Let’s Talk Art: An interview with multidisciplinary artist Michelle Lisa Herman on collaborating with AI software to expose bias

Published in  
February 27, 2024
What can artificial intelligence tell us about bias in art and technology? Art historian and museum editor Sarah McGavran’s interview with multidisciplinary artist Michelle Lisa Herman explores how Let’s Talk Art, a 2020 artist’s book written in collaboration with machine learning software, uses humour to reveal underrepresentation in both fields, particularly in terms of gender, disability, and race. Additional topics include exclusionary conceptions of art and artists, art speak versus accessible language, as well as objectivity and subjectivity.
PDF icon: click to go to full PDF of article

How to cite: Herman, M. L. and McGavran, S. 2022, ‘Let’s Talk Art: An interview with multidisciplinary artist Michelle Lisa Herman on collaborating with AI software to expose bias’, Culture Caleidoscoop 1: DOI 10.57031/culcal.v1i1.12192

A note from the interviewer, Sarah McGavran

Michelle Lisa Herman and I met at the artist’s solo exhibition Already, Never, Better, Faster, Stronger at the Stone Tower Gallery (Glen Echo, Maryland, USA) in February 2022.

I was especially drawn to Let’s Talk Art, an artist’s book composed of completely manufactured oral history interviews with fictive artists. Herman generated the texts using the Generative Pretrained Transformer 2 (GPT2) open-source artificial intelligence (AI) software, having ‘trained’ it using a publicly available online database of oral history artist interviews from the 20th and 21st centuries. Through unintentionally funny software-generated prose, the book highlights art world clichés while emphasising the experiences of white male artists, just like the original interviews. Even the table of contents makes this clear (see the video).

Extract from Let’s Talk Art by Michelle Lisa Herman, 2020. Animated video and artist’s book (written in collaboration with machine learning software). 1 © Michelle Lisa Herman

As an art historian specialising in gender and as a museum editor who strives to make academic research accessible to wider audiences through engaging prose, I felt resonance with Herman’s purpose. The project also sparked my curiosity: there are many ways to assess omissions in the art world, including historical research, statistical analysis of artists or subjects represented in museum or archival collections (like the set of interviews Herman drew from), and museum visitor and staff demographics.2 To learn more about what this unconventional method for analysing oral history artist interviews would add to the conversation, I asked if Herman would be willing to do a ‘meta-interview’ about Let’s Talk Art.

During Herman’s residency at the University of the Arts Bremen, we spoke over Zoom about the artist’s experiences in the fields of art and tech, the process behind Let’s Talk Art, and the larger issues it raises. The structure follows a traditional artist interview, where the initial questions focus on the artist. We have edited the interview transcript for clarity and concision.

Meet the artist

Sarah McGavran: Before we delve into Let’s Talk Art, I’d like to explore your perspective and the larger social and cultural contexts your work addresses. How do your experiences related to gender and disability as a multimedia artist – and in your work as a part-time web professional – inform Let’s Talk Art?

Michelle L. Herman: I think being a woman, and then particularly a woman with disabilities working with technology, there are a lot of what I call ‘barriers to entry’. This is true of both the art and tech worlds. I’ve worked with technology in my artistic practice, but then I also make a living as a web professional, and I think that these barriers really apply there as well. In addition to the added obstacles to entering these spaces that marginalised groups face, once you’re in, you’re continually being pushed out. For example, my knowledge and abilities are often discounted, particularly by men. I’ve overheard comments made about whether I could use a keyboard with my physical disability and had colleagues make assumptions about my technical abilities. This happens to me much more often than to my male counterparts. And it’s often harder to access certain equipment or venues.

Studies suggest that women-identifying folks are judged more harshly than men.3 So that impacts every situation in which I am judged: applications for shows, residences, awards, and more.

In a lot of ways, the book was my way of dealing with all of these barriers, obstacles, and forces, and trying to navigate both of these worlds while also trying to point out those gaps in the written history of art.

Sarah McGavran: In a research project4 for the local arts organisation Washington Project for the Arts, you’ve also called attention to the ways that the fields of art and technology perpetuate gender discrimination. Which women in those fields – or women artists working with technology – are important for your practice, and for this work in particular?

