Flooding is a major risk for coastal cities. Physical infrastructure, such as levees, are commonly used to reduce flood hazards. Typical planning practices often neglect  human-system responses to built infrastructure, such as the so-called “levee effect”. We illustrate how the levee effect can emerge from the censoring of flood risk information by levees and analyze what factors contribute to the effect strength. The levee filters decision-relevant information by eliminating nuisance flood events and reducing the impact of events which overtop the levee.This information-filtering effect shifts risk from smaller, higher-frequency events to rare, extreme flood events, as residents are unable to learn about underlying risk from less damaging nuisance flooding events. We explore the possible levee effect from a proposed levee construction project in Baltimore, Maryland by simulating household responses to dynamic flood events using the Coupled Human and Natural City Evolution - Coastal model (CHANCE-C). Additionally, we capture the uncertainty in flood damage estimates from future simulated development using the Uncertain Structure and Fragility Ensemble (UNSAFE) framework. We find that the susceptibility of levees to breaching is a major and often neglected driver of the levee effect and highlight the uncertain emergence of the levee effect even if the dynamics of community response to flooding events are kept constant.