Culturally Grounded and Trustworthy NLP: Perspectives from Islamic Media Ethics

Event at EMNLP 2026

📅 When: October 25, 2026
⏰ Time: 16:00 - 17:30 (Budapest local time)
📍 Location: Room G9-10, HUNGEXPO Budapest Congress & Exhibition Center

About the event

What makes an AI-generated answer trustworthy when cultural, religious, and ethical perspectives differ? This event brings together researchers and scholars to examine how AI systems represent knowledge, navigate cultural norms and personal preferences, and serve diverse communities across languages and modalities.

Drawing on Islamic media ethics and research on culturally grounded NLP, the presentations connect questions of truthfulness, verification, authority, and accountability with practical challenges in cultural representation and spoken and visual question answering. Together, they explore how AI development and evaluation can account for reliable evidence, differing interpretations, and people’s everyday experiences.

Discussion topics

  • Trustworthy Islamic knowledge systems: Retrieval, verification, and tool-assisted reasoning for reliable question answering, with traceable sources, appropriate uncertainty, and respect for differing interpretations.
  • Trust, authority, and accountability: Lessons from Islamic intellectual history, including al-Jāḥiẓ’s reflections on the written word, for knowledge transmission and human responsibility in generative AI.
  • Cultural norms and personal preferences: How systems navigate conflicts between social expectations and individual choices while recognizing disagreement within communities.
  • Cultural representation and religious bias: Differences within a shared language, including Bangladesh and West Bengal, and the risks of stereotyping or excluding communities in language and image models.
  • Multilingual and multimodal evaluation: Community-informed benchmarks and everyday interactions across languages, dialects, text, speech, and images, with attention to local knowledge and user experience.

Questions for the discussion

  1. What can Islamic media ethics and historical practices of knowledge transmission teach us about trust, authority, and accountability in generative AI?
  2. How should AI systems verify sources and communicate uncertainty or legitimate differences in interpretation?
  3. How can models navigate cultural norms and personal preferences without treating communities as homogeneous?
  4. How can researchers, scholars, and communities jointly evaluate fair representation and reliable everyday interactions across languages, dialects, and modalities?

Event format

The session combines short research presentations with an interactive discussion. The goal is to identify shared research priorities, resource needs, and opportunities for interdisciplinary collaboration.

Event schedule

Time Title Speaker/Presenter Mode
16:00–16:05 Opening Firoj Alam, Qatar Computing Research Institute In-person
16:05–16:20 AI Systems on Islamic Knowledge Capabilities Gagan Bhatia, University of Technology Nuremberg In-person
16:20–16:40 Silent Until Asked to Speak: Al-Jāḥiẓ (d. 869), the Written Word and Trust in Generative AI Mohammed Ghaly, College of Islamic Studies, Hamad Bin Khalifa University In-person
16:40–16:55 Whose Norms? Disentangling Cultural and Personal Alignment in Large Language Models Angana Borah, University of Michigan In-person
16:55–17:10 Unequal Representation of the Two Bengals Md Mohsinul Kabir, University of Manchester In-person
17:10–17:25 Multilingual and Multimodal Culturally Grounded Spoken and Visual Question Answering Shammur Absar Chowdhury, Qatar Computing Research Institute In-person
17:25–17:30 Closing