Yasser Bourahla
mOeX
Friday, September 25th, 2020, 9:30
INRIA, Montbonnot
Abstract
When agents learn knowledge, such as ontologies, about their environment independently it may be diverse, incorrect or incomplete. This knowledge heterogeneity could lead agents to disagree hindering their cooperation. Existing approaches usually deal with this communication problem by relating ontologies, without modifying them, or, on the contrary, focus on building common knowledge for itself. Here, we consider agents adapting ontologies learnt from the environment in order to agree with each other when cooperating. In this scenario, fundamental questions arise: Can this process improve knowledge correctness? Do all agents end up with the same ontology? To answer these questions, we design a two-stage experiment in which agents first learn to take decisions about the environment by classifying objects and then interact with each other to agree about the decision to take in face of various objects. The learnt classifiers are turned into ontologies that agents modify to agree on decisions. We show that agents can indeed reduce communication failure, improve the accuracy of their knowledge about the environment and yet not obtaining the same ontologies.