Plausible spatial behavior is a key capability that autonomous virtual characters need in order to provide ecologically valid social interactions. However, there is a lack of psychological data on spatial behavior in the larger scale social settings and over extended periods of time. In this paper, we present a social navigation model that aims at generating human-like spatial behavior for a virtual human in a social setting with group dynamics. We employ an engineering approach by defining a dynamic representation of interest and then using it as the psychometric function that regulates the behavior of the agent. We evaluate our model by means of two test cases that address different aspect of the model and serve as a proof of concept. Our work is a step toward models for generating more plausible social spatial behavior for virtual characters that is based on both internal dynamics and attributes of the social environment. Copyright © 2014 John Wiley & Sons, Ltd.