact, in many industries this is standard of practice. So let me give you concrete example. Boeing did in 94, the first ever airplane completely designed, completely modeled, completely simulated, end to end for the 4 million parts in a totally virtual fashion. Before anything was produced in a factory, it was the 777 aircraft. And so that was the first aircraft that was produced and designed totally virtually. At the time, everybody was calling that a digital mockup. Today Boeing is calling that a virtual twin. The benefit of that is manifold because they were doing that in a virtual world first. There was almost no errors in the parts definition and the engineering work, which means that the quality of the first 777 was better than the hundredth 767 at production time. So the virtual world, or let's say the digital mock up or the virtual twin is used to anticipate all the problems downstream and to ensure that the definition will be exactly as expected or as specified. And this is why aerospace has been almost the first industry to really, you know, endorse the digital mock up, virtual twin approach. Today there isn't a plane that is invented without its virtual replica or virtual representation. From the supply chain perspective, the 777, Boeing didn't 70% of the plane and outsourced and with suppliers, 30%. With a 787, the ratio is exactly opposite. So how do you control a supply chain that's worldwide with 200 suppliers and make sure that you control time, cost, IP, I mean, everything that is related to that very complex ecosystem, Guess what? With a virtual twin, and in that twin, it's not only the representation of the plane itself, it's the representation of all the supply chains and processes. People look at twins like it's a virtual photograph of an object. It's a management system. And so the real benefit, the real footprint, in fact, it's management. And this is why it's being endorsed in other industries more complex from the process standpoint, like pharma and ultimately in the everyday life, like cities, utilities and so forth.
- [14:37]
Rachel Franklin
Could I ask a quick question? Just because that perspective on digital twins, it's fascinating, but it's also super different from my standpoint. What are the unsolved challenges? Is this a developed approach or technology in manufacturing or are there still problems that they're trying to solve? Because on the city side. Okay, all right.
- [14:56]
Patrick Johnson
There are many. First, very concrete. And remember, I start always by saying it's a representation choice. So you put in a twin what you decide to put there. And therefore the other way around, what you don't put into twins translate into non answered questions. So let me give you a concrete example again, Aerospace. When Boeing changed the production methodologies for the 787, they switched to composites. At the time, material science was not a domain sufficiently well understood to be able to be virtualized and twinified, if I may use that expression. And so when Boeing raised a bar for manufacturing processes with composites, the ultimate consequence was twin with shape are not enough anymore. Geometry is nice, but I need a better understanding of the constituent of the material themselves, of the production processes of the material themselves, of the chemistry. So the representation suddenly was totally insufficient and needed complement representation more from the again chemistry and materials standpoint. And I could add billions of different facets if you will. So there is the exhaustivity or the completeness of the representation. That is the first issue. It's a never ending story. Basically. Rachel, you mentioned it just before with the behavioral observation, for example, from the social and human sciences. How do you model that? How do you inject that in a representation of a twin? We work for example, with consumer packaged goods companies and we know that it's not only the formulation of a product, of a shampoo or the packaging. It's also what they call the first moment of truth. How you react the first time you see the product. The second moment of truth, how you come back to that product, if that means anything to you. And those are just examples that we need to be able to model for health care. Quality of life, social determinants of health are also topics that need to be injected into twins. So it's a never ending journey because the representation need to be as sophisticated as you go. So the multi simulation, the multi discipline, that's a high challenge for engineering. How do you simulate something with all behavioral that you observe in reality in an extremely accurate and conformed fashion in a virtual setting? And so it's again a multi scale and a multidiscipline challenges.
- [17:54]
Liberty Vittert Capito
On one hand, that's not that different from any other kind of a scientific modeling we do, right? We're always trying to find a representation. There's all the complexity. But this has an added both the challenges as well as its power. So Patrick, for you, I want to ask you about the biggest challenge in terms of building a virtual twin or digital twin. Building a realistic model is going to be a grand challenge. I assume the data collection probably is another issue. Their computing Powers, there may be even team management. What are the things like as someone in charge of all those things from industry perspective, what as people say, you lose sleep on during the night.
