IGF 2025 Lightning Talk #136 Inclusive LLM design and development with open data & gender

    Internet Society Gender Standing Group
    Umut Pajaro Velasquez - Internet Society Gender Standing Group - GRULAC
    Speakers
    Umut Pajaro Velasquez - Internet Society Gender Standing Group - GRULAC
    Onsite Moderator
    Umut Pajaro Velasquez - Internet Society Gender Standing Group - GRULAC
    Rapporteur
    Umut Pajaro Velasquez - Internet Society Gender Standing Group - GRULAC
    SDGs
    5. Gender Equality
    10. Reduced Inequalities
    16. Peace, Justice and Strong Institutions


    Targets: The proposal connects to SDGs 5, 10, and 16 in the following ways: SDG 5: Gender Equality :Focus on gender lens: The proposal emphasizes exploring open data and collaborative construction with a gender lens. This directly addresses the need to consider gender biases when building AI models, which can perpetuate existing inequalities (SDG 5.1). SDG 10: Reduced Inequalities: Mitigating bias in LLMs: By using open data and including diverse perspectives, the proposal aims to create less biased deep language models (LLMs). This can help reduce inequalities in areas like job opportunities and access to information, which can be affected by biased AI (SDG 10.2, 10.3). SDG 16: Peace, Justice and Strong Institutions: Transparency and accessibility: The proposal highlights the importance of open data, making the construction process of LLMs more transparent and fostering trust (SDG 16.10). Collaborative construction: The collaborative approach promotes inclusive decision-making and strengthens institutions by involving diverse voices (SSDG 16.7).
    Format
    Roundtable
    Duration (minutes)
    30
    Description
    We will explore the synergy between open data and collaborative construction with a gender lens, unraveling how this combination becomes a fundamental catalyst for the creation of more equitable and bias-free deep language models. We will take a close look at how the transparency and accessibility of open data, along with the active inclusion of gender perspectives in its construction, contribute significantly to mitigating inherent biases in LLMs. Likewise, we will see how the use of practical strategies demonstrate how this approach not only addresses crucial challenges, but also drives innovation towards a future where NLPs and LLMs more accurately and fairly reflect the diversity of our society. In conclusion, we propose a path towards collaboratively building a more inclusive and equitable technological future.

    The session intends to organize a Lightning Talk where participants and speakers can engage in discussions centered around a specific topic. During the session, participants can ask questions, propose interventions, and take part in the Q&A segment that will have some online elements. The moderator will facilitate the discussion by guiding the speakers' responses and managing the floor. The agenda includes an introduction and background, policy issues discussion, open floor for participant comments, and a conclusion with key takeaways. Moderators will ensure that questions and comments are addressed effectively, and that communication flows smoothly.