From Language to Coordinated Action

Large language models have transformed what AI systems can generate and reason about. Operational teams, however, require more than language. Humans, AI systems, machines, and robots must coordinate actions, divide responsibilities, communicate uncertainty, adapt to changing conditions, and operate reliably in the real world.

This workshop examines the models, architectures, and methods needed to move from language-centered AI toward integrated language-action and world-model systems that can operate as interdependent teams. The emphasis is on effective orchestration across human and artificial agents, especially where safety, uncertainty, reliability, and consequences matter.

Why this matters

AI systems increasingly interact with tools, environments, machines, and people rather than operating only as isolated language systems. As these systems move toward physical and operational settings, errors that are tolerable in text generation may become unacceptable.

The open problem is not simply whether individual agents can perform their assigned tasks. It is how heterogeneous agents can function together as a reliable system: sharing information, coordinating decisions, adapting to uncertainty, managing trust, and producing predictable outcomes when the performance of the team cannot be understood merely as the sum of its individual members.

Questions We Want to Explore

Language to action

How should large language models, large action models, world models, and combined systems be integrated when AI must do more than generate language?

Teams and orchestration

How can humans, AI agents, machines, and robots be organized into effective interdependent teams with appropriate skills, roles, responsibilities, and decision processes?

Reliable operation

How should these teams manage uncertainty, trust, errors, security, governance, and control when operating in dynamic, hazardous, or adversarial environments?

Research areas

Language and Action Models

Large language models, large action models, world models, and combined language-action systems.

Human-AI Teaming

Interdependence, bidirectional trust, team structure, agent skills, task fit, and collaborative performance.

Orchestration and Decision-Making

Task allocation, timelines, decision mechanisms, uncertainty propagation, conflict resolution, monitoring, and control.

Reliable Autonomy

Safety, cybersecurity, governance, traceability, semantic drift, resilience, and operation in high-stakes environments.

Important Dates

November 20, 2026

Submissions Due

Workshop submissions due to organizers.

December 2, 2026

Author Notification

Acceptance and rejection notifications sent to participants.

February 22–23, 2027

AAAI-27 Workshops

Palais des congrès
Montréal, Québec

Participate

We encourage interdisciplinary work addressing the science, engineering, operation, and governance of human-AI-machine systems. For this workshop, submissions may be prepared in LaTeX or DOCX. The proposal specifies extended abstracts of 1–2 pages plus references, with a maximum length of 4 pages, and no required template. At least one author of an accepted paper must attend in person to present at the AAAI-27 workshops.

Workshop Format

We are conducting a 1-day workshop of invited speakers, posters during lunch, and end-of-day panel. The workshop is intended to combine formal presentations with substantial interaction among participants. Click here to see topics of interest.

Invited presentations

Long presentations are planned for approximately 45 minutes including Q&A.

Regular presentations

Regular talks are planned for approximately 15 minutes plus 5 minutes for Q&A.

Discussion

Interactive thematic discussion is planned following the presentations, with discussion among participants at the end of each workshop day.

Page updated

Google Sites

Report abuse