About the Workshop
The 1st PANDORA Workshop will take place at EMNLP 2026 in Budapest, Hungary!
Modern NLP systems are built on a simplifying assumption: that there exists a single correct answer. Yet in many domains, what is harmful, fair, polite, or persuasive—there is no single truth. Human disagreement is not noise to be removed, but a reflection of pluralism: diverse perspectives shaped by culture, context, and lived experience.
As NLP increasingly relies on human feedback to align models, the field faces a fundamental choice. Do we continue to collapse diversity into a single consensus, or do we design systems that can represent and reason over multiple, sometimes conflicting, viewpoints? This is not only a technical question, but a societal one: about whose values are encoded in AI systems and how they shape the world at scale. The workshop PANDORA (Pluralistic AI & NLP: Diversity-aware, Sociotechnical, Responsible Alignment) is a call to rethink this paradigm. We envision NLP systems that are diversity-aware, context-sensitive, and socially grounded. PANDORA aims to advance a new generation of methods, metrics, and practices for pluralistic AI across topics of :
- Data: Eliciting and representing pluralistic judgments, and separating error from meaningful variation.
- Modeling & Alignment: Learning from diverse and conflicting feedback without collapsing to one single “best” view.
- Evaluation: Context-aware assessment beyond single-ground-truth metrics.
- Normative & Sociotechnical Questions: Determining when pluralism is beneficial, when consensus is required, and understanding the resulting trade-offs.
Stay tuned for Call for paper!
You can follow us on BlueSky @pandora-workshop.bsky.social, and on Twitter/X @Pandora_EMNLP26
Contact us via email at pandora.workshop.1@gmail.com for any questions.