- [18:38]
Patrick Johnson
So I'd say what keeps me awake at night when it comes to twins is how are we going to come to the next level, which is a combined approach with generative AI and explicit representation. Generative AI is black box, so that's implicit representation. We don't know if it's a correct representation. We know it sometimes works. But if you are working on mission critical industries or regulated industries, you cannot rely on black boxes. So we are trying to revisit the foundation of AI. In fact, we're trying to do internal twins within generative AI so that in fact even generative AI is relying on unexplicit representation of a certain domain. And this is very, very new in the field. Nobody is really doing that. The second thing, we are in my company, so this is really related to my company. We are working a lot in healthcare. We believe that, you know, in the 19th century or before, the standard of representation was the drawing, the ecorche, as we, you know, French expression. The 20th century was the scan and the MRI. And the next generation of standard representation for health condition is going to become the twins. And so what keeps me at night is not so much the technical challenges, because we're getting there already in many, many settings, but it's the adoption by practitioners and physicians. How are we going to morph the current practice? How are we going to train the next workforce of the future? And it's not to oppose with the previous approaches, it's to complement them so that practitioner don't need to learn anything. The twins need to learn how practitioners do their work and live their daily work. So how do we make them as consumable and as relevant for those audiences? Those are the two things that keeps me awake.
- [20:53]
Liberty Vittert Capito
That's a fascinating angle. Yeah. I'm going to ask Rachel kind of similar questions, but from a social science perspective. Right. Obviously, I guess to build the system. You already mentioned the data and all this thing would be important. But I'm also curious, really from a user, like a social scientist, how these data, I call it data, but let's call it, you know, whatever this system is, how they be used for drawing important scientific conclusions or policy makings, how to be used and what is the process, just like Patrick has said, in terms of convince people to use them, or maybe you don't even convince them, you want to prevent them to abuse it. So what does this whole process looks like from your perspective?
- [21:37]
Rachel Franklin
Well, one of the first things I would say is actually to pick up on a comment you made, which is that maybe listeners are thinking, how do I do this? Like, I would like to do this in my research. And actually for any of us, regardless of the application. A really interesting thing about digital twins is that this is not a one person job, right? There is no such. I mean, maybe there's a unicorn out there who can do all of the things that are necessary to actually build, run, test a digital twin, but I don't think so, because, you know, if you take a city, for example, I think it's fair to say that when we talk about urban digital twins, much of the enthusiasm has come from civil engineers, some people in urban analytics, who you could classify as social scientists, but really falls slightly more on the engineering side. And this is important because when they think about a city, when they think about the skeleton of a city, they're thinking about street networks, utility grids, buildings, all the hard infrastructure that you see when you look out your window. And then I think, and I'm generalizing a little bit, but when they imagine sort of humans, humans are agents that you can move through the city. And we assign some behaviors to these agents, right? We can assign ages and behavioral patterns so that we know that, for example, if a metro station closes, we can imagine how the people will disperse and where else they'll go and which roads might be more busy. If there's a flood, we can imagine basically what's going to happen on the streets and what might happen to surrounding roads. You can imagine the policy potential, right? You think about congestion charging, you think about extreme weather, you think about urban development, urban growth, you think about economic development, right? Where should you build? Where should you build a city? If a city is expanding on what side of the city and what kinds of infrastructure should you emphasize in order to achieve the goal that you want in the end, which is probably minimizing pollution, easing friction of distance so that workers can do the thing that they're meant to do, which is to make money for someone, right? So you can start to see all the potential for an urban system. And then I think it starts to fall apart because the humans, they're not agents, right? We can put sensors on everything. We can sense the air quality, we can sense the footfall, we can sense the traffic. We are really, really good at monitoring so much at a very granular level what is happening in a city. But we can't put sensors on the people, we think we can. We think if we follow their phones and we know what they're doing with their phones, that we've like, that's the twin, right? Like, this is my digital twin. I'm holding my phone up. Then that is obviously not the case. And it's not even like a humanistic or qualitative argument that I'm making. It's a very data driven argument, which is that my preferences are very hard to identify and they're going to change over time. And a really great example would be thinking about urban growth. How do we think cities become bigger in terms of people? They can only become bigger in two ways. They can become bigger because people want to move to them, or they can become bigger because the people who live there have babies. Both of those things are really, really difficult to predict. Like, we know right now, for all of us that are on this call, something probably about our revealed preference when it comes to fertility. How many kids did we want to have or end up having? That does not tell us what people are going to do in 20 years. And to know what people in 20 years are going to want to do, we have to actually wait the 20 years.