Michelle L. Herman: Right now one artist would be Lynn Hershman Leeson because of her very early work using artificial intelligence and interactive technologies. She had the first interactive CD-ROM video art piece, Lorna (1979–1984). I’ve followed her work for a long time, and I actually just got to see her installation Logic Paralyzes the Heart (2021) at the Venice Biennale. Some others, in no particular order, are Sondra Perry, Laurie Anderson, and Dara Birnbaum.

A few months after I created the book, I interviewed Legacy Russell, the author of Glitch Feminism, who is now the director of the Kitchen in New York.5 I resonated with a lot of her ideas, especially the ‘glitch’ as a form of resistance – this purposeful failure to perform as expected.

AI for beginners

Sarah McGavran: The idea of failure to perform as expected makes a great segue to AI software, which still has glitches in everything from spellcheck and translation to captioning. Could you explain how GPT2 works and expand on this concept of the glitch in relationship to it?

Michelle L. Herman: This is a little more of a user-friendly answer than a technical one. The company that developed GPT2, or Generative Pre-Trained Transformer 2, is called Open AI. They developed GPT2 as an open-source artificial intelligence application in 2019 to generate natural language text. So it was designed to write text that would be really difficult to distinguish from that by humans. As I understand it, the system is considered pretrained, which means that you provide the system with a data set – or in this case texts – for it to learn how to emulate it. In this case, GPT2 was pretrained on a database of unpublished fiction books called Books Corpus. From there, you can ‘fine-tune’ the system by retraining it on different texts.

For Let’s Talk Art, I collected publicly available oral history interviews of artists. That’s the data set I used to ‘fine-tune’ or retrain the system. This resulted in the system being able to generate new fictional oral history interviews – formatting and all.

In terms of the glitch: I think a lot about how these systems – AI, really, any kind of technology-based application – are largely developed by white men. Since 2015, about 92 per cent of all programmers are male,6 and more than 50 per cent of them are white.7

It’s sort of a scary concept because in many ways we can see these technologies reflecting their creators, right? There are these gaps in the output, in representation, that illustrate how these technologies really represent that white male perspective. And so by pointing out the problems – the glitches – it’s my hope that we can understand and make that idea more visible, that these programs aren’t representative of everyone and that there are a lot of implications there when you consider how they are shaping our lives.

Sarah McGavran: You make a strong case for integrating different perspectives and subjectivities into these technologies so they can work better.

In the US art world over the past couple of years – especially in the wake of George Floyd’s murder – there has been increasing acknowledgement that museums and other cultural institutions don’t reflect the diversity of the general US population in terms of collections, programming, and staff. What do you think AI software adds to this conversation about art institutions?

Michelle L. Herman: Many of these institutions and the art world, like the GPT2 system itself, were developed by and around predominantly white men, so it makes sense they still reflect their perspective. AI shows how visible that fact is. In other fields or even social dynamics, things can be a little more nebulous or blurry – maybe because we can’t always see the whole picture? But when you see the results of many of these AI s, you can’t help but think, ‘Oh, that’s a problem’.

Sometimes it makes headlines when the AI is found to be blatantly racist or sexist, but we often write it off as a software problem and don’t consider how the system’s creators designed it to do these things.

For instance, Amazon had an issue where they were using AI to prescreen résumés of potential applicants. The data set they used were the résumés of mostly male candidates they received over a 10-year period. So, the system learned to reject women’s résumés based on keywords like collaboration, or the word women’s, in the context of having attended a women’s college.8 Another incident is Google’s image recognition system. It was labelling images of Black people as gorillas. When it was discovered, Google decided to eliminate ‘gorilla’ or ‘chimpanzee’ as possible labels rather than fixing the software.9 In both situations, the news coverage did seem to deflect the blame to the software instead of the 92 per cent male programmers who designed it.

The troubling thing is how many everyday technologies we interact with are using these systems. This will increase in the future. I think by pointing out these gaps and failures in representation in the people creating these technologies, the data they are using, their assumptions, and what happens when that goes unchecked, we can hopefully motivate the creators to address them.

Strategies for inserting yourself into the conversation

Sarah McGavran: You have to recognise a problem in order to do something about it. And that is exactly why it’s so important that you’ve inserted yourself into this conversation by making art with this technology. How does your own subjectivity come into the work? In some of the interviews, you inserted your name or your initials, perhaps as a way to acknowledge your own position?

