Selected Papers of #AoIR2023: The 24th Annual Conference of the Association of Internet Researchers Philadelphia, PA, USA / 18-21 Oct 2023 Myles, D., Chartrand, A. & S. Duguay. (2023, October). ‘If We Look at It from an LGBT Point of View…’ Mobilizing LGBTQ+ Stakeholders To Queer Algorithmic Imaginaries. Paper presented at AoIR2023: The 24th Annual Conference of the Association of Internet Researchers. Philadelphia, PA, USA: AoIR. Retrieved from http://spir.aoir.org. ‘IF WE LOOK AT IT FROM AN LGBT POINT OF VIEW…’ MOBILIZING LGBTQ+ STAKEHOLDERS TO QUEER ALGORITHMIC IMAGINARIES David Myles Institut national de la recherche scientifique Alex Chartrand Concordia University Stefanie Duguay Concordia University Introduction In the past decade, platform algorithms, implemented to curate, exploit, and predict user activities online, have been criticized for exacerbating social inequalities that disproportionally affect women and people of color (Noble, 2018). Today, researchers are increasingly interested in understanding how platform algorithms also affect LGBTQ+ users by automating cisheteronormative biases (Southerton, 2021). This paper presents the results of an exploratory study that seeks to identify issues that platform algorithms raise for LGBTQ+ communities and analyze their social implications. Phase 1 relied on a scoping review of the scientific literature (Arksey & O’Malley, 2005) and on an extensive press review (Chartier, 2003) to map the main algorithmic controversies relating to LGBTQ+ communities (2010-2022). By using algorithmic functions as a sensitizing concept (Latzer et al., 2016), we classified our corpus of public controversies into five categories: sorting algorithms, recommendation algorithms, filtering algorithms, and search algorithms, as well as a fifth category on cisheteronormativity in the tech industry. Phase 2 relied on two group interviews: one with Canadian social media managers of LGBTQ+ non-profit organizations and one with Canada-based LGBTQ+ tech workers (n = 11). Importantly, phase 1 controversies were used as prompts to elicit discussions among participants. In this paper, we present the preliminary results of our Phase 2 group interviews (Phase 1 results are discussed in Myles et al. [2023]). We pay close attention to how participants made sense of algorithmic controversies and their social implications for LGBTQ+ communities by queering dominant algorithmic imaginaries. In this context, algorithmic imaginaries refer to the ways “in which people imagine, perceive and experience algorithms and what these imaginations make possible” (Bucher, 2017: 31), whereas the notion of queering refers to the act of subverting the cisheteronormative ideals these imaginaries often tend to normalize (Butler, 1993). Queering AI Temporality A first way participants queered algorithmic imaginaries was to question dominant discourses that typically depict AI technology as being inherently new, innovative, or optimistically disruptive (Borup et al., 2006). For example, when asked about the ability of facial recognition software to predict sexual identity, Lou stated: “Everything is wrong about this. […] Predicting people's sexual orientation from what they look like, that was debunked decades ago, that’s obviously not possible.” In this context, queering AI temporality refers to how participants refuted the belief that these algorithms (and their resulting controversies) are somehow new. Similarly, when discussing gender and sexual categories provided by platform algorithms, Dax mentioned: “Our trans and gender nonconforming communities have been obeying lots of other categories for decades. I think the real power struggle relates to who gets to set boundaries between categories.” Thus, a recurring trope in our group interviews was the re-inscription of algorithmic controversies within an LGBTQ+ temporality, one that sets them as part of a long-lasting history of gender and sexual oppression. Queering the Configured User Second, participants queered algorithmic imaginaries through the reconfiguration of the ideal-type user embedded in sociotechnical systems (Woolgar, 1990). For example, Judith said: “There is an idea that you can look homosexual or you can look female or male and I think it is easier for society to accept these ideas because it comes from a machine […] The fundamental issue is that the technology is built by humans who have a ton of biases […] The data that the machines have access to is, in of itself, limited to whatever the people have resources for, which is often the ability to access databases of mostly white people, white cis men specifically.” Thus, participants generally agreed that the tech industry is partly responsible for pre-configuring white, cisheterosexual users in digital platforms. However, a tension emerged when it came to identifying the best strategies to counter these biases in the tech industry. For example, Lou confided: “I have to do this work because if we, as queer people, don't speak up then, like, the people who had the idea to make these systems, they won't notice. […] If we don't say anything, it doesn't get addressed. But then, also, my mental health is impacted by having to speak up about these things every time, again and again and again.” As such, a conflict of allegiance emerged between the responsibility for LGBTQ+ tech workers to effect