In entertainment applications, artificial intelligence techniques have most often been used to implement embodied agents or to automatically generate artistic content. A more recent development concerns using AI to support the user experience through new AI-based interactivity techniques. This is especially of interest for the development of artistic installations based on interactive 3D worlds. A major difficulty in developing such installations is to properly translate the artistic intention into actual elements of interactivity, which in turn determine the user experience. The starting point of this research was to facilitate the description of high-level behaviors for virtual worlds that would form part of virtual reality (VR) art installations. In our approach to interactivity, the consequences of user interaction can be dynamically computed to produce cascaded effects eliciting a specific kind of user experience. This chain of events is computed from first principles embedding elements of the artistic brief (the artist's initial conceptual description of the interactive installation and the intended user experience). In other words, AI techniques are used for their ability to represent actions and to compute analogical transformations on them to create a user experience.