 live now. And let me, the same thing that the attendees are probably very used to hearing me say, let me wait for these settings to finish loading. And then let me come back over here and make sure that everyone is marked into this session and wait for just a second for everyone to get moved over and logged in. And that's probably enough delay. So with that, all right, welcome back everyone again. Just in case, just in case you missed this at the end of the last talk, there is the zoom link there for you to head over to the exhibit visit and social hour after this, that that that zoom link will not open until seven o'clock Central Europe to the end of the current hour that you're on right now. So so you click it, but it won't do anything. But there it is for you here after this after this talk, and we will get a get a great, great tour of the places and spaces exhibit. But that's for after this before that, we have one more very exciting talk in our in our lineup today. Very pleased to introduce Dario Rodiguero from the Harvard Metal Lab, who is here to talk to us about some fantastic visualization and present presentation work that this is just a really neat project, supervision, celebrating the 50th anniversary of the Ecole Polytechnique Fédérale de Lausanne, if I'm remembering my acronym correctly, through the history of 8000 doctoral theses. So this looks this looks looks fantastic. And without without me further ado, please take it away. Okay, thank you, Charles, for the invitation and the very nice introduction. Good evening, everyone. My name is Dario Rodiguero, and I am a postdoc at Harvard University. I am a member of the metal lab and knowledge design unit at the intersection of arts and humanities and affiliated to the Bergman Klein Center for Internet and Society, whose mission is to explore and understand the cyberspace through its development dynamics, norms, and standards. My research is founded by the Swiss National Science Foundation with two consecutive fellowships, one at the MIT and the current one at Harvard. I have a bachelor's degree in computer science and master's degree in psychology. My doctoral studies are about architecture, sociology, and digital humanities. And this allow me to interact with specialists from different disciplines in high interdisciplinary environment. Today I present a work of science mapping that is titled supervision. Supervision is a word play. Its meaning is two fold. From one hand, it is about the supervision of doctoral students. From the other hand is about distant reading, which is the capacity of seeing a phenomena from a distant perspective. Supervision is a data visualization conceived as a digital installation. It has been exhibited for the 50th anniversary of IPFL, the Swiss Institute of Technology in Lausanne, or the Ecole Polytechnique Fédérale de Lausanne in French. And its idea is very simple. It's about presenting to the visitors the university's history by visualizing its doctoral thesis. So the IPFL is a federal institution in the French speaking part of Switzerland. The campus is composed of different buildings, some of which are remarkable for their architecture. And the Rolex Learning Center is probably the most representative one. It was designed by the Japanese studio Sana. And it's visible in the photograph as the rectangular building with holes in the middle, like a gruviera if you want. But certainly, you also notice that that there is a longer shape of building on the left. This is the IPFL pavilions, which is the Museum of the Federal Institute. Okay. IPFL pavilions is an exceptional building created by Kengo Kuma with a wood. Its original name was under one roof because it brings together three separated spaces. One for the Montreux Jazz Festival, whose archives have been digitalized by IPFL. One for the Data Square and all the projects related to big data at the IPFL. And one for public exhibitions. In the spaces dedicated to public exhibitions, the director of IPFL pavilions, Sarah Kanderdine, organized an exhibition called Infinity Room. You see the poster on the right. In the moment of its 50th anniversary, Infinity Room is an exhibition that aims to make visible the history of IPFL from different perspectives. So provision was the project intended to explore the perspective of education. When I was a doctoral student at IPFL, my studies focused on the visual representation of academic practice through the concept of affinity. One of the outcomes of my studies is the affinity map, a data visualization which represents the diversity of a practice in the academic environment. The affinity map is a network visualization using a collaboration metric based on publications, advising and teaching. In addition to these metrics, the affinity map also features a lexical distance also called cohort analysis. Using the TFIDF algorithm, it is possible to arrange laboratory according to the lexical similarity. TFIDF works with literary corpora. It allows to find the most relevant keywords for a document. Put simply, TFIDF is about the frequency of terms. The more two labs are closed in the space of the network visualization, the greater is the similarity in terms of dictionary between these labs. In the affinity map, if the affinity map aims to represent an organization through their labs and members, supervision is different. Supervision is an exploration of the language over the time. It is not about visualizing the present, but it is about representing the past by covering a much longer period of time. The investigation to create the data visualization started from the digital archives of APFL. APFL collects a publication of different types such as books, journal articles, conference paper, presentation and so on in a system called InfoScience. Among all these material, we found intriguing to work