Civics of Technology Announcements
Next Book Club: Hyperscale: The Ambition and Excess of Big Tech’s Data Empires by Paris Marx on Tuesday, December 15 at 8:00 PM Eastern. Register here.
Dr. Bruna Damiana Heinsfeld is an Assistant Professor at the University of Minnesota, Twin Cities (UMN). Her research is rooted in the Critical Studies of Education and Technology (CSET) framework, focusing on critically examining the complex intersections of technology, society, and education. Her work seeks to illuminate the ideological, cultural, and economic forces influencing these dynamics, fostering a deeper understanding of their implications for education.
I write this as a Brazilian woman and an assistant professor in the United States, where I have lived for eight and a half years. My upbringing in Brazil continues to shape how I understand education, politics, and collective life. Following the Brazilian election from abroad has also made a connection particularly visible: the platforms through which Brazilians communicate and organize are owned by corporations headquartered in the country where I now live, conduct research, and teach.
On October 25, 2026, Brazil elected Luiz Inácio Lula da Silva to a fourth presidential term, reversing Senator Flávio Bolsonaro’s first-round lead (Paraguassu et al., 2026). The result concluded a campaign marked by disputes over the conditions of political communication. After the election, one question remains: What happens to democracy when a private company controls who gets heard and nobody can hold it accountable?
Elections depend on opportunities to encounter arguments, express commitments, and organize collectively. These activities increasingly unfold within applications that also carry family conversations, work arrangements, commercial exchanges, and cultural life. Political communication emerges through these overlapping relationships: a neighborhood group can organize transportation to vote; an entertainment account can become a space for political debate.
In Brazil, WhatsApp occupies an essential place within this landscape. A 2025 survey found it installed on 99% of respondents’ smartphones, alongside Instagram on 91% (Paiva, 2025). Both applications are owned by Meta, which also owns Facebook. This concentration gives one corporation authority over multiple routes to public participation. Opening an account elsewhere offers limited relief when the community one needs to reach remains within Meta’s systems.
Democracy, sovereignty, and corporate power
Lula, a former metalworker and union leader associated with the Workers’ Party (PT), governed from 2003 through 2010 and returned to office in 2023. His opponent, Liberal Party (PL) senator Flávio Bolsonaro, largely continued the conservative political agenda of his father, former president Jair Bolsonaro (Hughes, 2026), whose conviction for plotting a coup after his 2022 electoral defeat formed part of the election’s political backdrop (Brito et al., 2025).
Within this context, restrictions on political communication acquired significance extending beyond individual accounts. Internal Meta communications acknowledged problems affecting Lula’s accounts and inappropriate recommendations involving Flávio Bolsonaro (Mello, 2026). Further reporting documented difficulties finding Lula’s profile, displaying comments, and sharing videos. Some incidents affected Lula specifically; others occurred within broader platform failures (T. Dias & Motoryn, 2026). Ordinary users were affected too. Reported cases included a pro-Lula WhatsApp community suspended under a ‘terrorism-related’ classification (Oliveira, 2026).
These incidents fed a narrative that Meta was acting as a political instrument favoring Bolsonaro (Arnaudo, 2026). The narrative exposes a consequence of corporate opacity: when a foreign private entity controls influential channels of communication, unexplained asymmetries create grounds for suspicion that the public lacks the evidence to resolve.
Student organizations faced restrictions while participating in a nationwide mobilization involving assemblies, demonstrations, and voter outreach (União Nacional dos Estudantes, 2026). Four university organizations in Rio Grande do Sul had their Instagram accounts suspended after political posts and calls to assemble, according to prosecutors. A federal court ordered their restoration (Romagna & Rodrigues, 2026). The streets offered a space for political presence, but assembling there also depended on communication networks. What happens to the right to organize when the channels used to exercise it become inaccessible through decisions participants cannot meaningfully challenge?
Whoseolitics enter the AI space?
In Brazil, each presidential candidate is identified by a number that voters enter into an electronic voting machine: 13 for Lula and 22 for Flávio Bolsonaro. These numbers function as campaign symbols, appearing on shirts, posters, and other materials. During the runoff, users found that Meta AI treated them differently. When asked to produce a design supporting Lula with the number 13, the chatbot recognized the request as electoral content and refused to generate it. Yet equivalent requests supporting Flávio Bolsonaro with the number 22 produced campaign images (Reis, 2026a). The restriction invoked to block support for one candidate was not consistently applied to his opponent (Spagnulo, 2026).
The asymmetry extended beyond explicitly partisan requests. When journalists asked Meta AI to depict fictional animals as candidates, without mentioning either politician, a party, or an electoral number, the system automatically added 22 to the images. It also generated other numbers, but never 13 (Reis, 2026b). A politically meaningful symbol was therefore introduced into images whose prompts had not requested it. Why could the system recognize Lula’s number as grounds for refusing an image while allowing (and sometimes supplying) his opponent’s number?
What political associations enter an apparently generic request to depict a ‘candidate’? Whose assumptions become visible when a system supplies an electoral number the user never requested?
