 International Women's Day to everyone. If you were joining us for the first time today and would like some Japanese interpretation in your ear, please go to the back to grab an earpiece. Thank you so much for joining us at the third keynote session as part of C. Hub's annual symposium, which as you know brings together bold ideas and innovative thinkers to explore themes that enhance our ability to foster excellence through inclusion. And this year's International Women's Day focus is invest in women, accelerate progress. The speaker you're about to hear epitomizes these themes. It is my absolute pleasure to introduce you to Catherine Dignasio, an associate professor of urban science and planning at MIT. She is also the director of the Data and Feminism Lab, which uses data and computational methods to work towards gender and racial justice, particularly in relation to space and place. Catherine is one of those brilliant people whose curiosity knows no bounds, which has allowed her to work outside traditional boundaries and bring people, groups, and ideas together in creative and meaningful ways. Her presentation today, which will be delivered online as you can see, is entitled, A Feminist Approach to Data Science and Communication. She will speak for about 30 minutes and then we will have time for your questions. Without further ado, let's give Catherine a warm, oist welcome. Hello everybody. Here, I'm just gonna take a second to share my screen. There we go. All right. So, yes, good morning. Evening for me, but morning for you and happy, happy International Women's Day. I think that's fantastic that I get to celebrate with you even before it's my women's day here. So I'm Catherine Dignasio and I'm going to be presenting about A Feminist Approach to Data Science and Communication. And I do just wanna mention, I'm very grateful to Kathy Takayama for the invitation to be with all of you here today. So first, I just wanna thank the Wampanoag people and the Massachusetts peoples on whose lands I am a guest and on whose lands I completed much of the work that I'm speaking about today. So what I'm speaking about comes from Lauren Kline and I's book called, Data Feminism. This was published by the MIT Press in 2020. And in the book, we present seven principles for doing more ethical and equitable data science. And these principles are based on the teachings of intersectional feminism. So first today, I want to talk a little bit about our motivation for writing the book. So why is it important to bring feminism and data science together at this particular moment in time? I also wanna define what type of feminism is informing our work because there are many feminisms. And then I wanna show some examples of the seven principles of data feminism in action. So that's our roadmap for today. So just to get started. So as Lauren and I were writing this book, we were increasingly aware that data was becoming a tremendous form of power. And we mean that sort of metaphorically, but we also just being a very literally, like the corporations who have the most sort of money and power and resources in the world are corporations that are leveraging data. We were learning from Sophia Noble who wrote the book, Algorithms of Oppression from authors like Virginia Eubanks who wrote Automating Inequality. We also were looking at the work of Joy Borumlini and the Algorithmic Justice League who have been doing some amazing work to showcase racial bias in facial recognition system and algorithms. And what all of these books are pointing to is the fact that data are indeed incredibly powerful. However, that power is wielded unequally. And more specifically, it's wielded by a small and homogenous group of corporations and other well-resourced institutions. So these are the ones that have the resources to design and deploy data and AI systems. And when they're only deploying AI systems in the service of their own profit, it comes at the expense of everybody else. And this is where feminism comes into the equation. So what we explain in the book is how feminism and in particular intersectional feminism has been focused on precisely this, on imbalances of power and on the structural forces that cause imbalances of power for a very long time. So in other words, we might be looking at these new systems like artificial intelligence or algorithm, chat GPT, large language models. But the basic idea that systems are optimized for certain groups and not for others, this is not at all a new idea for feminists. So one of Lauren and I's main motivations, in fact, for writing the book was because the popular press who kept writing about data and algorithms, they kept being surprised. There's always this surprise coming out in the news that this or that system is biased against women. Like it's demoting women's resumes, for example, for a hiring algorithm or a soap dispenser that is racist, that doesn't dispense soap for dark skin. So all these examples of bias and the journalism would throw up their hand and be like, how could this soap dispenser be racially biased? Like that must be impossible or how could a hiring algorithm discriminate against women? But in fact, in a sense, maybe I should just back up. So we were just frustrated by the fact that there was always surprise at these things because as a feminist, we can say that these kinds of flaws or errors in the system, they're not surprising at all. In fact, they are completely predictable, but they're only predictable if we know to draw from feminism and or other theories of social inequality. Okay, so we'll come back to the principles of data feminism in just a minute, but first let me back up and give you a very short introduction on feminism. I don't know where people in this audience are or whether people have ever taken a class in women and gender studies. And most people, if they know anything about feminism, they think that feminism is about women and that's correct, but there's more. So feminism at its core entails a belief in equality for all genders. And so if you hold this belief, if you think that all genders of people are equal, this is a