Excerpt from ‘The Human Experience’
MS. HERMAN: Well, there’s this thing that you can do, or you can’t do, that applies to many artists. And I think that’s one of the things I think about the art world. It’s not like, ‘Don’t get married. You can’t be a great artist.’ Well, I think that’s part of the appeal of art.
MS. REEVES: Yeah, and what about, in your questions, are you interested in the world of art, in the arts, and also the world of commerce? Or what is the reason behind that?
MS. HERMAN: Well, I’m talking about the art world.
MS. REEVES: Okay. Do you see yourself in the art world for a long time?
MS. HERMAN: I like the idea of the craft and the artists. And the essence of craft is human, collaborative, and creative. And also the paint is human, the ink is human, the clay is human, the paint is human, the clay is human, and I think that’s one of the reasons that I like the craft. And I don’t know that I’ve ever done anything like that in my life. I just always felt that that was what I liked to do.

Michelle L. Herman: I made decisions about what was to be included in and omitted from the book. There is a curation aspect to it. I wanted to make sure that the interviews weren’t too similar.

I also curated the content. I spent a long time trying to get the system to actually generate interviews that went further. I noticed there was a gap in that I wasn’t seeing many (or any in some cases) discussions about gender, race, or disability. I felt this was really telling. So I actually kept trying to generate them – to make them appear. It was almost like I was trying to rewrite the history by manifesting these interviews, but it was really challenging, and even the ones that I did get were very surface level. In terms of inserting myself, there’s maybe two parts to that, where I went through all of the interviews and changed the names that the system generated. Sometimes I did that just by rearranging letters, or I found a new name that starts with the same letter. And I did that because I didn’t want to refer to any real people, living or not. So it’s possible that I may have started by changing a name that started with ‘M’ to my name, just out of convenience. But then there were places where I was thinking, ‘I’m going to insert myself in this history that doesn’t exist as a way to rewrite that narrative to make it more representative of someone like me: people with disabilities, women’.

Sarah McGavran: I’d love to talk a little bit more about the gaps or absences in the fictive interviews. Digital art is not discussed. And then there are absences in terms of who is represented. Disability is a major absence.

Michelle L. Herman: Yeah, digital art in the sort of grand narrative and history of art is relatively new, so I think that it just wasn’t in that initial data set of the ‘real’ oral history interviews that I used to fine-tune the system. Because of that, the results would then also have that lack of representation as well. As for the absence of disability: I don’t think I ever did find a single interview where disability was discussed, and I’m not sure the word ‘disability’ even comes up. And again, yes, it’s sort of pointing to that lack of representation in that initial data set.

Sarah McGavran: Another near absence is race, which comes up only once in a fictive interview called ‘The Right Racial Soil’. It seems to be discussed by a white critic in tokenising terms, in the context of a conference or another event. What is your takeaway on the following passage?

MS. DIPLOMI: Black writers and art historians from all over the country and the East Coast, and interracial something of convenience. So I liked that […] But then I was also interested in the fact that they were here.

Michelle L. Herman: Unfortunately, I can’t explain why it generated something like that. There’s a flippant tone in that passage, and that is something that might occur in that data set when more substantial issues are being discussed. And so maybe it’s trying to kind of emulate that, but it’s such a superficial kind of touching on this subject in a way that doesn’t totally make sense.

Sarah McGavran: Maybe some of us don’t have all the language that we need, or maybe there’s even a sense of discomfort in talking about race that results in that offensive tone.

Michelle L. Herman: Yeah, AI is almost like a mirror of the people who developed it and what the data set is.

Sarah McGavran: What do you think this technology got right? And what did it get wrong?

Michelle L. Herman: There’s something really interesting about the cadence and the rhythm of the different interviews […] Now, when I read a real interview, I think, ‘this sounds like an AI generated it’. So, it got the sentence structure right, the natural rhythm of conversation. I think it is super interesting that it was able to emulate this very organic thing.

I expected the results to be pretty male-centric, but I was actually surprised at the extent to which that was true. I didn’t realise how bad it was going to be.

Self-help for artists: How to fill those gaps of race and gender

Sarah McGavran: Let’s talk a little bit about how you selected and categorised these interviews: you move from more general art world tropes to issues of race and gender and then on to art and language. Why did you organise the book in this way?