The uncertainty about how these outputs arose is central to the problem. Users encounter politically consequential differences but cannot inspect the processes producing them. Whether those differences were intended does not determine whether the institution responsible should answer for them.
Meta denied favoritism and attributed the incidents to technical problems and safeguards (Meta, 2026). Its Oversight Board nevertheless criticized recurring election-integrity failures, including excessive enforcement against Lula’s advertising accounts (Oversight Board, 2026). Calling a failure ‘technical’ does not settle its political significance. Even if nobody deliberately sought to influence the election, Meta designed and deployed the systems, determined the resources available to monitor them, and established the procedures through which affected people could seek redress. Unpredictability cannot absolve a company of responsibility for introducing systems into crucial social processes. The framing of harm as an incidental malfunction also risks obscuring the structural arrangements that produce it (Broussard, 2023).
During an election, time is part of the harm: restoring access after an opportunity has passed cannot necessarily restore that opportunity. Accountability becomes harder when the corporation whose conduct is disputed also holds much of the evidence needed to assess it (Arnaudo, 2026).
Automated systems can thus function as accountability sinks: users are directed toward filters, models, and technical failures, while responsibility becomes difficult to locate and enforce. Opacity serves an institutional function when it allows decisions to have material consequences without providing an effective route to challenge them (Heinsfeld, 2025). The company need not intend each outcome to benefit from an arrangement that makes those outcomes difficult to scrutinize.
The labor of being heard
People sought ways to ‘dribble the algorithm’: coordinating engagement, circulating material through personal networks, and combining digital outreach with organizing in the streets. The campaign’s final weeks included expanded efforts to mobilize people who had skipped the first round (M. Dias, 2026).
This ingenuity also reveals a burden. Testing visibility, documenting restrictions, rebuilding audiences, and finding alternative routes of communication demand labor from people attempting to participate. Electoral victory does not retroactively make those conditions equitable.
For US readers, the broader concerns should sound familiar. Facebook’s role in the 2016 election came under scrutiny through Cambridge Analytica (Lapowsky, 2018). The company deceptively harvested Facebook users’ information for voter profiling and targeting (Federal Trade Commission, 2019). A separate investigation documented Russia’s use of social media to influence the election (Senate Select Committee on Intelligence, 2019). These cases involve different mechanisms and do not establish that Facebook determined either country’s electoral outcome. They do, however, highlight a recurring problem: democratic communication depends on privately governed systems over which the public exercises limited authority.
What critical technology education demands
The educational implications extend beyond elections. When AI generates a ‘student,’ a ‘competent professional,’ or a ‘trustworthy source,’ which identities and forms of knowledge appear? Which are distorted, omitted, or refused? What goes unquestioned when an output matches expectations we already hold?
The Brazilian examples offer a starting point for examining both representation and access: whose request is accepted, whose expression is restricted, and whose complaint receives attention. They also raise institutional questions. What happens when a school or university adopts systems whose classifications affected people cannot contest?
This is what is at stake in the work of Civics of Technology: our capacity to recognize and confront technological power as it enters the conditions of collective life. A corporation does not have to act directly or knowingly against a political movement for its systems to undermine democratic participation. It can produce that harm through the deployment of poorly understood systems, inadequate safeguards, and ineffective mechanisms of redress. Critical technology education must confront the institutional arrangements that allow such consequences to persist without meaningful accountability. Students and educators need opportunities to ask how such authority was acquired, whose interests it serves, and how it can be challenged. Otherwise, we risk educating people to adapt to conditions they should have the power to transform.
That transformation requires asking what communication infrastructure governed through more accountable institutions could look like. Demanding better explanations and remedies from big tech matters, but responsiveness granted at a private company’s discretion remains precarious. What forms of public or community governance could give participants enforceable rights and a meaningful role in determining the rules? How could such institutions protect expression while remaining answerable to the people whose communication they organize? These are questions of democratic design that education can help place within collective reach.
Brazil’s election leaves us with an urgent question: how much of our democracy are we willing to place at the discretion of systems whose owners answer to us only when compelled? No public should have to outsmart a corporation to exercise its political voice.
