Inclusive AI:
Rethinking AI-based Multimodal Interaction for Diverse and Underrepresented Users 

This workshop marks the inaugural edition of “Inclusive AI”, an initiative focused on addressing social interaction challenges in AI systems, with a particular emphasis on inclusivity for diverse and underrepresented user populations. It builds on a growing recognition across the HCI, HRI, and AI communities of the need for cross-disciplinary venues to advance socially interactive AI that is equitable, adaptive, and user-aware.

The potential of AI that engages with people socially, so called social AI, is widely acknowledged, yet its applicability to individuals with disabilities remains critically underexplored, despite these users may benefit from such technologies. Many social AI systems rely on normative assumptions about communication: typical social cues, behavioural patterns, or language use. This can lead to misrecognition, exclusion, or interaction breakdowns when users’ behaviours diverge from these norms. Even when interaction is mediated through conversational AI, inclusivity is not guaranteed. For example, conversational systems often assume proficiency with abstract, indirect, or culturally specific language forms when eliciting or refining user intent, creating barriers for users with limited language proficiency or diverse cultural backgrounds. In high-stakes domains, such as healthcare, employment, or autonomous systems, these failures can have serious and irreversible consequences. New methodologies are needed to explicitly account for variability in social and communicative behaviours. This raises pressing questions: to what extent can social AI meaningfully address the needs of individuals with disabilities, and what new methodologies, models, and technological foundations are required to ensure genuinely inclusive interaction? 

This workshop directly addresses these challenges by examining persistent disability biases in AI systems, insufficiently diverse training datasets, foundational technical issues, methodological blind spots in modelling diverse interaction behaviours, and ethical frameworks that rarely account for the specific experiences and vulnerabilities of this population. Building upon the expertise of the organisers, the workshop aims to involve the ICMI community in identifying and acknowledging the societal risks posed by biased or inaccurate social AI, highlighting the urgency of developing technologies that support, rather than exclude, people with disabilities. To this end, participants will be involved in participatory, interactive activities aimed at articulating a shared research agenda that positions inclusion as a central pillar, rather than a peripheral consideration.

Important dates

• Paper submission deadline: 15th June 2026

• Notification of acceptance: 15th July 2026

• Final camera-ready deadline (hard): 2nd August 2026

• Workshop day: 5th October 2026

Paper submissions

We will invite extended abstract papers (1–4 pages) to be shared also on the workshop website, supporting knowledge exchange and contextualizing the discussions. The accepted papers will finally appear in the Adjunct Proceedings of ACM Public Library under the ACM Open Access Policy as reported in the ICMI'26 conference website.

We will solicit extended abstract papers that cover the topic of the workshop and can trigger interactive discussions. Submissions will undergo an internal review process conducted by a Program Committee composed of the workshop organisers and other invited members. The review will focus on relevance to the workshop themes, potential to stimulate discussion, and diversity of perspectives.