Michelle L. Herman: When I was generating these batches, I would go through them and just essentially copy and paste the interviews that were my favourites into a Google Doc, and at some point, I just had so many that I decided, ‘I need to organise this’.

So I started categorising them loosely based on subject matter. I think at the beginning they were more general because that’s the corpus of data that it had. Once I saw that, I decided I wanted to find out what the gaps were and how to fill those gaps of race and gender, and that’s when I started trying to generate them. So they come in at the end because that’s where I was in the process.

At the beginning, I was thinking about it as a self-help book for navigating the art world, or a manual for ‘how to art’. That’s also part of the reason that the beginning is more general and talks about art school and stereotypes of what a typical artist is concerned with or does. I wanted to complicate it a little more towards the end.

The art and language at the end in some ways summarises what the whole thing is about. In a way, it’s questioning the limits of language – what isn’t being discussed.

And then: what is art? Is art supposed to communicate or not? That’s something I was really thinking about at the time: what is the purpose of art? I once really thought that art was supposed to communicate, and I think I’ve come to a different understanding about what I want my art to be. While there is a level of communication, there’s also an interplay with the viewer, who then takes from that and creates their own meaning. I also do love the art historical reference to the conceptual art collective Art and Language.10 That ties into the book well too.

Sarah McGavran: At the end, I started wondering, if art is supposed to communicate, then who does it communicate to? What training, background, or resources do you need to understand this obscure form of communication?

Michelle L. Herman: Something I always think about is this idea of ‘barriers to entry’. As a woman and a woman with disabilities, being neurodiverse, as well, I’m more cognisant of my limitations. I’m constantly second-guessing myself, and so the art speak and performance you have to do as an artist is something that I struggle with all the time. I’m interested in how the book pokes fun at these expectations and helps to highlight the absurdity of this sort of theatre and the way that artists are expected to perform.

Sarah McGavran: Some of these clichés made me think of the process of self-legitimisation for artists: going to art school, getting an MFA, learning art speak, etc. If all of these things are necessary, then again it seems like it’s an impossible barrier to get past for a lot of people, unless we start rethinking what it means to be an artist.

Michelle L. Herman: I was originally thinking about it as this self-help book because, as someone who works to support my practice and has to navigate this time balance, it’s sometimes frustrating to see these traditional depictions of what an artist is supposed to do and be. It’s not attainable. They clearly have somebody supporting them to get where they are, and that’s not a reality for everyone.

And so the self-help aspect is like ‘okay, if you’re not in that position, here’s how you can fake it till you make it’.

Humour and absurdity: It’s funny, but it’s not

Sarah McGavran: Let’s return to humour and absurdity, which play important roles in your practice, and especially in Let’s Talk Art. How does humour help address sensitive topics like bias and stereotypes?

Michelle L. Herman: In a lot of my work, humour comes through partially because that’s an aspect of my personality. But I also think that humour has this incredible way of making complex subjects more digestible. It lets you bring your guard down so that you can be open to seeing something in a different way. There’s a great book, Concrete Comedy by David Robbins, about comedy’s ability to challenge systems of power.11 So, I’m really interested in strategically employing comedy.

Sarah McGavran: Some of the interviews also have a dreamlike or even mystifying quality. One passage that really struck me seems to conflate being an artist and a mother with a disastrous car crash. And for me as an art historian, I was also reminded of Andy Warhol’s car crash series. How do you interpret the following passage?

MS. MESSER: Yeah. We had a car accident in New York. We were driving in Wilton […] and we had a front end collision […] And we were crossed in between two cars. And I’m running to the hospital. And they had a car that was going to Wilton with my baby. And they were going to do a car accident and I had a child in the car with the car. And I was on the side of the road, and I was in a hospital. And I was in a hospital and I was dying. And I wanted to do a car accident.

MR. FLIP: You have a son, and you’re suffering from one of your accidents.

Michelle L. Herman: 'Rather than making a painting, I wanted to make an accident.' I think that sort of absurdity was something that I was interested in. One of the reasons that I have been attracted to using AI systems like this is because it's sort of a modern-day version of the cut-up technique used by the Dadaists and Surrealists, where they would cut up magazine and newspaper articles and then rearrange the words to get this element of chance. That passage specifically really seems like it’s by artist Hugo Ball or writer and artist William Burroughs. So, I think that chance element is also a really interesting aspect. The super absurd juxtapositions maybe let you look at things in a different way. I really love your interpretation of how this disastrous car crash and the career of a woman artist are conflated with having a child. When you put these things together, those sorts of interpretations come up. I’m really excited about hearing what other people get from that.