References
Arnaudo, D. (2026, October 23). In Brazil’s runoff, Lula faces Bolsonaro, Trump—and big tech. Tech Policy Press. https://www.techpolicy.press/in-brazils-runoff-lula-faces-bolsonaro-trump-and-big-tech/
Brito, R., Magalhaes, L. N., & Andreoni, M. (2025, September 12). Brazil’s Bolsonaro sentenced to 27 years after landmark coup plot conviction. Reuters. https://www.reuters.com/world/americas/brazils-bolsonaro-sentenced-27-years-after-landmark-coup-plot-conviction-2025-09-12/
Broussard, M. (2023). More than a glitch: Confronting race, gender, and ability bias in tech. MIT Press. https://mitpress.mit.edu/9780262047654/more-than-a-glitch/
Dias, M. (2026, October 26). Takeaways from Lula’s win in Brazil: Polarization and scrutiny of digital platforms. The Washington Post. https://www.washingtonpost.com/world/2026/10/26/takeaways-lulas-win-brazil-polarization-scrutiny-digital-platforms/
Dias, T., & Motoryn, P. (2026, October 23). Novos documentos internos da Meta registram falhas no perfil de Lula em stories, comentários e busca na reta final da campanha. CTRL+Z. https://newsletter.ctrlz.org.br/p/novos-documentos-internos-da-meta
Federal Trade Commission. (2019, December 6). FTC issues opinion and order against Cambridge Analytica for deceiving consumers about the collection of Facebook data, compliance with EU-U.S. Privacy Shield [Press release]. https://www.ftc.gov/news-events/news/press-releases/2019/12/ftc-issues-opinion-order-against-cambridge-analytica-deceiving-consumers-about-collection-facebook
Heinsfeld, B. D. (2025, April 20). Is the author dead? Addressing bias and accountability in educational technology [Paper presentation]. American Educational Research Association Annual Meeting, Denver, CO, United States. https://doi.org/10.3102/2181633
Hughes, E. (2026, October 24). What to know as Brazilians choose between Lula and Flávio Bolsonaro on Sunday. Associated Press. https://apnews.com/article/brazil-election-lula-bolsonaro-0ab5c798e84a3f3a3a5eb659a5ca0fac
Lapowsky, I. (2018, March 17). Cambridge Analytica took 50M Facebook users’ data—and both companies owe answers. WIRED. https://www.wired.com/story/cambridge-analytica-50m-facebook-users-data/
Mello, P. C. (2026, October 16). Em mensagens internas, Meta admite que contas de Lula foram prejudicadas por restrições; engenheiros falam em falhas técnicas. Folha de S.Paulo. https://www1.folha.uol.com.br/poder/2026/10/em-mensagens-internas-meta-admite-que-contas-de-lula-foram-prejudicadas-por-restricoes-engenheiros-falam-em-falhas-tecnicas.shtml
Meta. (2026, October 18). A Meta não está interferindo nas eleições brasileiras. Estes são os fatos. https://about.fb.com/br/news/2026/10/a-meta-nao-esta-interferindo-nas-eleicoes-brasileiras-estes-sao-os-fatos/
Oliveira, M. (2026, October 16). Meta bloqueia conteúdos políticos pró-Lula e contra Flávio acusando terrorismo e nudez. Agência Pública. https://apublica.org/2026/10/meta-bloqueia-conteudos-politicos-pro-lula-e-contra-flavio-acusando-terrorismo-e-nudez/
Oversight Board. (2026, October 20). Oversight Board statement on protecting election integrity on Meta’s platforms. https://www.oversightboard.com/news/oversight-board-statement-on-protecting-election-integrity-on-metas-platforms/
Paiva, F. (2025, February 27). Telegram: Penetração no Brasil cai 6 pontos percentuais em um ano. Mobile Time. https://www.mobiletime.com.br/noticias/27/02/2025/telegram/
Paraguassu, L., O’Boyle, B., & Magalhaes, L. N. (2026, October 25). Brazil’s Lula wins historic fourth term, Flavio Bolsonaro refuses to concede. Reuters. https://www.reuters.com/world/americas/brazil-heads-polls-runoff-between-leftist-lula-bolsonaro-jr-2026-10-25/
Reis, G. (2026a, October 14). Inteligência artificial da Meta se recusa a criar artes de campanha de Lula e libera as de Flávio; entenda. Correio do Povo. https://www.correiodopovo.com.br/not%C3%ADcias/pol%C3%ADtica/elei%C3%A7%C3%B5es/inteligencia-artificial-da-meta-se-recusa-a-criar-artes-de-campanha-de-lula-e-libera-as-de-flavio-entenda-1.1756034
Reis, G. (2026b, October 16). IA da Meta gera imagens de candidatos fictícios com o número do PL. Correio do Povo. https://www.correiodopovo.com.br/not%C3%ADcias/pol%C3%ADtica/elei%C3%A7%C3%B5es/ia-da-meta-gera-imagens-de-candidatos-ficticios-com-o-numero-do-pl-1.1756636
Romagna, D., & Rodrigues, É. (2026, October 17). Justiça manda Meta restabelecer contas de movimentos estudantis do RS bloqueadas no Instagram. GZH. https://gauchazh.clicrbs.com.br/politica/eleicoes/noticia/2026/10/justica-manda-meta-restabelecer-contas-de-movimentos-estudantis-do-rs-bloqueadas-no-instagram-cmvcy5az802f6014l3vzb7btf.html
Senate Select Committee on Intelligence. (2019, October 8). Senate Intel Committee releases bipartisan report on Russia’s use of social media [Press release]. https://www.intelligence.senate.gov/2019/10/08/press-senate-intel-committee-releases-bipartisan-report-russia-e2-80-99s-use-social-media/
Spagnulo, S. (2026, October 14). IA da Meta favorece criação de imagens de Flávio Bolsonaro. Aos Fatos. https://www.aosfatos.org/noticias/ia-da-meta-favorece-criacao-de-imagens-de-flavio-bolsonaro/
União Nacional dos Estudantes. (2026, October 6). UNE lança campanha “Estudantes com Lula” para derrotar a extrema direita. https://www.une.org.br/noticias/une-lanca-campan