Abstract. In ubiquitous computing environments, we are surrounded by significant amounts of context information about our individual situations and the situations we share with others around us. Along with the widespread emergence of ubiquitous computing and the availability of context information comes threats to personal privacy that result from sharing information about ourselves with others in the vicinity. We define an individual's context to be a potentially private piece of information. Given the individual context of multiple participants, one can compute an aggregate context that represents a shared state while at the same time preserves individual participants' privacy. In this paper, we describe three approaches to computing an aggregate measure of a group's context while maintaining a balance between the desire to share information and the desire to retain control over private information. Our approaches allow dynamic tuning of information release according to trust levels of the participants within communication range. By evaluating our approaches through simulation, we show that sharing aggregate context can significantly increase the rate at which a group of co-located users learns an aggregate measure of their shared context. Further, our approaches can accomplish high quality context sharing even in situations with low levels of trust, assuming the availability of a small number of highly trustworthy partners.