<aside> âšī¸ This presentation was given at PAIRS 2026 in New Delhi on 18th February 2026 at 09:30 IST.
đī¸ Thread on PAIRS Discussion Server (Discord) (register first)
</aside>
Due to technical issues, a recording of this session is now available.
Online gender-based violence (GBV) and misogynistic hate speech are significant barriers to digital participation for women and gender minorities across South Asia, a region with an estimated 50 million female digital users at risk in India, Nepal, and Sri Lanka alone. Dominant AI-based content moderation systems, often developed without this context, consistently fail to capture the culturally-specific and intersectional nature of this harm. This leads to algorithmic failures that silence survivors and enable abuse, undermining digital equality and safety. This presentation is a case study of piloting VoiceValor, a "thick-big data" tool developed by Dignity in Difference that attempted to offer a practical, survivor-centric framework for algorithmic accountability. Designed with and for young South Asian feminist YouTube creators, the tool empowers survivors to actively identify and correct biases in online gender-harm indicators, enabling them to address algorithmic failures with their own cultural insights. During the pilot, over 10,000 users from LGBTQI+ and other marginalized groups have engaged with the platform. This participatory, survivor-led process has already identified over 5,5000+ distinct algorithmic failures in detecting culturally specific misogyny.
Beyond algorithmic correction, VoiceValor provides users with access to support resources and 20 domain-specific AI personas. These personas act as coaches, helping creators develop counter-narratives and strategies to navigate digital toxicity. As a case study serving communities in South Asia, VoiceValor presents a novel method for co-designing and refining AI systems by centering on the "thick data" of lived experience, specifically addressing algorithmic failures of classifiers based on big data. Our work aligned with the call for practical toolkits that embed community power into the AI lifecycle, offering critical, translational insights from the Global South on building a safer, more equitable digital future.