Interactive intelligent agents use cognitive models to antic- ipate and simulate human behavior, and a fundamental pil- lar of human cognition and interaction is narrative. As a result, agents need to understand human comprehension of various types of narratives. A key component of modeling comprehension is the perception of interestingness of con- stituent actions and events in the narrative. In this paper, we briefly review previous theories of interestingness, drawn from cognitive psychology and narratology. We propose expanded computationally amenable theory of interest which takes into account both cognitive and experiential aspects of perceived interest. To empirically validate the theory, we present a narrative generator for abstract animations inspired by Heider and Simmel’s experiments (Heider & Simmel, 1944). The generated animations are parameterized along the dimensions of our proposed theory. We present the results of a user study with this generative system and report on the effects of visual narrative parameters on perceived interest.