feminist belief. So it's a belief, first of all. But secondly, if you take a look at the world around you, you can realize very quickly, in fact, that this goal of equality of all genders has not been realized in the world. And so that brings us to the second definition of feminism, which is that feminism is political action. It involves political activity to realize this belief in equal rights for all genders. To make that goal truly a reality in the world that we live in. And then feminism has a third definition. And this is that it's a set of theories and ideas. It's an intellectual heritage. It goes back decades and even centuries where thinkers, activists, writers, poets and artists have thought about this problem of why we don't have gender equality. That is the basic starting point of feminism. And they've theorized it, they have taken action. But what it also means is that we have guidance from the past to think about issues of inequality today. And these, this guidance, feminist thinking, feminist intellectual heritage begins by thinking through with respect to sex and gender. But then the past 40 years of scholarship and also our current political reality have brought many more dimensions of inequality into the conversation, including race, class, sexuality, ability, immigration status, and so many more things. And so this brings us to the core sort of feminism that Lauren and I are drawing from in a book, which is intersectional feminism. So this is a type of feminism that comes to us from the work of women of color feminists and black feminists in the United States in particular. And this perspective takes the position that we can't only think about sexism in order to understand inequality. And so we get this idea, people think about this idea as being new, but in fact, this is an old idea. It was proposed at least as far back as the 1800s by Francis E.W. Harper, for example. And she and others are theorizing directly from the experiences of black women who at the same time face both sexism and racism. And so Kimberly Crenshaw uses the metaphor of the intersection to talk about the place where these two forces come together. So racism and sexism compound and combine. And so we have to analyze them together. We have to look at them together. Other intersectional work has built off of this work and looked at other forces of oppression. So things like classism, colonialism, ableism, and other dimensions of structural inequality. And so what we get from intersectional feminism is a framework to critique structures of power to understand why certain people may experience oppression on the one hand or privilege on the other hand. And so intersectional feminism helps us open up some of our analytical tools and make our analysis more complex. So really think about like, what are the root causes for why these inequalities exist in the world? So what we do in data feminism is we draw from these teachings of intersectional feminism, along with other ideas from feminist activism and critical thought, in order to arrive at these seven principles for doing more ethical and equitable data science. So the principles are these, and I'm not gonna explain them right now in depth because I wanna give you examples of what these principles look like in action. So there are things like examine power, challenge power, rethink binaries and hierarchies, elevate emotion and embodiment, embrace pluralism, consider context and make labor visible. And one of the things I'll draw your attention to is that we actually have two principles about power. And this is because of this sort of analysis of unequal power and analysis of inequality is so central to the feminist project. So if we're starting to try to rectify inequality, first we have to understand it. We have to examine power. And then secondly, we have to challenge that power. We have to think about how do we rebalance so that things are structured better. And so the goal of these principles is to really think about how do we use feminism to do data science? It's really about like how do we draw from it but to inform practice, to inform our work with data science and also to inform our work with data communication and visualization which I'll talk more about in just a minute. In the book it's structured like this. So we have one principle for each chapter. And what we do is we try to talk about the feminist theory that let us formulate that principle and also give multiple examples of that principle in action in the work of data scientists, data visualization designers and other folks working with data. So to say like the stuff is already happening. People are already elevating emotion and embodiment in data. So as a way of illustrating these principles in action, I'm going to pivot and discuss some of the examples that we talk about in the book. So one of the first points that we make, I will actually maybe I should back up. One of the connections that I wanna make in our present moment, many of you have probably been a part of and seen all of the hype about chat GPT and large language models and AI. And I certainly hope at your institution you're talking about these things because these are big public conversations right now. We still talk about data feminism even in the midst of all this AI hype to call us back to the idea that all AI algorithms and large language models are ultimately powered by data. So a lot of the biases and inequalities that we see in these algorithms and in these chatbots and so on, these are models that are reproducing data problems. So I just want to sort of make that connection or make that link between data and AI, since AI is so much in the news right now. So one of the first points that we make in the chapter about examining power, the first chapter is that data bias, so bias in our data sets, it starts even before the data set exists. So in data feminism, we use the example of this artwork. So this is an artwork called the Library of Missing Data Sets. It was created by Mimi Unwocca and we use this to talk about this a little bit. So she makes the point in two ways. First, this is a GitHub repository. It's