Reasoning about what other people know is an impor-tant cognitive ability, known asepistemic reasoning, whichhas fascinated psychologists, economists, and logicians.In this paper, we propose a computational model of hu-mans’ epistemic reasoning, includinghigher-orderepistemicreasoning—reasoning about what one person knows about an-other person’s knowledge—that we test in an experiment usinga deductive card game called “Aces and Eights”. Our startingpoint is the model of perfect higher-order epistemic reason-ers given by the framework ofdynamic epistemic logic. Wemodify this idealized model with bounds on the level of fea-sible epistemic reasoning and stochastic update of a player’sspace of possibilities in response to new information. Thesemodifications are crucial for explaining the variation in humanperformance across different participants and different gamesin the experiment. Our results demonstrate how research onepistemic logic and cognitive models can inform each other.