In this article we present a biologically inspired approach to predictive task-dependent gaze control for active vision systems based on maximization of visual information. In this context we introduce two novel concepts: Information Content of a view situation and Incertitude. Based on these concepts, we present a method for selecting optimal subsequent view directions which maximize the relevant visual information thus contributing to an improved performance in typical autonomous robot locomotion tasks such as obstacle avoidance, self localization, and navigation.