The Dada precedent: The absurdity of technology

Sarah McGavran: The Dadas latched onto the absurd in response to the unprecedented destruction of World War I, which was the first war fought with modern technology. They rejected everything that had come up to it, including traditional artistic media and processes. Instead, they came up with different ways to generate art, as in the randomly selected poems that you mentioned.

Some Dadas, like Hans Arp, abhorred technology. How does this piece, which uses technology as a medium, work with and against the Dada precedent?

Michelle L. Herman: We’ve talked about how these technologies that are largely developed by white men are shaping our future. All the Facebook algorithms to keep you on there, everything that we touch nowadays is some kind of software that shapes our lives and our ability to move in space. One strong aspect for me is to critique that.

I am rejecting this technology by emphasising its problems and failures, and to argue that something needs to be done to change its trajectory.

So, I am interested in technology as both a medium and as a subject to critique.

Excerpt from ‘You Have to Be Very, Very Complex’
MS. PATRICK: Donald Trump.
MR. FINCH: Yes, he’s an illustrator.
MR. FINCH: And I think he has the same kind of sense of humor and creativity that I do. And I am very impressed by those things. But I think they’re very, very, very important when they’re shown, and I think they’re important in a sense. But I also think the art world has become so much more sophisticated and sophisticated, that the qualities of the artist are – I think – they’re much more important.

MS. PATRICK: And I think that’s true of many. You know, I don’t know, I’m just a very, very, very, very keen, very intelligent person. And I think that’s what you have to do. You have to be an artist, and as you’re going to be an artist you’re going to be very, very, very complicated. You have to be very, very, very, very, very, very, very, very, very, very complex.
MS. PATRICK: And I was very, very, very, very, very, very, very, very, very complex. But I think that’s what I did. And I think that’s what you have to do as well, and that’s what you have to do.

Sarah McGavran: There’s a lot of repetition in the fictive interviews. For example, there’s one called ‘You Have to Be Very, Very Complex’, where the interviewee repeats the word ‘very’ nine times. How does this repetition speak to the larger issues at hand?

Michelle L. Herman: There’s a couple of answers there: it is the system, essentially getting sort of hung up on a word and can’t determine the word that should come after it. That’s possibly because the data set also has some repetition. But I also love how the human brain and software are not that different. I know I get hung up on phrases or words that I’ll repeat in my head. I don’t know if it’s just me. I can’t stop thinking of a word until I figure out something else to fixate on.

But then, in the one that you referenced, I like that very emotional emphasis. I read it as, ‘very complex’, which plays into the kind of performance that is expected of artists. You can’t just be […] mainstream […] you have to be very complex. I’m interested in poking fun at this venerated performer-artist personality and exploring it.

Where do we go from here?

Sarah McGavran: What are some of the larger implications of Let’s Talk Art?

Michelle L. Herman: As an artist, I hope that the work inspires someone to explore technology in their work, particularly women and women-identifying people. How can they integrate that kind of material into their practice and, hopefully, lower that barrier to entry a little bit. Sometimes even for someone like me, who is working in technology, there are hurdles that you have to get over – I like to call them paywalls. It’s difficult to even make those connections in the industry, and so it’s my hope that the work would invite people to talk to me about it and inspire them.

Sarah McGavran: That also makes me think that a fundamental issue here is the stereotype that women aren’t good at science. One of the things that we need to do is expand that data set by bringing more women into these fields so that we don’t keep ending up with results like this.

Michelle L. Herman: The perception that women don’t like science and technology is something that I’ve certainly experienced. But then once you’re in it, there are all these factors that are essentially pushing you out. And so it feels like, ‘Okay, you’ve made it over one barrier, but there are like ten more coming at you’. It’s the culture itself, the attitudes, the way we treat people within that industry. There’s a toxic culture that I think needs to change too. So, beyond just bringing people in, we need to figure out how to keep them here, because women, women of colour, they’re getting pushed out. There was an article I read that said that half of all women in tech leave the tech industry at age 35.12 That’s when they just can’t take it anymore.

