Angelo Ferrando, Daniela Briola, Rem W. Collier, Viviana Mascardi
Matteo Baldoni, Franziska Klügl, Andrei Olaru, Alexandru Sorici, Adina Magda Florea (a cura di)
Multi-Agent Systems: 22nd European Conference, EUMAS 2025, Bucharest, Romania, September 3–5, 2025, Proceedings, Part I, capitolo 9, pp. 142–162
Lecture Notes in Computer Science (Lecture Notes in Artificial Intelligence) 16528
Springer Cham
luglio 2026
Recent trends in artificial intelligence increasingly conflate agency with large language model (LLM)-driven toolchains, often under the banner of “Agentic AI”. While these systems offer impressive generative capabilities, they typically lack the structured deliberation, persistent intentions, and transparent reasoning found in classical symbolic architectures. In contrast, Belief-Desire-Intention (BDI) agents support goal-driven behavior through explicit mental states and formally grounded control loops. In this paper, we critically compare these two paradigms and present a conceptual proposal for a hybrid architecture that strengthens Agentic AI by embedding it inside a BDI-style deliberative mechanisms. We outline a conceptual reasoning cycle in which a central controller maintains symbolic beliefs, goals, and intentions, and dynamically delegates planning to modular decision services—either symbolic planners or LLM-based generators. A meta-cognitive layer monitors performance and refines decision strategies over time. This architecture is illustrated through an agent-based scenario, highlighting how symbolic and generative decision making can be orchestrated under principled control. Our contribution is primarily architectural and conceptual: we aim to reframe Agentic AI as a deliberative neuro-symbolic paradigm, providing a foundation for future empirical validation, adaptive decision policies, and formal verification efforts.
parole chiave
Agentic AI, BDI Agents, Hybrid Reasoning, Neuro-Symbolic Architecture
evento origine
rivista o collana
Lecture Notes in Computer Science
(LNCS)
pubblicazione contenitore
Multi-Agent Systems (curatela, 2026) — Matteo Baldoni, Franziska Klügl, Andrei Olaru, Alexandru Sorici, Adina Magda Florea
pubblicazione associata
progetto finanziatore
ENGINES — ENGineering INtElligent Systems around intelligent agent technologies
(28/09/2023–27/02/2026)
FAIR-PE01-SP08 — Future AI Research – Partenariato Esteso sull'Intelligenza Artificiale – Spoke 8 “Pervasive AI”
(01/01/2023–31/12/2025)
funge da
pubblicazione associata per
