Both generative and reactive planners are based on a similar plan choice mechanism: plan applicability is verified in the current world state. In dynamic and uncertain application domains, world knowledge may be insufficient to select an appropriate plan. In this paper we propose a plan choice model that is not limited to static precondition testing. Instead, it allows information acquisition actions to be executed in order to select the plan which is appropriate to the current situation. This result is obtained with an Object-Oriented domain representation and interpreting plan choice as object dynamic classification. This classification is not done with respect to a static hierarchy, but is conceived as “knowledge chunks” composition, based on the actions performed on the objects.