Sarah McGavran: Given the context of Culture Caleidoscoop, do you see any avenues for socially engaged, community-based projects using AI?

Michelle L. Herman: I think there are definitely a lot of applications, again, being cognisant of what the software does, is capable of, and what it represents. I’d love to see AI-generated biographies of women artists, inserting them in some way throughout art history or using that as a basis for an intervention on Wikipedia (another male-dominated technology where 87 per cent of editors are male).13

Postscript by the interviewer

Although I initially framed this interview to focus on Herman’s use of AI software to expose bias, our conversation taught me that Let’s Talk Art is about so much more than that. The work challenges us to think more broadly about how women, individuals with disabilities, people of colour, and members of other marginalised groups can insert themselves into these narratives and fields and make change.

I also hope this interview motivates others to elevate the voices of artists with different backgrounds, identities, and abilities so that, together, we can begin to correct the imbalance in the art world’s ‘data sets’ through expanded representation.


1 A more in-depth preview is available on the artist’s website: https://www.michellelisaherman.com/artwork/artist-book-lets-talk-art/.

2 The Guerilla Girls, an anonymous artists’ collective, uses statistics as the basis for tongue-in-cheek works of art and public demonstrations that shed light on art world inequities. For more information, see the Guerilla Girls website, accessed 31 July 2022, https://www.guerrillagirls.com.

3 There are numerous studies pertaining to various fields in which women are judged more harshly than men. For more information, see Ilana Yurkiewicz, ‘Study Shows Gender Bias in Science Is Real. Here’s Why It Matters’, Scientific American, 23 September 2012, https://blogs.scientificamerican.com/unofficial-prognosis/study-shows-gender-bias-in-science-is-real-heres-why-it-matters/; Paola Cecchi-Dimeglio, ‘How Gender Bias Corrupts Performance Reviews and What to Do about It’, Harvard Business Review, 12 April 2017, https://hbr.org/2017/04/howgender-bias-corrupts-performance-reviews-and-what-to-do-about-it; Joseph Shapiro, ‘Federal Report Says Women in Prison Receive Harsher Punishment than Men’, NPR, 26 February 2020, https://www.npr.org/2020/02/26/809269120/federal-report-says-women-in-prisonreceive-harsher-punishments-than-men.

4 Find more information about the research project in this presentation: Michelle L. Herman, ‘Wherewithal Research Presentation’, Michelle L. Herman, 12 July 2021, https://www.michellelisaherman.com/wherewithal-research-grant-presentation/.

5 Legacy Russell, Glitch Feminism: A Manifesto (New York City: Verso, 2020).

6 Lionel Sujay Vailshery, ‘Software Developer Gender Distribution Worldwide as of 2021’, Statistica, 23 February 2022, https://www.statista.com/statistics/1126823/worldwide-developergender/.

7 Gail Sullivan, ‘Google Statistics Show Silicon Valley Has a Gender Diversity Problem’, The Washington Post, 29 May 2014, https://www.washingtonpost.com/news/morning-mix/wp/2014/05/29/most-google-employees-are-white-men-where-are-allthewomen/.

8 Jeffrey Dastin, ‘Amazon Scraps Secret AI Recruiting Tool that Showed Bias against Women’, Reuters, 10 October 2018, https://www.reuters.com/article/us-amazon-com-jobs-automationinsight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G.

9 Tom Simonite, ‘When It Comes to Gorillas, Google Photos Remains Blind’, Wired, 11 January 2018, https://www.wired.com/story/when-it-comes-to-gorillas-google-photos-remainsblind/.

10 Read more about Art and Language, a conceptual artist collective founded in 1967, on the Tate website: https://www.tate.org.uk/art/art-terms/a/art-language.

11 David Robbins, Concrete Comedy: An Alternative History of Twentieth-Century Comedy (Copenhagen: Pork Salad Press, 2011).

12 Erin Carson, ‘Half of young women will leave their tech job by age 35, study finds’, CNET, 29 September 2020, https://www.cnet.com/tech/tech-industry/half-of-young-women-willleave-their-tech-job-by-age-35-study-finds.

13 ‘Community Insights/Community Insights 2020 Report’, Wikimedia Meta-Wiki, updated 6 April 2021. https://meta.wikimedia.org/wiki/Community_Insights/Community_Insights_2020_Report.


More Articles