- [25:08]
Liberty Vittert Capito
You know, Patrick, I want to go back to something you said about the technology doesn't keep you up at night. It was the adoption by the practitioners. The thing that, you know, sort of keeps me up at night is the privacy and the ethics around all of this stuff. And how do we really guard against bias? How do we guard against surveillance in terms of the ethics? We can get really high level, you know, on this stuff of like, well, you know, we're gonna protect the population, we're gonna protect whoever. But that for people's everyday use, you know, which is the sort of incredible stuff that you're working on. I'll give you an example. I bought a scale, I'm trying to lose my baby weight, so I bought a scale at cvs. So I unpackaged a scale, I put the batteries in and I went to go stand on it actually this morning, literally across the screen, instead of saying my weight, it said use app. It literally would not let me get on a scale and just tell me the number. It wanted me to go onto CVS and use their app to record my weight, to be able to see what I weighed. It was sort of a mind boggling moment to me. This is like the, the highest level of invasion of privacy I can possibly imagine.
- [26:25]
Patrick Johnson
I agree.
- [26:25]
Liberty Vittert Capito
So, you know, everything from sort of the very large sense that I want to understand from you of ethics and privacy down to like, you know, me standing on the scale naked this morning, wondering how much more baby weight I have to lose.
- [26:39]
Patrick Johnson
Of course. So let me give you very concrete answers. In my company, we were very, very involved with the COVID waves. Why so? Because as we expanded into life sciences, it so happened that we basically provided the platform to test the COVID vaccines. Those vaccines were clinically tested on our platform. And why do I mention that is because in the clinical trial processes today around the world, it all starts with a patient consent. And it's not just a form that you sign. It's a process that is being monitored every time a tiny piece of your data is either first used or used a second time, or used a trillion times. So this, this culture, if you will, of traceability and let's say sanity check, where you still agree to give your consent. And the consent is not just a sign off, it's always for specific usage, for a specific purpose in a specific setting at a specific time. So it's extremely well defined. But in fact, what we've been doing, we've been expanding that concept and that approach to any personal and actually even industrial data there is. So for example, when Boeing is working with suppliers and Boeing does the fuselage and the supplier does the wing, at some point in time they need to connect. But maybe there is some very sensitive IP of Boeing related to the fuselage that the partner doing the wing is forbidden to see. And so we've been putting in place all those guardrails. The term is privacy enhancing technologies for personal health data for personal non health data for companies, IP data for networks and supply chains, stakeholders data and ip. And so this is a very specific value proposal that we deliver. This is called in our jargon industrial or intellectual property lifecycle management. And so it's not only the data that you look for, let's say at one day, it's the life of that data. And every people using it or leveraging it, or creating something out of it, we trace that. We always check the consent, the agreement and the outcome related to that data. But this is really something extremely important to us. Why? So everybody is saying that data is the oil of the 21st century. But more than that, data is a part that could define us. And we all know that by combining metadata, even pseudonymization or anonymization becomes very challengeable. So because we know that and because our market is really the industrial market, so you know, it's all about reputation and trust, the ultimate value that we deliver Are all those guardrails privacy enhancing techniques, intellectual property traceability and trust management. And so we do that for all the data. And to get back to your example with the scale and stuff, of course, a lot of players in the digital domain are trying to have you lock in on their system. Here you cannot use the scale without using the app. As you know, in Europe, rgpd, among the RGPD principles, you have the right to leave, you have the right to transfer your data, and nobody has the right to oppose. Nobody.
- [30:45]
Rachel Franklin
I am not sure that I think that health is the best benchmark to be using because we've had now, you know, a good century of thinking about what consent means and how, how medicine and health can be misused. So we have lots of guardrails in place. And I think crucially, when it comes to medicine and health, there's a recognition that you need population level data often to be able to have interventions. And it might be because I'm a geographer, I could have led with this earlier, but I think where we should be thinking is about all of the other arenas in our lives and what is happening in terms of our privacy. And some of it is lock in with having to use apps, as Liberty said, but some of it is recognizing that, for example, every time you use a credit card, they know exactly how much you spend. And from that they can estimate how much you make, but also they know what you're spending your money on. And the thing that I think we don't think about very often is they know exactly where you're doing this. And I think probably at some level, most of us are comfortable with the fact that this is what it means. I have a credit card and that is the job of the bank to have those data. How do we feel about banks packaging those data and selling them? I never gave consent to that. And yet that is exactly what is happening. Every time I move somewhere with my phone, every time I open an app, what the apps would really like is to not only have me giving them information about myself, how much I weigh, what music I'm listening to. Today was Spotify Wrapped, where everyone gets really excited about how much information they've given to Spotify over the course of a year about what they listen to. Right? But we're so pleased that they serve it back to us all wrapped up, when in reality we've just handed a big piece of our behavior over to a company.
- [32:34]
Patrick Johnson
The product is you.