with doctoral thesis for a very simple reason. If for the affinity map we have been obliged to work with abstract, indeed the parsing of different papers of different provenance is very time-demanding. Doctoral thesis offered the opportunity to work with a pretty homogeneous material. To have an overlook on such material, we created a very simple data visualization you see in this slide. One dot is a thesis and our exaltant distribution is time-based. A color is gray when theses are not related to a specific department. It's orange where there is no abstract and it's green when we have the full metadata. Looking at this data visualization, it is visible at first glance the frequency of theses over time and the lack of metadata when color is orange or gray. In addition, it is also visible when the exalt date of defense is not available. During ten years in two different periods thesis records were just saved by here and for the rest we have the precise date of the year. This is a little bit new. The frequency of theses over the years is represented as a circle which allows to also the network visualization in the inside. It is noticeable how first thesis date back to the beginning of the 20th century. Even if EPFL has been officially founded in 1969, this situation was previously part of the University of Lausanne and this explains the present of doctoral thesis before the year of foundation. The TFIDF has been applied to all the doctoral theses using a string of texts composed of the title, the abstract, the keywords and the first five pages of the PDF parsed. Parsing these five pages enables to collect information for theses with no abstract but it even allows to collect abstract in both French and English. This data visualization shows a network of theses in which the metric employed is based on the frequency of words. In the arrangement is given using the TFIDF algorithm in addition to the UMAP conceived especially conceived for large data set. In this data visualization dot scholar stays for the department in which the theses have been defended. The color has been used to understand if the overall arrangement of the network visualization was meaningful or not. And it is pretty visible that the colors are well distributed. For example, doctoral students in architecture are grouped in the rose peninsula that is visible on the left side. I don't know if you see my mouse but it's this area. The gray mask at the bottom which is very visible. It's about the theses that were discussed in French with no English abstract especially before the foundation of IPFL. This is the reason of this cluster at the bottom of the data visualization. And again, this is the problem of this this morning with Christophe Mellaterra about the the problem the linguistic problem between different theses. So as I previously said, the idea was to host the network of visualization inside the circular timeline to create a unique object. And to enable interaction a three-dimensional mouse was chosen because its capacity to work on three dimensions at the same time. So the orthogonal axes are useful to move vertically and horizontally for selection. So basically it's like joystick. While the rotation is perfect to move over time like a control knob turn right for the past and turn left for the future. As a digital humanist I think that one of the major contribution of the field stays in the fact that humanists have became the designer of their own tools entering the world and the practice of design. I can cite in this sense a book that connects humanities and design which is titled digital underscore humanities by Bordek, Drucker, Landfielder, Prestner and my professor at Harvard University Jeffrey Snap. And it has been published by MIT Press in 2012. Like in Darwin's theory of evolution the design process works more or less in the same way. It aims to explore and compare different variations of the potential forms that the data visualization can take. These images show some steps and variations of a supervision. They are all different. During the design process a lot of attention was given to the timeline, to the selection and even the typography of course. Even if a lot of scholars employ data visualization as they are provided it is fundamental to think that graphical forms are situated and produced by graphic and interaction designers. As a data visualization is a visual vector to convey information, communication plays a great role in shaping the appearance of data. This slide is to mention that the data visualization is a complex object of technology which might be worthy of attention by the domain of science technology studies. Among the specialists you can see in these pictures we have the director of APFL, Pavillon's Sarah Kandardine, decorator Giulia Bini, the designer Patrick Donelson and the expert of data visualization, Philippe Riviere. All these peers had a specific role in the construction of the final object including me. And like cartographic objects data visualization are complex artifacts in which different specialists work together for a common goal. This is the final version of a supervision in full view. It can be navigated over time, moving left and right makes change in the selector that you can see in this point. The selection makes active in green those theses that have been published in that period. On the contrary moving in the space allows the users to select a specific thesis. In the example it is highlighted Roberto Seiga's work which is displayed through its metadata. And you have of course the position of his thesis. And you have on the right the metadata like the ear of the defense Enak, which is the faculty. You have the title of Ecologie Alpine. You have his name, the author's name. And you have even the directors of the thesis. And here we have four different