Bayesian modeling of task dependent visual attention strategy in a virtual reality environment