OntoAgentic AI

Workshop Description

A symbolic architecture in the orchestrating role

Most current agentic and neurosymbolic systems are effectively based on generative AI and typically use symbolic, knowledge-based AI in supporting roles. This workshop explores the alternative paradigm in which a symbolic cognitive architecture has the orchestrating role, while neural models are invoked as tools for the subtasks they perform best.

We call this stance OntoAgentic AI: agentic AI orchestrated by symbolic, knowledge-based systems and neurosymbolic AI which is symbolic-neural rather than neural-symbolic.

The name designates a field, not a system: “onto-” signals agency grounded in explicit, inspectable models of the world and of the agent's own knowledge, whatever form those models take in a given architecture. A central goal of the workshop is community building — to make this shared commitment explicit, examine where approaches converge and diverge, and begin coordinating a research agenda for the field. A featured application area is cognitive robotics, where the requirements of trust, verifiability, and explanation are non‑negotiable.

This design commitment is shared, in different forms, by work on:

• Soar
• ACT-R
• Companion
• ICARUS
• CLARION
• DIARC
• MIDCA
• Sigma
• ARCADIA
• OntoAgent