Learning Agents' Reliability through Bayesian Conditioning: a simulation study