selections all from the department of architecture. And here I come so far you have seen the digital version. But it is important to remember that this kind of data visualization has been designed for an exhibition. So it occupies a space. In these two photos Patrick Chouard and I are setting up the projector to find the best balance between resolution and distance of the visualization. This is a set of photographs of the final installation. A vertical furniture host the 3D mouse that the visitors use to control the data visualization. Some information are therefore conveyed with the installation to the public. One is about the life lamp which covers 50 years officially but is visible how the origins date back to one century ago the origin of IPFL. One is about the sites. This reading is the way to represent all the doctoral students together, all the thesis together. Other visual methods like for example Google search are based on lists which obliges to split records in different pages. And one other interesting insight we can get from the visualization is the frequency. Visitors can notice the increase of thesis over the time. Obviously this evolution is related to the sites of IPFL that are augmented considerably. It is interesting to notice the number of PhD that the institution annually deliver to the job market. The estimate today indicates more than a PhD day. Furthermore, the visualization reading can be more interesting if the visitor is part of the visual representation. In this picture you see Kirill Benzi, a colleague from IPFL that defended the his thesis in 2017. And all my work are not limited to analysis of data but I'm specifically interested to observe the reaction of the readers when they recognize themselves in the data visualization. In a moment in which there is an ongoing discussion on ethics and privacy of data, very few scholars are concerning how this information is presented. My work opens the room to a discussion in which it is the subject represented that evaluates the quality of the visualization and reflects about the production of data, especially today that the scholars are invited to share their information to open data protocols. So this is a video, yes. This is the last slide of my presentation. The animation you see is the idle function. We added to show to visitors that the visualization was live and interactive. If on the internet the user has the habit of interacting with data visualization, the use of data visualization in museum is a field that we are just started to explore. And for example, in museum we are used to have static artifacts. So it has to be visible that you can interact with this data visualization. And in my conclusion, I have three points that emerge from this presentation. One, of course, is the dimension of time, which is still a problem in the network of visualization. Indeed, we can create an animation like this one. We can create several pictures of different years, different periods, but it doesn't exist something that is able to connect all these pictures. So I like to investigate in this sense, putting a lot of efforts not just in the algorithm to display the notes on the screen, but even in the way in which networks, network visualizations can be represented in innovative ways. Another part is the part of the public exhibition. We have seen Katie Borner that she's organized by a lot of years, an exhibition of data visualization. But I think that the first data visualization was exhibited by the moment in New York 20 years ago. So it's a while, but data visualization are becoming more and more part of exhibition spaces. And the last one, of course, is related a little bit to the way in which we draw networks is the idea to bring the design, to bring the design process into the craft of this data visualization. And then I think it's an interesting point, especially in a moment in which designed universities are a little bit separated for the major subject of research. And you have to discover that the integration of different skills in doing research is just beautiful. Just a few moments, Charles, if you want to enter, I wanted to share with you even the live version because it's very different. So just a moment. Yeah, fantastic. Please do. While you're switching over, I think I can ask one very easy question that came up that you'll be able to answer very briefly. Do you have any videos of people interacting with the exhibit in the space, in the pavilion that we really don't know if you have? Yes, not for this one because during the exhibition I was in the United States and I was able just to be there for the installation. And otherwise, usually it's my job to observe like an anthropologist how people interact. And that was crucial in my thesis because my thesis is a representation of a university. And so it was risky, but the idea was to ask feedback to the scholars that are represented in the visualization. And this is part of my thesis and will be also part of my book that will be published soon in French and English in open access. So from June, from June, from Metis Press, Geneva. And this is the interactive map. So it's nice because you have this idle function and you can select, of course, here EC stay for computer science department. SB is basic sciences and ENAC is the Faculty of Architecture. But what is interesting is that you can navigate so you can go back and you have this feeling of quantity. And of course, when you go back, you have many, many cases in the part in the south that is a little bit lexically detached from the rest. And the other thing that is interesting to notice in the live version is the length of the selection bar because of course, in the past, you had very few doctoral tesis. So when you advance, you have more points. You have more tesis, but the bar, as you notice, is