The Platform Question: Power, Accountability and Global South The Platform Question: Power, Accountability and Global South Reimagining AI from the Ground Up Our project embraced this politics of imagination. Rooted in feminist and design justice frameworks, we facilitated workshops to help people create alternate ‘data points’ about themselves using art and storytelling as tools to reimagine the very foundations of AI, its datasets. At Hyderabad’s Digital Citizen Summit 2024, we worked with tech makers, engineers, students, and professors. To reimagine what datasets could look like, we first needed to understand what the participants knew of the current AI landscape and its impact on human realities. This framing helped participants recognise AI not only as a future-forward innovation but also as an embedded decision-making infrastructure shaping the representational landscape of the web. One participant shared a deeply personal story. He had worked on an AI-driven medical tool designed to predict emergencies. But the training data lacked female-specific health indicators, a gap that only became clear when a close acquaintance was diagnosed with breast cancer. The absence of women in the design team meant no one had flagged critical missing data. 128 One participant shared a deeply personal story. He had worked on an AI-driven medical tool designed to predict emergencies. But the training data lacked female-specific health indicators, a gap that only became clear when a close acquaintance was diagnosed with breast cancer. The absence of women in the design team meant no one had flagged critical missing data. This powerful moment reiterated our point that missing data reflects missing people, missing perspectives, and ultimately, missed possibilities. At the AWID Forum 2024, in Bangkok, conversations widened further. Here, our participants included artists, activists, human rights defenders, sex workers’ rights groups, policymakers, and digital rights advocates. Despite the geographic and professional diversity, one theme persisted: most participants still associated AI primarily with generative tools. This unfamiliarity opened up space for reflection around What would your data look like, feel like, sound like? What would it resist being called? Participants began by searching everyday prompts beautiful bodies, breakfast, expression, and reflecting on whether the search results represented their realities. Queer participants, for instance, had to enter multiple, highly specific prompts inclusive, feminist, non-binary bodies, before seeing representations of themselves. People of color and disabled participants encountered similar erasures. Especially among participants from non-Western contexts, it revealed the dissonance between lived realities and the worlds The Missing Blocks of AI rendered legible by AI systems. Once again, it became clear dominant representations online reflect the priorities of those with the most access to and control over data infrastructures. Similar to the absence of a kitchen from Lucknow, this is a design flaw and a failure of imagination as well as a refusal to make space for the multiplicity of lives that fall outside the hegemonic dataset. The erasure became even starker when participants translated prompts into their native languages Burmese, Amharic, Hindi, and Thai. The results often returned nothing relevant, if anything at all. This underscored an uncomfortable truth, the language of the internet and by extension, AI is English. Anything outside it becomes invisible, unindexable, unintelligible to the machine. In response, participants reimagined these datasets by generating their own. Using categories like breakfast, bodies, and home, they transformed the generic sticky notes into artifacts of resistance, using basic tools such as pens, colored pens, watercolors, and embroidery threads. Under the breakfast sub-prompt “coffee,” participants added bunna and jebena, Ethiopian terms for traditional coffee, alongside espresso and cappuccino. Under meals, they introduced foods like papad (a crispy lentil flatbread), masko daal (black lentils), aloo gobi (potato and cauliflower stir-fry), and rabari (a sweet dish from India), pushing back against Eurocentric culinary defaults. The reimaginings extended to bodies. Participants from Thailand, India, Singapore, Ethiopia, Myanmar, the US, UK, and across Africa reflected, shared, deferred, and held space for every person’s stories. They brought rich, layered insights into how AI is experienced, resisted, and reimagined. They drew trans bodies, faces with African locs, annotated in Thai, Burmese, Mandarin, Hindi. These were acts of creativity, interventions, care and resistance. Each drawing, annotation, and translation carved out space for new representations, recognitions, and the lived reality of those traditionally excluded. When structural injustice is coded into AI systems, the solution then becomes to integrate feminist AI principles of care, context, and cocreation, and question who gets to define what counts as data, who classifies it, who benefits from it, and who annotates it. So that when a young girl from a small town of the Global South, wants to tell her story through images of her kitchen, and her dishes, she can do so and see her reality in the vast, vast landscape that AI is. The Missing Blocks of AI Dominant representations online reflect the priorities of those with the most access to and control over data infrastructures. 129