social change and the necessity for them to protect themselves against additional harm occurring in the process. Queering Algorithmic Resistance Third, participants queered algorithmic imaginaries by subverting the notion of ‘algorithmic resistance’ (Velkova & Kaun, 2021). Namely, they discussed whether true resistance could really occur by countering platform algorithms in practice. For example, Alexander shared the tactics they developed to counter the lack of content moderation on Facebook by constituting their own list of prohibited keywords. Yet, several participants were ambivalent towards using digital platforms professionally, as evidenced by Rafael: “It’s like a double-edged sword. Earlier, when I said I was for artificial intelligence in marketing, that’s if I could take advantage of it, if it wasn’t harmful. When you do LGBT comms, that’s kind of the fence you’re sitting on.” Other participants, like Dax, argued that discontinuing the use of proprietary platforms was a more effective strategy to resist oppressive algorithms: “In my experience, we end up using alternative circuits that do not go through these extensive algorithms. […] Just like 20, 30 years ago, we had flyers and word of mouth. Well, we also work through social capital, and I mean… our community’s algorithm, it also works really well.” This tension relates to broader academic discussions that alternately identify effective technological resistance as a form of strategic misuse (Jauréguiberry & Proulx, 2011) or, rather, as conscious technological disuse (Granjon, 2011). Queering Algorithmic Translation A final way participants queered algorithmic imaginaries was how they ‘translated’ (Akrich et al., 2006) algorithmic controversies in terms that made sense with their positionality as LGBTQ+ stakeholders. This entailed apprehending these controversies beyond purely technicist implications, extending them to political (Anna: “What if they start weaponizing [sorting algorithms] in different countries or use [them] to identify queer people?”) or socioeconomic (Rafael: “The issue is capitalism, that search engines end up working around notions of profit.”) considerations. Interestingly, Dax provided their own translation of the algorithmic functions used in our research design: “I’ve renamed these algorithms […]. If we look at it from an LGBT point of view, sorting algorithms could be called assignation algorithms, search engines could be called gatekeeping algorithms, filtering algorithms could be called policing algorithms, and recommendation algorithms could be called subjectivation algorithms.” This excerpt reveals the limitations of a technicist or functionalist framing of algorithmic controversies and the need to better account for LGBTQ+ imaginaries when analyzing the social implications of AI technologies. Beyond Algorithmic Imaginaries We conclude this paper by highlighting the importance of mobilizing key stakeholders, especially those from marginalized communities, to contest the dominant discourses through which society makes sense of AI technologies. Indeed, research on algorithmic justice should be firmly grounded in the preoccupations of these communities. Our study also points to the importance of developing research geared toward concrete actions to effect sociotechnical change. Beyond algorithmic imaginaries, research on algorithmic justice should advocate for better allyship from policymakers and tech representatives to alleviate the burden experienced by LGBTQ+ stakeholders. References Akrich, M., Callon, M., & Latour, B. (2006). Sociologie de la traduction: textes fondateurs. Presses des MINES. Arksey, H., & O'Malley, L. (2005). Scoping studies: towards a methodological framework. International journal of social research methodology, 8(1), 19-32. Borup, M., Brown, N., Konrad, K., & Van Lente, H. (2006). The sociology of expectations in science and technology. Technology analysis & strategic management, 18(3-4), 285-298. Bucher, T. (2017). The algorithmic imaginary: Exploring the ordinary affects of Facebook algorithms. Information, communication & society, 20(1), 30-44. Butler, J. (1993). Bodies That Matter: On the Discursive Limits of ‘Sex’. Routledge. Chartier, L. (2003). Mesurer l'insaisissable: méthode d'analyse du discours de presse. PUQ. Granjon, F. (2011). Fracture numérique. Communications, 88(1), 67-74. Jauréguiberry, F., & Proulx, S. (2011). Usages et enjeux des technologies de communication. Érès. Latzer, M., Hollnbuchner, K., Just, N., & Saurwein, F. (2016). The economics of algorithmic selection on the Internet. In Handbook on the Economics of the Internet (pp. 395-425). Edward Elgar Publishing. Myles, D., Duguay, S., & Flores Echaiz, L. (2023). Mapping the social implications of platform algorithms for LGBTQ+ communities. Journal of Digital Social Research, 5(4), 1-30. Noble, S. U. (2018). Algorithms of oppression. New York University Press. Southerton, C., Marshall, D., Aggleton, P., Rasmussen, M. L., & Cover, R. (2021). Restricted modes: Social media, content classification and LGBTQ sexual citizenship. new media & society, 23(5), 920-938. Velkova, J., & Kaun, A. (2021). Algorithmic resistance: Media practices and the politics of repair. Information, Communication & Society, 24(4), 523-540. Woolgar, S. (1990). Configuring the user: the case of usability trials. The Sociological Review, 38(1), 58-99.