2nd Snack: AI, Community Power, and Non-Extractive Technologies

The session featured two keynote speakers: Nadia Nadesan, a researcher at Platoniq, who addressed the relationship between AI and queer communities, and Eva Navarro, a professor of computer science and head of technology architecture at LaNuestra, who presented a humanistic and feminist perspective on technology architecture.

The event combined prepared questions, group discussions, and an open conversation about bias, representation, data extractivism, digital violence, and the possibilities for a more community-centered AI.

Queer in AI

Nadia proposed understanding the queer presence in AI through three lenses: political advocacy and community safety; artistic and cultural criticism; and the creation of data, archives, and visibility. She discussed networks such as Queer in AI, projects to detect transphobia developed by trans and non-binary communities, and works such as The Zizi Project, where flaws in drag deepfakes reveal the normative biases of the models. Her central question ran through the entire session: “How can AI infrastructures be accountable to queer and trans communities, rather than merely extracting information from them or categorizing them?”

The conversation then turned to AI biases in everyday life and in high-impact systems. Nadia explained that a seemingly minor error—such as a generative tool consistently assigning the masculine gender to an artist like Karol G—reveals deeper patterns of invisibilization. Eva expanded on this analysis with examples from medical imaging, automated recruitment, and music recommendations: when data on women, people of color, or vulnerable groups is lacking, systems produce discriminatory diagnoses, filters, and recommendations. “Unfortunately, biases exist. It’s not something we’ve already overcome just because we’re talking about it,” she noted.

Human Prompts to Understand How We Interact with Technological Tools

Cristian also prompted a reflection on our physical relationship with technology: not only what these tools contain, but how they make us move, express ourselves, and navigate the world. Nadia responded with a clear image: “Technology is a portal through which we access the world.” She cited Google Maps and maps inherited from imperialist perspectives as examples, reminding us that even the most everyday interfaces contain ways of viewing, organizing, and simplifying reality. AI, in that sense, is not just an external tool: it mediates the way we navigate the world, and we would do well to pause and ask whether that mediation is aligned with our bodies, experiences, and ways of life.

LaNuestra as Humanistic AI

Eva also presented LaNuestra, a digital community launched by Cristina Fallarás as an offshoot of the Cuéntalo movement, designed to collect, protect, and amplify testimonies of gender-based violence. The project is not intended to be a static archive, but rather an infrastructure for care, memory, and mutual support, offering therapeutic, legal, and community-based support. AI can help organize, anonymize, and categorize testimonies, but always in a supervised, ethical, and non-extractive manner. “This data is very real; it’s gold—it comes directly from women who have been subjected to violence and abuse,” Eva explained, emphasizing that official data captures only a small fraction of the violence that actually exists.

In her explanation, Eva advocated for a humanistic and feminist AI that doesn’t start by asking what can be designed, but rather who it serves, who decides, and what forms of community it enables us to build. LaNuestra aspires to become a catalyst: a tree with strong roots from which other democratic archives, technological learnings, and support networks can grow. That is why she even envisioned a future LaNuestra Academy, where young people and people of all ages can learn programming, data science, and AI from a solid foundation, without delegating knowledge to prompts or tools that are not fully understood.

The most critical part of the session focused on generative AI. Eva questioned its indiscriminate adoption: “Why haven’t we asked ourselves why we have to use it?” She pointed out problems such as training data contamination, errors, misinformation, plagiarism, data theft, the exploitation of moderators in the Global South, water and energy consumption in data centers, digital violence against women, and military applications. In the face of the fascination with large models, she advocated returning to basic questions of necessity, scale, and responsibility: we don’t always need to “use a sledgehammer to crack a nut.”

Tensions and Opportunities

Olivier closed the session by summarizing some of the main tensions. The group had identified risks related to algorithmic discrimination, lack of representation, the ecological footprint of AI, endless comparison, aesthetic pressure, self-esteem, the definition of “normal,” and the invisibilization of bodily, functional, cultural, and emotional diversities.

In light of these challenges, it was proposed that we continue working on specific dilemmas over the next fifteen days on the project’s platform. Among them: how to anonymize testimonies without diminishing their political impact; when AI should intervene in a vulnerable community; what data is worth collecting; who should control it; and how to prevent technology that promises to address violence from ultimately amplifying it.

The closing session left several dilemmas to be addressed further: what data to collect, under what community oversight, how to anonymize testimonies without diminishing their political impact, when AI should intervene in vulnerable communities, and how to prevent technology that promises to remedy violence from ultimately amplifying it. The shared conclusion was that the question is not simply whether AI is good or bad, but what kind of AI we want, who designs it, with what resources, based on what data, and to whom it must be accountable.

Dilemmas for Continuing to Work

“Connected Loneliness” Dilemma: https://openspaces.platoniq.net/processes/R1/f/540/debates/144

“Attention Economy” Dilemma: https://openspaces.platoniq.net/processes/R2/f/542/debates/145

“Digital Communities” Dilemma: https://openspaces.platoniq.net/processes/R3/f/543/debates/146

“Personal Self-Exploitation” Dilemma: https://openspaces.platoniq.net/processes/R4/f/544/debates/147

“Hate Speech” Dilemma: https://openspaces.platoniq.net/processes/R5/f/545/debates/148