a simple list on the internet and it's a list, it's the list that you see depicted on the right. And it's a list of data sets that do not exist in the world, but that Mimi Unwocca thinks should exist. So these are things like trans people killed or injured in instances of hate crime, people excluded from public housing because of criminal records and so on. So these are sort of urgent and interesting data sets, data sets that the public would benefit from, but no institution is collecting this information. And then there's a second way to see this, which is the artwork form. And that's the cabinet that you see in front of you where it's this white filing cabinet. And each of these data sets is tagged as a tag on a single folder. And when you're the visitor in the gallery, you can go through and page through the files in the file folders and you can look at the names of the data sets that are missing. But when you reach in, you take out a folder, you open it up, there's no information in the folder because no official data have been collected about that topic. And so the point that the artist is trying to make is that these data sets are missing for a reason. And the reason is that there's a really profound imbalance of power with respect to data collection in the world today. So who has the power to determine which data are important to collect on which data are not important to collect? Generally speaking, it's governments and other money institutions that make these decisions. And generally speaking, minoritized groups, community-based groups do not have the power to build and then maintain these large data sets. So this is why a feminist approach to data and to AI begins with an analysis of power because far too often even just the data that we have access to, this has already been over-determined by the imbalance of power in the world. So there are many things, many urgent and important things that we don't have information about. And they follow along, I should say too, the connection point here is they follow along gendered lines as well. So for another example, kind of building on this idea of missing data, feminicide is another very profound case of missing data sets. So in the book, we tell the story of Marias Casalguero, she's a Mexican citizen. And so she resolved to head straight towards this problem of missing data and collect the missing data herself. So for those of you who don't know, feminicides are gender-related killings of women and girls. They include cisgender and transgender women. Feminicide is a human rights violation. So it's recognized as one of the most egregious human rights violations by the United Nations and other international institutions. It is legally defined as a crime in almost all countries in Latin America and that includes also in Mexico. But the state, even though there is a law in the books, the state does not collect systematically data about feminicide. So Marias Casalguero was frustrated by the lack of formal action on this topic. And so she has single-handedly compiled the largest archive of feminicides in Mexico. And so how does she do this? So since 2016, she has spent two to four hours a day logging these cases on a Google map. So the Google map that you see here. And she finds them in media reports through WhatsApp group chats, through attorney general websites. And then when she finds information, she puts it in her database and logs it on her map. And so using this map, which is now an eight-year-long archive, she has helped families locate loved ones. She's provided data to journalists, to NGOs. She's testified in front of Mexico's Congress multiple times. And so in data feminism, we talk about this as what you might call feminist counter data. So this is a kind of a activist data collection that steps in when the state and other institutions have systematically failed to ensure the basic safety or the basic rights of their population. And so this is one way that one could use data to challenge power is this collection of counter data. And I just wanna take a moment to mention that writing about Maria Salgaro's work in Mexico led me to interview her and then that led me to interview other people like her. And it turns out there are many data activists working in this way on the issue of feminist side globally, not only in Latin America. And subsequently I've ended up undertaking a collaborative project for the past five years, which has culminated in this book that is coming out next month called Counting Feminicide. And so in this book, I describe some of the creative and intellectual and emotional work of feminist side data activists. So again, this idea that data activism is a way to challenge power and a way to push for social change. But let's keep going. So we've also seen, so kind of continuing with this thread of missing data and also with this idea of examining power and challenging power, we've seen another example of missing data play out very invisibly in training data for large language models. So these are the models that drive apps like chat GPT, for example. So what we know about large language model data, the training data, aside from that fact that it's large is that it comes from the internet. So vast quantities of internet data have been sort of harvested to train these models. And so the internet is many things, but what it is not is a full and equal representation of the world. The internet data, if you think about platforms like Reddit or Tumblr or things like this, it skews young, it skews male, it skews towards the US and the global North, it skews towards toxic speech, all right? And it is certainly not the source of truth that people interpret the output of chat GPT to provide. However, the makers of LLMs, including OpenAI and others mostly do not disclose their training data. So we basically have no way of knowing or testing or auditing just how biased these models actually are. And so a recent paper released with chat GPT or GPT-4 explicitly disclosed the data sets that had been trained on. And that's what this post is talking about here on this slide. And, but we can still reverse