bdi-mental-states

Model beliefs, desires, and intentions as endurants linked by motivational properties, and mental processes as perdurants that generate them. Ground every belief in a world state via refersTo, attach justifications, and assign temporal validity intervals using Turtle serialization.

T2B2T is a bidirectional pipeline where incoming RDF triples are translated into belief instances via BeliefProcess entities, BDI deliberation selects intentions and executes plans, and results are projected back into RDF using bringsAbout so downstream systems consume agent outputs as linked data.

Map each cognitive chain link to SEMAS HEAD/CONDITIONALS/TAIL production rules, or map ontology individuals to JADE agent belief sets and JADEX goal classes. The reference documentation includes Python translation layers and Java code for both platforms.

Yes, Logic Augmented Generation serializes the ontology into the prompt context, then validates generated triples against ontological restrictions before acceptance. Inconsistent outputs trigger retry loops with error feedback until constraints are satisfied.

Beliefs without temporal bounds cannot be garbage-collected or conflict-checked during diachronic reasoning. Assigning hasValidity intervals enables SPARQL queries for mental states active at specific times and detects overlapping conflicting beliefs.

Common pitfalls include conflating mental states with world states, omitting temporal bounds, using monolithic beliefs instead of compositional hasPart structures, skipping explicit justifications, and building reasoning chains deeper than three levels which inflates inference cost.