Last CfP of 1st CLaRAMAS workshop at AAMAS 2026

[Stefano MARIANI]

Dear Colleague, 
we hope this email finds you well :)

This is the last Call for Papers for the 1st International Workshop on “Causal Learning and Reasoning in Agents and Multiagent Systems” (CLaRAMAS), hosted by the 25th International Conference on Agents and Multiagent Systems (AAMAS).

The workshop seeks contributions at the crossroads between learning of and reasoning with causal models, and engineering agents and agent-based systems (both programming and learning approaches).

The extended deadline (March 1st) is strict.
The workshop is held on May 26th.
Accepted papers will be published in Springer CCIS series.
Prof. Emiliano Lorini will deliver a keynote speech.

Please, feel free to extend this invitation to your collaborators that you believe may be interested in sharing their work in progress on the subject.

Check out the workshop’s full scope & aims and the Call for Papers (also reported at the end of this email) on the workshop website: https://claramas-workshop.github.io/claramas2026/ 

 Looking forward to receive your feedback, 
 our best regards.

 The CLaRAMAS Program Chairs
    --- Stefano Mariani, Mehdi Dastani, André Meyer-Vitali, Julien Siebert

 ##### Call for Papers #####

The concept of an “agent” represents a foundational abstraction in software engineering, encapsulating the notion of agency—namely, the capacity of a software entity to bring about effects in pursuit of specific goals within its operating environment. Exercising agency requires the ability to interpret the structure and dynamics of that environment and to anticipate its responses to the agent’s actions. In essence,

    agency hinges on understanding and leveraging the causal relationships among observable and controllable variables (e.g., through sensors and actuators).

Such causal reasoning is indispensable for planning actions that reliably achieve intended objectives—a principle reinforced by recent research on causal inference in emerging “agentic AI” systems.

This requirement extends naturally to multi-agent systems (MAS), a cornerstone of distributed artificial intelligence, where multiple agents coexist and interact within a shared environment. These interactions – whether cooperative or competitive – contribute to individual and systemic goals, either through direct communication or indirect influence on the environment. Consequently,

    effective coordination in multi-agent settings depends on a causal understanding of interdependencies among agents’ behaviors.

Only by modeling these reciprocal influences can agents achieve robust and purposeful collaboration (or competition) toward their respective objectives.

 ⚠️ However, this fundamental role of causal modelling of the agent-environment and agent-agent relationships is not yet widely and deeply discussed in the AAMAS community. ⚠️

Accordingly, CLaRAMAS welcomes submissions dealing with the following topics of interest:

• how to integrate causal learning in agent architectures and MAS
• how to carry out causal learning of agent-environment and agent-agent relationships from the standpoint of an individual agent or of the MAS as a whole
• how causal modelling and learning can be integrated in agent-oriented software engineering methodologies
• exploring causal explainability, safety, and robustness in agent(s) design, for instance in robotic and multi-robot systems
• how causal learning may integrate with learning-based approaches to agent design, such as with Reinforcement Learning for counterfactual reasoning, credit assignment, policy evaluation, policy improvement in single- and multi-agent systems
• theoretical foundations of causal learning and reasoning in single- and multi-agent systems, including the relationship between sequential decision making (e.g., MDPs) and Pearl structural causal models, integration of game-theoretic formalisms, etc.
• practical applications of causal learning and reasoning in MAS
• cooperative planning, prediction, and diagnosis using (perhaps, partially) shared causal models
• combining causal learning and reasoning with planning and adaptive control, including model-based, model-free and hybrid approaches
• cooperative causal discovery and inference in MAS
• neuro-symbolic AI via causal models
• evaluation and benchmarks for causal MAS applications, including datasets, metrics, simulators, and reproducible experimental pipelines

 Check out the submission dates and instructions at the CLaRAMAS website: https://claramas-workshop.github.io/claramas2026 

 ##### #####