Jason Smiles: Incremental BDI MAS Learning

This work deals with the problem of intentional learning in a  Multi-Agent System (MAS). Smile (Sound Multi-agent Incremental  LEarning), a collaborative learning protocol which shows interesting  results in the distributed learning of well known complex boolean  formulae, is adopted here by a MAS of BDI agents to update their  practical reasons while keeping MAS-consistency. An incremental  algorithm for First-Order Induction of Logical Decision Trees  enables the BDI agents to adopt Smile, reducing the amount of  communicated learning examples when compared to our previous  non-incremental approaches to intentional learning. The protocol is  formalized extending the operational semantics of AgentSpeak(L), and  implemented in Jason, its well known java-based extended  interpreter.