shorter. And I think it's just impressive that the idea to have a PhD defense each day and deliver 365 doctoral students, I think even more to the job market. Very cool. Let me, there's a couple of very nice questions in the chat. So let me go ahead and jump on these. So one from Eugenio Petrovic, so what was the visitor interaction like? Did the visitors immediately grasp the meaning of the visualization easily or did they need some explanation to interpret it? Do you have some captions or something like this as well or not? I mean, that was a big issue. You have even to think anyway that this visualization was the museum of a technical school. So people were aware about engineering data and all this stuff. And a part, a larger part of the visitors were a member of DPFL. So and you have to think even that all these people that defended the tesis in the past and were looking for themselves during the exhibition. But of course, these are new, strange objects to the common public. I think it is intelligent to make them simple because for example here, the bar chart is, I mean, the evolution is visible to the very eye. But it's, I think it's even interesting to push a little bit the technological boundary to ask to the visitors to do something more, something more. In any case, it was part of the expectation of the exhibition and not easy in general. Very cool. Yeah, that whole different whole different meanings to user experience design than I think a lot of us are normally used to. That's really, it's really neat. But you have even to think that now all the major universities have their own museum, even at the museum of MIT. I've seen a beautiful data visualization, not related to science, but even to archives. And you have people working in the sense of providing this visualization to the museums and even the Harvard Art Museums that was recently restored by Renzo Piano at the top floor as a beautiful room with a larger display that is open to new application and interactions. I have to, I have to visit. I haven't been since they redid the building. That's really cool. It's cool. So a next question from Rose Trappes who is wondering about changes in, oh, yeah, great. Cool question about changes in collaboration between fields. So can you get at least a sort of rough impression of how say interaction between disciplines changes over time just by looking at sort of the proximity between the nodes at a different time points or the movements of the kind of massive of theses over the over the over time in the graph? No, this is not possible in this kind of a visualization because if the the popping up is temporal, the distribution is just static. But of course, if you know the DAPFL, the new branches of of theses are related to the opening of new departments. And this is the big change because when a department opens, you have even funding for finance doctoral studies. Otherwise, and this is very subtle and we have to thank the designer for this. When you select a theses, you see the destruction behind and when the destruction is wider than the other part, this is the ramification of the faculty. So for example, you see this one. You have certain ramification and this is which is computer science is really in the opposite direction. And here, basic science is there. But sometimes you have a strange stuff because you see here in this case, you have really a division. You have a big mass on the top right and a small mass in the bottom left. This is because of course, interdisciplinarity, but even because within departments, you have a division in institutions. Very cool. Yeah, that's really, no, that's really, that's really neat. Thanks. Okay. I have another question here from from Nicola Bertoli, who asks, interested in knowing more about the, you mentioned the Darwinian nature of the network, was that just a metaphor, say a little bit, a little bit more about that. That is cool. No, that was just an idea because you know, when I was in the United States, Darwin is popping up continuously. And so for example, when I do design, when I design a logo, when I design an interface, you have sketches at the beginning, ideas, but then there is a craft work of refinement of all the elements to make, I don't know if I will, but you have to make it, make it working. And so starting from that point, you have really a ramification from the first sketches, you have different ideas, and from one variation, we have other variation, and was this my idea. So it was not related to networks precisely, but it was related to the nature of design, to the selection of your best idea, if you want, and even about the agreement with the others, because when you design data visualization in a team, but it's like it's easy, I mean you have to do an agreement with the others. That's cool. Yeah, I don't know, I like that way of thinking about the process. I actually want to pick up on that. This gets back to something that actually we were talking about, for lack of a better way to put it in the green room while we were getting connected. So on this kind of process idea, so I know, yeah, you mentioned that these are big, real, collaborative, serious team projects, and it takes obviously a large group of designers to build something that's this sort of articulated. I wonder what your thoughts have been about, again, I'm going to do this, I'm going to do another kind of asking you to ramble question again, because I think it might be helpful, about how to manage these kinds of collaborations and how these kinds of flows between, for example, people on the design side, people on the art side, people on the, obviously you've had to have, I'm sure, some historians at the university or people with the hard domain knowledge about what's been going on at the EPFL over these years, the people crunching out some