Session
Organizer 1: Civil Society, Latin American and Caribbean Group (GRULAC)
Organizer 2: Civil Society, Latin American and Caribbean Group (GRULAC)
Organizer 3: Civil Society, Latin American and Caribbean Group (GRULAC)
Organizer 4: Civil Society, Latin American and Caribbean Group (GRULAC)
Organizer 2: Civil Society, Latin American and Caribbean Group (GRULAC)
Organizer 3: Civil Society, Latin American and Caribbean Group (GRULAC)
Organizer 4: Civil Society, Latin American and Caribbean Group (GRULAC)
Speaker 1: André Fernandes, Civil Society, Latin American and Caribbean Group (GRULAC)
Speaker 2: Umut Pajaro Velasquez, Civil Society, Latin American and Caribbean Group (GRULAC)
Speaker 3: Jalal Abukhater, Civil Society, Asia-Pacific Group
Speaker 4: Yee Man Ko , Technical Community, Asia-Pacific Group
Speaker 5: Luana Araújo, Civil Society, Latin American and Caribbean Group (GRULAC)
Speaker 2: Umut Pajaro Velasquez, Civil Society, Latin American and Caribbean Group (GRULAC)
Speaker 3: Jalal Abukhater, Civil Society, Asia-Pacific Group
Speaker 4: Yee Man Ko , Technical Community, Asia-Pacific Group
Speaker 5: Luana Araújo, Civil Society, Latin American and Caribbean Group (GRULAC)
Format
Roundtable
Duration (minutes): 60
Format description: The roundtable format is ideal for this session, as it fosters direct exchange between experts, affected communities, and the audience, ensuring an inclusive and horizontal debate. The duration allows the debate to go beyond surface-level comments and delve into deep analyses of the technical, political, and cultural dimensions involved. Initially, concrete cases will be presented to demonstrate how AI can erase or distort cultural expressions from the Global South, providing a real basis for the discussion. Then, the extended time facilitates an interactive and critical dialogue among experts, activists, and representatives from affected communities, enabling the exchange of experiences and the confrontation of different perspectives. Finally, this duration supports the collective development of proposals and recommendations, transforming the debate into concrete actions to mitigate epistemicide. In this way, 60 minutes are indispensable for promoting an inclusive, in-depth discussion with the potential to generate effective impacts.
Duration (minutes): 60
Format description: The roundtable format is ideal for this session, as it fosters direct exchange between experts, affected communities, and the audience, ensuring an inclusive and horizontal debate. The duration allows the debate to go beyond surface-level comments and delve into deep analyses of the technical, political, and cultural dimensions involved. Initially, concrete cases will be presented to demonstrate how AI can erase or distort cultural expressions from the Global South, providing a real basis for the discussion. Then, the extended time facilitates an interactive and critical dialogue among experts, activists, and representatives from affected communities, enabling the exchange of experiences and the confrontation of different perspectives. Finally, this duration supports the collective development of proposals and recommendations, transforming the debate into concrete actions to mitigate epistemicide. In this way, 60 minutes are indispensable for promoting an inclusive, in-depth discussion with the potential to generate effective impacts.
Policy Question(s)
(A) How can AI systems be designed to actively prevent epistemicide and foster genuine epistemic diversity, ensuring fair representation of Global South knowledge systems?
(B) How can regulatory frameworks and public policies effectively mitigate the ongoing cultural erasure caused by global AI models, ensuring inclusive and diverse data governance?
(C) What ethical, institutional, and legal responsibilities should major tech companies uphold in preserving, promoting, and integrating the cultural and linguistic diversity of the Global South?
What will participants gain from attending this session? Participants will be able to reflect on how AI systems reinforce colonial dynamics of knowledge production, leading to the erasure of cultures, languages, and epistemologies from the majority world. Additionally, they will engage with practical examples from two distinct regions of the Global South that illustrate these dynamics, such as AI models’ failure to accurately recognize and reproduce the cultural expressions of Brazil’s Northeast and black and marginalized communities in Cartagena, Colombia; They will also reflect on the ethical and political implications of this phenomenon. Finally, through exchanges with experts and affected communities, participants will explore strategies to challenge AI epistemicide, including regulatory frameworks, advocacy, and community-driven data initiatives.