engineer these systems. And so this has been the subject, I think, of really interesting work. Again, with this idea of challenging power, how do we peer into these black boxes and examine some of their biases? And so one of the things that Lauren is involved with right now, led by a graduate student at Berkeley named Lucy Lee, involves trying to reconstitute some of the training data so it can be more closely examined for biases. And so one of their findings so far with this work is that even within English language training data, meaning like all of the texts are written in English, texts that comes from certain regions of the world is deemed as lower quality texts by the model. And it ends up being excluded from the training data in higher percentages. So this is consistent with other research that has showed how AI plagiarism, often, or the plagiarism detectors often flag non-English speakers or non-native English speakers at higher rates than native speakers for plagiarism. There's a kind of bias against non-native English speakers. So up till now I mainly talked about examining power and challenging power in data sets and data systems, but how do we make data science and AI processes more inclusive? This is why this idea, this principle of embracing pluralism comes in. So to formulate this principle, we drew from fields like human-computer interaction and participatory design. And these have long track records of research around public process, participation, and co-design. And in data feminism, we characterize these approaches as embracing pluralism. And so what that means is listening and learning from a wide range of sources of knowledge and perspectives throughout the data science and data communication lifecycle, including the part of like training models, annotating data, training machine learning models. All of these are things that are opportunities for participation and for involving more people in the process. And so, yeah, I'm gonna keep going. I'm not in the time when I don't have that much time. So I'm gonna keep going if I can tell you about a couple more principles. And so this idea of participation is something we've actively mobilized in my collaborative project about feminicide. This is the project my new book is based on. And so what we've done on this project is we've worked really carefully to bring together these data activists across Latin America and globally who are working on the topic of feminicide. And we bring them together to build community. We bring them together to conduct qualitative research. And we have brought them together to do co-design and participatory design of AI and machine learning technology. And one of the things that has come out of this is that the several tools that we've built with activists where activists have participated in all stages of the data processing pipeline. So we've worked closely with activists to conceptualize digital tools. We've worked closely with activists to annotate training data, to train models, machine learning models, and then to evaluate those models as relevant for their work. So the main takeaway here is that AI, data science, computer science, and then data communication all have a lot to gain by this idea of involving people, real-life people at every stage of these what we typically consider to be more technical processing. Okay, so the previous examples have focused mainly on the issue of power and people, but a major idea that comes with feminism has to do with critiquing binary structures of power. Feminist theories have helped to show us how binary distinctions are often hiding hierarchies with one group on the top and another group on the bottom. The distinction between the idea of man and the idea of woman is the obvious reference point here. So typically we inherit the common sense idea of gender, which is wrong, which is this idea that there are two genders. So first of all, this is empirically wrong, right? There are more than two genders. But second of all, what we see as a gender binary often is hiding a hierarchy, right? Because often actually it's the men that are on top and the women are on the bottom here. And so feminists in general are very skeptical of binary. So when we encounter a binary, how do we think about this and use this to question other binaries and hierarchies that we encounter in the world? And these would be things like the distinction between nature and culture, between subject and object, or what this next example takes up, which is the artificial distinction between reason and emotion. So in an Anglo-Western context at least, we've been taught that reason is somehow better than emotion. We see this play out in data communication in particular. So often the best practices for data communication involved a very clean design, very minimalist aesthetic, is kind of like just the facts sort of presentation. But in data feminism, Lauren and I question that and we say, why? Why are these our best practices? Minimalism, neutrality, kind of a flatness of affect. Why are these our best practices? Especially when research shows that emotion helps us to learn, it helps us to remember things and it helps us to connect with each other. So what about visualizations, data visualization, that deliberately leverage emotion? And that's what this next example helps us think about. What you're actually looking at here are screenshots of animation. So what happens here is that these arcs that you see on the screen, if you go to the webpage, they get traced out one by one. So there's an arc, there's an arc, and then they speed up and then they start to come very quickly onto the screen until the picture builds and the picture is this one that you see in front of you. And what this is showing is the number of gun related deaths in the United States in 2018. So each of the people killed by a gun in that year is represented as one of these arcs on the screen. And the arcs are traced one by one, but then they get faster and faster and faster. So it's very overwhelming to watch. It's almost really unbearable to watch it. And that's exactly the point. So it's an animation