text analysis stuff. So how have those flows worked in your experience? Can you talk a little bit about what that's what that's been like for you in a project like this? Because I think that's something that many of us would would like to do as well as you've done it in this in this environment here. Sure, I mean, this project was realized in 2019, but of course we started to talk about the visualization of the whole EPFL years before, and I've met Philippe Riviera in Lyon, and he's a very good data visualization guy. And we traveled from Lyon to Paris together, and at a certain point I told him we have to work together. And just after a couple of months I got this founding to create this map that was, I mean, it was good. It was 10,000, 10,000 Swiss francs. So I told him we can pay you, we can work together, and then we started to work on the data first. And I was doing the parsing. He was playing with UMAP, TFIDF, and when we created the first visualization, very raw material, we forwarded it to the designer. And for me it was the first time I worked with a pure designer that was not me for a data visualization. And it was nice. And there you have just to leave room to the creativity. You don't have to block. You have, I think it's one of the main rules of the management. You have to delegate. Delegate, accept, and I mean, if the team is good, the results will be very, very, very good. But you know, for example, then we had this good team. We worked for a couple of months through Slack. And then just one month before the exhibition, we were together in Lausanne to discuss. And then we accelerated and we obtained the final version. But, you know, then you have students around the lab. And so you ask them, so look in the visualization, what do you see? So it's very randomly. And for me, this is the basis of the design process. It's not the process of design thinking, but it's schedule for the managers. The process of design is creativity. And even the capacity to work with the others, this is a very good point. That's great. No, that's super helpful. Thanks. Oh, a question coming in from Christoph Malatter, who asks, oh, yeah, good, good question. Do you have a rough guess on how many person days were needed to get this, to get this project done across, across you and the team just to give the rest of us an idea of scale here? I mean, for, to realize this, I mean, I think it was three months for each of us. Yes. So, 1990, 1990. Sure. Okay, that's the total is this. That's still impressively fast, actually. Yeah, good work. Thank you. Very cool. Okay, let me see. I'm going to do my, let's do my stalling thing again and see if, if anybody else pops in. Actually, in the meanwhile, let me, let me ask what, I have a feeling, this is another very open-ended question, but what were some of the unexpected challenges that you had in transitioning from something that looked nice and was really reasonably interactive to you in your web browser in your, in your, in your own lab to something that worked in the physical space for the person at the, at the standing at the installation. Was it, was there, was there not, was it really seamless because you always had that in mind or did you, did you come into some points of friction and, and, and if so, what, what did those look like? I mean, for that I'm very inspired by the word of architecture as I, I told you I did this PhD in architecture because there was no design school and that was the closest one, but basically for the architects it's very difficult to imagine the interaction with people and the building. So basically, but even to, to the expectation, maybe now it's a little bit better with 3D rendering, but the volume of, of the building is this and the interaction of people with the building or the space in general is impossible to predict. And the same is for data visualization and I like to create a public visualization because, yes, a browser is great, we can talk together, I can present stuff, I can share stuff with it, which is very powerful, but at the same time you don't have people together. So for example, a part of my thesis was just about the conversations you can have in front of a visualization because of course we have a visualization saying fats and we have a visualization about the exploration. All these, I mean, at least this one for me, it's an interface for exploration. So the best stuff is coming out from people in front of the visualization interacting together and this is unpredictable. Very cool. That's great. Yeah, that's a, that's a, that's a really nice, that's a really nice point. I like that. I like that idea a lot and it's a, it's a kind of data that we don't, we often don't. I mean, you can do kind of user experience trials or something on a website, but we don't normally get that kind of, that kind of situated experience when we just sort of launch a website off into the world and hope that people interact with it in the way that we expect them to, which they never do, right? They never click on the things we expected them to click on. So, yeah, no, that's a, that's a really, that's a really great point. With that, we're, we're basically out of time. So let me go ahead and, and, and wrap things up there. Thank you so much. This is really, really cool. I really enjoyed getting to see this and then let me just say once more to everybody up at the top of this crowdcast chat is a Zoom link. We'll be starting there in 15 minutes. So we'll be starting there at quarter past, quarter past the hour for yet more cool visualizations. This is the cool visualization block of the conference. So yet more cool visualizations, more, more examples along the lines of, of some of what came up in Dr. Berners talk at the very start of the conference. So we'll see you all there in just a few minutes. Thanks very much.