Ethical and Technical Challenges for Identity-Aware AI
Workshop at ECAI 2025, Bologna, Italy, October 25-30.
https://ecai2025.org/workshops/
Workshop theme: What makes each of us unique, and which ethical and
technical challenges does this imply?
Overview
What makes us unique? Language (and thus the automatic processing of it) is
about people and what they mean. However, current practice relies on the
assumptions that the involved humans are all the same, and that if enough
data (and compute power) is present, the resulting generalizations will be
robust enough and represent the majority.
This approach often harms marginalized communities and ignores the notion
of identity in models and systems. Our interdisciplinary workshop aims to
raise the question of “what makes each of us unique?” to the AI community.
We seek to gather researchers from diverse fields to understand how the
identities of all stakeholders — e.g., the individuals projecting their
views in texts, the individuals perceiving the texts, the individuals
mentioned and those not mentioned in the texts — should be considered in
future research in AI.
Workshop Goals
- The development of a shared and interdisciplinary understanding of
identities and how identity is treated in AI.
- The development of new methods that push the effective, fair, and
inclusive treatment of individuals in AI to the next level.
Topics of Interest
We invite submissions on the following topics:
- *Approaches to model subjective phenomena:* Personalization and
perspectivist methods that leverage disaggregated labeled data, encoding
annotator metadata on their beliefs, moral values, sociodemographic
features, or personal narratives. ML methods to address the challenges of
“learning from disagreements” both from the development of new models and
the collection of data to train such models.
- *Methods for detecting and controlling bias in models and data:*
Techniques to audit fairness, enforce fairness constraints, and learn fair
representation from data, in order to enhance the fairness of models while
maintaining their predictive reliability. Ethical challenges for LLMs in
identity-aware dialog and tasks: diversity, stereotypes, harms.
- *The role of sociodemographics in LLMs:* Such as which characteristics
(and disagreements) they embody and how to measure their capacity for
representing and reasoning about diverse types of identities.
- *Challenges for applying AI methods to model socio-political
phenomena:* Including polarization, impact of media consumption on
public opinion formation, agenda setting, deliberation support, and how
integrating identity into AI methods can influence the accuracy for these
tasks.
- *NLP work at the intersection with social psychology:* The
methodological foundation for quantitative investigation of
identity-related topics. The reflection on best practices to reliably
measure complex constructs such as morals and values. Detection and
analysis of personal narratives across cultures.
- *Accountability of AI in the eye of the general public:* The role of
LLMs, and the responsibilities of AI and NLP developers for ethical use of
identities.
- *NLP work at the intersection of survey science:* The use of LLMs to
model and simulate individuals and subpopulations; the role of LLMs in
personalizing information elicitation; and methodological approaches to
address data contamination and response validation when LLMs are used by
either researchers or respondents.
Submission Types
We welcome the following types of submissions:
- Long papers: Up to 8 pages (excluding references)
- Short papers: Up to 4 pages (excluding references)
- Non-archival submissions, student project presentations, mixed-media
submissions: No page limit
- For non-archival submissions, we welcome creative formats
including:
- Art, poetry, music
- Blog posts
- Jupyter notebooks
- Teaching materials
- TikToks and videos
- Findings papers
- Late-breaking papers
- Extended abstracts
- For creative format submissions, please submit a PDF containing:
- A summary or abstract of your work
- A link to your work (if hosted externally)
- Any additional context or documentation
Submission Guidelines
- All submissions will be double-blind reviewed
- Submissions should follow ECAI formatting guidelines
<https://www.ecai2024.eu/calls/main-track> with the latex template here
<https://ecai2024.eu/download/ecai-template.zip>
- Submit your paper through EasyChair
<https://easychair.org/conferences/?conf=identityawareai2025>
- Accepted papers will be published in the workshop proceedings through
CEUR
Workshop Format
The workshop will be a half-day event featuring:
- Keynote speeches from leading experts in the field
- Paper presentations (oral and lightning talks)
- Participatory design activity to develop a shared interdisciplinary
vocabulary, identify current gaps in datasets for studying identity, and
design a vision for collecting new datasets
- Special student project session
We are committed to ensuring that our workshop is accessible to all. The
workshop will be held in a hybrid format, allowing both in-person and
virtual participation.
Important Dates
- Submissions: 22 August
- Notifications: 26 September
- Camera-ready: 3 October
- Workshop: 25 October
Diversity & Inclusion
We actively encourage submissions from underrepresented communities and
countries. The workshop organizers will provide mentorship and thorough
feedback, especially to first-time authors and reviewers.
Organizers
- Pranav A (University of Hamburg)
- Valerio Basile (University of Turin)
- Neele Falk (University of Stuttgart)
- David Jurgens (University of Michigan)
- Gabriella Lapesa (GESIS, Leibniz Institute for the Social Sciences &
Heinrich-Heine University of Düsseldorf)
- Anne Lauscher (University of Hamburg)
- Soda Marem Lo (University of Turin)
Contact
For queries, please contact: [email protected]
Join us at Identity-aware AI 2025 to contribute to this important
conversation!
_______________________________________________
Corpora mailing list -- [email protected]
https://list.elra.info/mailman3/postorius/lists/corpora.list.elra.info/
To unsubscribe send an email to [email protected]