and a visualization that it just goes on for too long because there are too many people being killed by guns in the US. It's an epidemic. And I should say this, just to explain this a little bit, on the bottom here is a timeline. And so each arc that goes across the screen is one person's life. And as their arc, as their life is cut off, that arc turns to gray. And the gray part is what they would have lived to, the age that they would have lived to if their life hadn't been cut short by violence. Catherine? So the data are very sound. They come from a national crime data set. It's very statistically sound. And yet when this first came out, this was viewed with suspicion by the visualization community because it made us feel. It was communicating with us through emotion. And you can see that here where the number of the statistics go up and it computes the sort of speculative statistic of many, many, almost a half a million stolen years. And so people were like, yeah. You had asked me to tell you when you're at 30 minutes and you're here, we're here. Oh, good. Okay, great. I just have one more slide after this is perfect. Thank you. And so people were suspicious because it made us feel things. It was a visualization designed to provoke emotion. But a feminist approach would say that it's not a problem at all that it made us feel something. In fact, it's more compelling as a visualization because it blends emotion with reason. And so in the chapter about elevating emotion and embodiment we talk about how we can rebalance our data communication toolkit to really think about how we balance reason and emotion. And we can talk more about this in the Q and A of folks are curious. So the final point I want to make is probably obvious, but it's just that data feminism insists on an expanded definition of data science and data visualization and an inclusive approach to artificial intelligence. And if we expand our definition of data science we can see that some of the most exciting work in the world today is being done by artists, by journalists, by community organizers, librarians and activists. Some of this work does look like traditional data science but other work looks like artworks, data-driven sculptures and artworks, data murals, data-driven journalism. And so we have hundreds of examples like this in the book which we selected to illustrate our points and inspire people to action because just as data might be at the root of many of our problems today in the world we think it can also be part of the solution. So yeah, thank you so much. It's truly an honor to be part of this symposium. There's many ways to get in touch with me and I'm really looking forward to the Q and A and comments that you all might have. Thank you so much. Excellent. Thank you, thank you, Catherine. We have about 10 minutes for questions which I know isn't enough. Nonetheless, Catherine has assured me that she's open to any and all questions and we recognize of course that we come to this topic, to these topics from different history and different experiences. So we have the mics at the ready so please raise your hand and we will run right over to you. Hi Catherine, I'm Huang Hun from Singapore. I have two concepts that we think quite a lot about in feminism and what that is, third space and the issue of questioning the logic of the dominant hierarchy. I do hear some of the echoes in the seven principles but I would like to hear a little bit more elaboration on that. The third space concept, is that what you're advocating for in that whole continuum between objective data and subjective narrative? And is that supposed to be introducing a new logic? Is that clear? I think I did hear the two concepts that you mentioned. I'm sorry, maybe the audio is a little bit muddy. Heather, is there a way you can repeat the two concepts at the same lunches? Third space and logic. Logic and logic. As in logical structure because the dominant hierarchy has already defined a certain frame. So I'm just wondering whether in this meeting of this redefining of data feminism, is that a way of talking about a new third space? Is there a way of challenging the logic of the institution? Yeah, great, thank you. I think I understand the question. Yeah, I think what we're trying to do is take some of the logics, it's really it's an epistemological move and it's to say, how do we think about other ways of knowledge creation that are not only, like how do we bring one of feminism's contributions, right? Is to expand who's at the table in knowledge production. That's where the pluralism idea comes from, to say that many voices produce a kind of more robust knowledge, so like kind of expand who's present in the knowledge making table. But also to say that there's not only one way to do even data science, right? Which we tend to think of as something very quantitative, something that has a kind of very specific set of methods, but what we actually show in the book is that data science, especially again, if we've used this expanded idea of data science, data science can involve community based participatory work where people are painting murals, you know what I mean? From conversations about data and statistics. And so it also has to do with how we define data, which for us does not mean only quantitative information. It also means, it just means any systematically collected form of information. So we could be collecting photographs, like any kind of archive can also be considered as a form of data. It could be qualitative data, could be testimony, could be stories. And so I think those are some of the feminist moves to, yeah, I think kind of come into the cracks of what is perceived as a kind of very rigid frame of like data, data science, quantification, also something that's like yielding a lot of economic value and say actually there's a whole different way to do these things and here's what it consists of. So I'm not sure that entirely answers your question, but hopefully that starts to answer your question. I was going to say it's a big question. Perhaps I can invite you to follow up. Maybe we can share the contact information