- [32:35]
Rachel Franklin
Yeah, the product is us. And they're making money off of it. And so I Think the thing about apps and phones that we're not thinking about very often is that not only are we, are they making money off of our individual relationship with that company, but the data are being packaged across platforms and then sold on for someone else. And often there's a very high locational component to this. It's not that I think that we should be concerned. What I think is that the value proposition is not the same as it is for medicine. And this gets us back to digital twins, especially for cities or social systems. There's an opportunity cost in investing all of our time and money in this particular paradigm or framework. And there is a point, if you talk about well being or happiness in a city where you might ask, could we not take those millions and millions and millions of dollars and spend them in a different way more quickly? That would make people happier, right? So, you know, people talk about food insecurity as a wicked challenge, for example, and there are people who would argue that digital twins would make it better to think about food production and food access, but you could also just take that money and give it to people to buy food. Right. And so I think a risk when we talk about digital twins and when we talk about data and when we talk about privacy, is this is a very big edifice and is this where we would like to be devoting our attention? And maybe it is. I think in a lot of cases it clearly is. Like I'm very, very happy to fly on a Boeing airplane that is, you know, the result of a twin in terms of city systems or other, any systems that involves human behavior and actually where the outcome is humans, like humans are going to be happier or more productive. Then I start to get a little bit more skeptical because I'm not sure that those trade offs are worth it. Is it worth somebody knowing how much I weigh and where I bought my coffee half an hour later, like whether or not, like what I'm eating, what I'm buying in my groceries. You start to tie all of these things together and then I think you should be a little bit more nervous because is that the kind of world we want to live in, that someone's making money off of it and that they know this about us and that there might be a better way to achieve the outcomes. And just one final point is that all of this is privately held data. If you leave out the phones, if you leave out the humans, and the phones matter for the humans because it's, I think, largely how we're trying to get at individual level behaviors within cities, if you leave that out, all the other data are largely public. You know, like the sensor data for all of these large cities. I can go pull the data for air quality sensors for any large city. Like it's free data. Right. If I look at what Paris has available on its open data platform, I have the right, even not living in France, to use a lot of these data. As soon as we start talking about the humans, a lot of that data's moved behind industry walls. And so I think a question that we need to ask ourselves is, do we really want these data to be owned by someone where a faucet can be turned on and off? That the kind of water coming out of the faucet is controlled by someone else? How much water is coming out of the faucet is controlled by someone else. And if the faucet can be turned entirely off, is also owned by someone else. That's where I think when we talk about data, where I start to get a little bit more nervous. And I wish we talked about it more and I wish we had more informed consent.
- [35:48]
Liberty Vittert Capito
Now. We're already running out of time, but we still need to do one more thing, which is we always end by a magic wand question. So we don't want to break that tradition. So be very quick, start with Rachel. Let me ask the magic wand question that if you could have waved your magic wand and what would be the one thing you would have picked up in terms of bring the virtual or the digital twin to the next level? What will be the one thing you say I want to happen?
- [36:18]
Rachel Franklin
Public buy in. And by that I mean actually like public as in government buy in. Government data, government systems. The people should own these things. Yeah, that's it.
- [36:28]
Liberty Vittert Capito
Great. Patrick, what's your answer?
- [36:31]
Patrick Johnson
So it's going to be a very personal topic. We are using twins for elderly to make sure that we adapt, we do home improvement so that they can stay put where they want. And twins have the abilities to help them, you know, have a better life where they want to live and using the power of the virtual world to relate with their relatives. And so that's my magic wand to have twins more used for elderly for next generation homes.
- [37:07]
Liberty Vittert Capito
Well, thank you. And I guess, you know, my wish will be having what do you call digital twin, virtual twin of me that can do many things so I can go fishing, drinking wine and talking to you guys instead of jumping from one zoom to another. Thank you to both of you, Richard and Patrick. This is a fascinating conversation. We definitely will do another sequel especially when the special issue on Digital Twin comes out, because clearly there's so many questions. Thank you again.
- [37:37]
Patrick Johnson
Thank you so much.
- [37:38]
Rachel Franklin
Thank you.
- [37:40]
Liberty Vittert Capito
Thank you for listening to this month's episode of the Harvard Data Science Review Podcast. To stay updated with all things HDSR, you can visit our website at HDSR, MITPress, MIT.edu, or follow us on Twitter and Instagram hedsr. A special thanks to our executive producer Rebecca McLeod and producer Till Tina, Toby Mack and Aaron Keeswetter. If you liked this episode, please leave us a review on Spotify, Apple, or wherever you get your podcasts. This has been the Harvard Data Science Review. Everything Data Science and Data Science for Everyone.