When Silence Speaks: Feminist AI Addressing Gender-Based Violence
This project aims to make online gender-based violence (GBV) visible – violence that is often hidden through silence and omission in digital environments. The goal is to develop a technical framework capable of detecting and analysing these “missing data” using advanced natural language processing and knowledge graphs.
Background
Online gender-based violence is often invisible. It is not limited to explicit abuse, but also includes silences – when survivors’ voices are erased, misogyny is normalised, and responsible institutions disappear from the narrative. These invisible patterns shape how violence is understood and contribute to the persistence of inequality.
Project goals / research focus
The project builds on a pilot study from spring 2025 that applied a critical perspective to different forms of structural “silences” in the development and deployment of AI tools, including marginalised voices, gender-based invisible labor, and the environmental impact of AI use.
The new research project focuses on further developing and testing the conceptual framework introduced in the pilot, with a specific emphasis on gender-based violence. Using advanced AI and natural language processing, the project develops tools that can detect not only hate and threats, but also the absence of key voices and perspectives. Knowledge graphs are used to map which actors and events are mentioned and which are missing – revealing structural erasures in digital narratives about violence.
The vision is to provide researchers, journalists, and communities with tools to expose and challenge the hidden patterns that sustain online gender-based violence.
Advancing gender equality
In collaboration with feminist organisations and survivor communities, the project develops new datasets and methods to assess how well AI systems can address these silences. The result is a new form of feminist AI that not only listens to what is said, but also to what is left unsaid – interpreting silence as a signal of structural injustice.
Researchers
The project is led by Amir Payberah, Docent at the Department of Computing and Learning Systems, KTH, and Lina Rahm, Docent at the Division of History of Science, Technology and Environment, KTH.