after. While folks are thinking, oh, Melanie, please. So this is a sort of a reflection and a comment, a question to Catherine and also to my scientific colleagues. So I work at OIST, but I'm not a scientist. And I was very, very intrigued by the point you drew out about putting emotion or judging emotion as somehow lesser than reason. And that's another really big topic as well about how that came about and so many links to what you were talking about, about how history has determined how we frame things. But, and also how women are seen because women are often identified with emotional responses in the workplace. But this is actually partly just a kind of comment on that and say how stimulating I found that was, but also a bit of a question to any of my scientist colleagues here today because science, the kind of conventional, successful scientist and conventional science I think as a non-scientist, it elevates rational decision. It actually, and this is very much to your data point that data numbers are regarded as somehow having the authority and they safely inform decisions and everything you've been telling us this morning shows that definitely needs to be questioned. So I'm really interested in what my science colleagues think about how that hierarchy between reason and emotion actually informs their working lives and how they feel the model of what good looks like, what a good scientist looks like and how those values are reflected in the career structure and career progression. And leading on to that is what efforts we can make to change those, to enrich those things. So thank you for such a stimulating point. I don't know if you have a response over here or another question. Hi, yeah. No, just responding to that as a scientist. I think as an emotional scientist. So for me throughout my career, obviously the science is the science but when you want to communicate your science and especially when you're trying to communicate with maybe audiences that not so accessible. So audiences that aren't really necessarily really wanting to be engaged with your science. I have used work with artists to enable me to do that. And so I think that sort of allows you to bring in that emotion but it's really interesting in working with an artist because they have a very different approach to their work in that their whole experiences, the whole life experiences, everybody that meet everything comes into their work. Whereas a scientist were traditionally taught that that sort of gets left at the door and it's the same as women when you're pregnant, you're not pregnant at work. You can't be pregnant at work. So, and all those things. So I think there is something. And also I talked yesterday about being neurodivergent and being one of the things that I do a lot is sort of I jump, I have, and people would call that, oh it's female intuition. And again in a very derogatory way. Whereas I can see things big jump away. And so that's how my research works is that I take big jumps and that's what enables you to be a really good scientist. But it's put back on you as a negative thing because the logical thing is to do thing in a very sequential little step sort of way where yeah, she will get there. But yeah, so I think, yeah, lots to discuss. Thank you. As it happens, we have our PsyArt consultant in the front row here today. So that's a wonderful connection. Katherine, maybe we have time for one more question but I can already see all the discussion that you've stimulated. So I think our breaks and our lunch hour will be buzzing today. Yes, one more at the back. Great. I just want to comment and build on the previous comment. So I think for some, it's important to recognize that for some scientists, while there's a sort of shared norm across modern sciences to sort of leave yourself at the door of the lab, like outside, that some scientists are forced to do that more than others. And a lot of how scientific concepts are structured and how data are analyzed. For example, let's say talking about biological sex in terms of strict binaries. When you actually start looking at that, there are variations in sex characteristics that vary along various spectra. And so it's really a more dynamic and multi-dimensional, high-dimensional variable. Biological is what we mean, what we usually talk about with biological sex but that's usually compressed down to binary categories. And that for maybe most scientists who maybe happen to be cisgender and no sex, so not intersex, that doesn't appear as a problem because that fits their everyday experiences of gender and sex. But for transgender scientists like me and other people, maybe also intersex scientists, that categorization automatically leaves them out as aberrations or outliers and we don't really make it into the data or we're kind of marginalized in the data by the structure, the analytical structure itself and the concept, binary concept itself. But we're kind of forced to play along with that most of the time because that's what other people have done and that's what a lot of literature is based on. So we need to kind of address previous findings from a literature, so we have to play along. And so we're forced to leave out our own experience of sex and gender at the door, which I think is unjust and needs to be recognized more and I think this goes for many other systems of oppression as well, along racial lines, for example. So I just wanted to comment on that. Thank you, thank you very much for adding. I'm so sorry, but we have to wrap here, the day is ticking on. Please remember to respond to the symposium survey and you can be entered to win a copy of Catherine's book, Data of Feminism. To all of you here, thank you so much for joining us for this presentation and for the symposium. Thank you for bringing open minds and curious outlooks. And Catherine, thank you so much for sharing your time, expertise and passion. You have given us much to think about and please stay on the screen after we just wrap so that we can get a photo with you in the background. I think that's because I must thank you very much. Thank you so much. Thank you.