The Belief-Desire-Intention Ontology for modelling mental reality and agency
This paper introduces a formal, modular Belief-Desire-Intention (BDI) ontology designed to bridge declarative and procedural intelligence, demonstrating its effectiveness in enhancing Large Language Model coherence and enabling bidirectional reasoning between RDF triples and agent mental states within neuro-symbolic systems.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are trying to build a robot that doesn't just follow a rigid list of commands, but actually thinks like a person. This is the world of Artificial Intelligence, specifically a branch called "cognitive science," where researchers try to give machines a mind. To do this, they often use a famous mental recipe called BDI: Beliefs (what the robot thinks is true about the world), Desires (what it wants to happen), and Intentions (what it has decided to actually do). Think of it like a video game character: it believes there is a dragon in the cave, it desires to get the treasure, and so it forms the intention to grab its sword and fight. For decades, scientists have used this recipe to build smart agents, but there's a problem: these "minds" often speak different languages. One robot might describe a "belief" in a way another robot can't understand, making it hard for them to work together or for humans to trust their decisions. We need a universal dictionary—a shared map of how these mental states connect—so that AI can be transparent, logical, and able to explain why it did what it did.
This paper introduces that universal dictionary. The authors have created a formal "BDI Ontology," which is essentially a highly structured, computer-readable blueprint for how a rational agent's mind works. Instead of just describing the BDI model with words, they built a rigorous map using a method called "eXtreme Design," which starts by asking specific questions (like "What triggered this thought?" or "When did this desire start?") and then builds the map to answer them. They didn't just draw the map; they tested it in two very different ways to see if it actually works.
First, they tried to teach a super-smart AI chatbot (a Large Language Model) using this map. They asked the AI to solve logic puzzles about what a robot should do in different situations. When the AI was given the BDI map as a guide, it got better at spotting contradictions. For example, if a robot believes it is at home but desires to check into a hotel, the map helped the AI realize this is a logical mess and generate a clear explanation of why the robot shouldn't do it. The results suggest that giving AI this structured "mental grammar" helps it produce more consistent and explainable answers, though the authors note this was a small test and the map isn't a magic fix for every error.
Second, they plugged this map into a working robot system called SEMAS. Here, the map acted as the bridge between the robot's internal thoughts and the real world. They showed how a robot could take raw data (like a list of facts about scientists), turn them into "beliefs," use the BDI rules to decide what to do next, and then turn those decisions back into data. It's like a translator that lets a robot's internal "mind" talk directly to the internet's "knowledge base."
The paper finds that this BDI Ontology successfully acts as a "conceptual and operational bridge." It connects the abstract idea of what an agent thinks with the actual code that makes the agent move. While the authors are careful to say this is a foundational step and not a finished product, their experiments suggest that using this shared, precise language makes AI systems more transparent and capable of reasoning about their own thoughts. They argue that for the next generation of AI to be truly smart and trustworthy, it needs this kind of structured, shared understanding of how beliefs, desires, and intentions dance together.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.