SDGs
Description:
"Epistemicide" was a term coined by Boaventura de Souza Santos to refer to the suppression and inferiorization of local knowledge and an entire cultural diversity in favor of a single epistemological model. The Brazilian theorist Sueli Carneiro adds that this serves as a mechanism for the burial of knowledge, especially that of dominated peoples. Djamila Ribeiro further states that this process constitutes a violation of human rights, limiting the available perspectives. In Artificial Intelligence systems, this has become an alarming phenomenon, particularly in the Global South, where cultures, languages, and epistemologies are systematically erased or distorted by AI models. This workshop will explore how these systems reinforce epistemic inequalities by prioritizing data, knowledge, and values from the Global North while marginalizing or misrepresenting cultural expressions and knowledge systems from historically oppressed communities. We take as a starting point an observation on how language models behave when reproducing expressions such as cordel—a traditional form of Brazilian Northeastern literature—and Champeta—a genre and culture from Black and marginalized communities in Cartagena, which is often replaced by a generic image of Black Caribbean music. Based on this, we will discuss how the exclusion of specific cultural expressions is not merely a matter of data training but a structural consequence of AI design, data collection, and validation within a Eurocentric and market-driven logic. The workshop will bring together AI experts, digital rights activists, scholars, and representatives from affected communities to explore alternative approaches that break away from this erasure and promote epistemic diversity in AI systems. The session will be interactive, combining case presentations, roundtable discussions, and participant contributions. Our goal is to foster a critical debate that will serve as input in a report on how policies and regulations can mitigate digital epistemicide, ensuring that technological innovation respects and strengthens global cultural diversity.
"Epistemicide" was a term coined by Boaventura de Souza Santos to refer to the suppression and inferiorization of local knowledge and an entire cultural diversity in favor of a single epistemological model. The Brazilian theorist Sueli Carneiro adds that this serves as a mechanism for the burial of knowledge, especially that of dominated peoples. Djamila Ribeiro further states that this process constitutes a violation of human rights, limiting the available perspectives. In Artificial Intelligence systems, this has become an alarming phenomenon, particularly in the Global South, where cultures, languages, and epistemologies are systematically erased or distorted by AI models. This workshop will explore how these systems reinforce epistemic inequalities by prioritizing data, knowledge, and values from the Global North while marginalizing or misrepresenting cultural expressions and knowledge systems from historically oppressed communities. We take as a starting point an observation on how language models behave when reproducing expressions such as cordel—a traditional form of Brazilian Northeastern literature—and Champeta—a genre and culture from Black and marginalized communities in Cartagena, which is often replaced by a generic image of Black Caribbean music. Based on this, we will discuss how the exclusion of specific cultural expressions is not merely a matter of data training but a structural consequence of AI design, data collection, and validation within a Eurocentric and market-driven logic. The workshop will bring together AI experts, digital rights activists, scholars, and representatives from affected communities to explore alternative approaches that break away from this erasure and promote epistemic diversity in AI systems. The session will be interactive, combining case presentations, roundtable discussions, and participant contributions. Our goal is to foster a critical debate that will serve as input in a report on how policies and regulations can mitigate digital epistemicide, ensuring that technological innovation respects and strengthens global cultural diversity.
Expected Outcomes
(1) Increase awareness among policymakers, researchers, and AI developers about the existence and impacts of AI epistemicide. Develop a series of educational materials for social media dissemination explaining what epistemicide is and how it relates to AI-based technologies.
(2) Develop recommendations that provide input for public policies that ensure cultural and epistemic representation in AI systems. These recommendations will be further developed in a policy brief that will be distributed more specifically to government entities, and will be translated into Portuguese and Spanish.
(3) Connect civil society organizations, academia, the private sector, and the public sector to collaborate on addressing the challenge of AI reproducing colonial dynamics. Based on the set of contributions, hold outreach lectures on the topic, targeting different sectors and stakeholders.
(4) Produce a report that gathers the key reflections and proposals developed during the workshop.
Hybrid Format: To ensure an inclusive experience for both in-person and online participants, we will implement multiple engagement strategies. A skilled moderator will actively balance contributions from both audiences, alternating between in-person questions — collected via cards and live microphones — and virtual inputs from the chat. Additionally, we will use interactive tools such as Mentimeter for live polling and QR-code surveys, allowing participants to share their views dynamically. A collaborative digital board, via platforms like Miro, will enable attendees to contribute references, personal experiences, and solutions in real time, ensuring the discussion’s impact extends beyond the session. This hybrid approach fosters meaningful interaction, making both onsite and online participants feel equally involved.