The Third Workshop on Social Influence in Conversations (SICon 2025) Held at ACL 2025, Vienna, Austria
The Third Workshop on Social Influence in Conversations (SICon 2025) invites submissions on recent advancements in social influence dialogue systems and other NLP tasks that can expand these systems' social influence capabilities. Our workshop aims to foster inclusivity and interdisciplinary dialogue, bridging insights from diverse fields, including NLP, psychology, sociology, and decision science. SICon 2025 aspires to provide an open and engaging forum to connect researchers from academia and industry working on related topics, especially in negotiation, persuasion, and other forms of social influence.
This year’s workshop also emphasizes cultivating connections between subfields and creating opportunities for newcomers to adapt and thrive within this emerging community. To this end, we encourage submissions from diverse perspectives and career stages, providing a platform for collaboration and growth.
We aim to promote discussion around key questions such as:
How can social influence systems model users and plan optimal responses systematically?
How can linguistic theories from the social sciences (e.g., persuasion, negotiation tactics) enhance social influence systems?
What unites and differentiates various social influence tasks structurally?
What ethical challenges do AI systems that engage in social influence face, and what guardrails must we implement before deploying them in real-world contexts?
Topics include, but are not limited to:
Analysis-focused contributions
Analyzing linguistic behaviors in social influence (SI) interactions, including persuasion appeals, emotional expressions, therapeutic methods, or deception (e.g., strategic games like Diplomacy or Catan).
Associations between linguistic behaviors and SI task outcomes (e.g., agent performance, user satisfaction).
Examining user attributes (e.g., demographics, personality traits) and their association with SI outcomes.
Cultural differences in SI interactions.
SI in text combined with other modalities (e.g., face-to-face multi-issue negotiations).
System design contributions
Dialogue systems for SI tasks such as strategic games, negotiation, persuasion, argumentation, recommendation, or emotional support.
Multi-agent SI systems.
SI systems leveraging large language models (LLMs).
SI systems informed by linguistic theories from social sciences.
Addressing unintentional SI in human-facing NLP systems.
Contributions advancing sub-goals in SI tasks
Detecting influence strategies, outcome prediction, stance detection, and argument mining.
Partner/opponent modeling and emotion recognition in SI interactions.
Other contributions
Datasets capturing forms of social influence, especially in underrepresented languages (e.g., human-human datasets, wizard-of-oz studies).
Guardrails for Fairness, Accountability, Transparency, and Ethics in AI systems with SI capabilities.
Opinion or position papers on social influence.
We refer to the ACL 2025 website for paper templates (the official ACL style templates) and requirements for both submission types, regular workshop submissions, and shared task submissions. We will allow additional space (not counted towards the above page limits) for supplementary material, including an appendix. We encourage all authors to include careful considerations of the limitations, broader impacts, and ethical considerations of their work. We also encourage authors to provide their code and data, where applicable, in the interest of reproducibility.
Regular workshop submissions are archival short (4 pages) and long (8 pages) papers. There is also a non-archival track for extended abstracts (2 pages) covering ongoing work on social influence NLP. Both tracks' accepted papers will be publicly available on the workshop’s website.
If you are unsure whether a topic is suitable, contact the workshop organizers at sicon-chairs@googlegroups.com.
The review will be double-blind. Please do not include any self-identifying information in the submission (covers all three submission types above). This includes anonymizing the already-published work by removing acknowledgments, self-citations, etc.
Non-archival submissions are expected to receive less detailed reviews compared to archival submissions.
SICon 2025 will use a separate ethics review, as required. Please ensure that any ethical concerns are appropriately addressed in the paper.
Parallel Submissions: Parallel submissions to other venues are NOT allowed.
Mode of Presentation: To foster thorough interaction and feedback, presentations will be primarily based on poster sessions. However, we will recognize three exceptional papers by providing dedicated slots for lightning talks (5 minutes each).
Authors of accepted archival papers should upload the final version of their paper to the submission system by the camera-ready deadline. Authors may use one extra page to address reviewer comments, for a total of nine pages + references.