Computer simulations have become key to modeling human behavior in many disciplines. They can be used to explore, and deepen our understanding of, new algorithms and interfaces, especially when real-world data is too costly to obtain or unavailable due to privacy or competitiveness reasons. In information retrieval, simulators can be used to generate inputs from simulated users -- including queries, clicks, reformulations, judgments -- which can then be used to develop a deeper understanding of user behavior and to evaluate (interactive) retrieval systems. The trust that we put in simulators depends on their validity, which, in turn, depends on the data sources used to inform them. In this paper we present our views on future directions and challenges for simulation of queries and clicks from transaction logs.