The paper estimates the causal effect of a health treatment on patients' beliefs, preferences and memories about the treatment. It exploits a natural experiment which occurred in the United Kingdom during the COVID-19 vaccination campaign. UK residents could choose to opt into the vaccination program, but not which vaccine they received. The assignment to a vaccine offered little objective information for learning about its qualities, but triggered strong psychological demand for reassuring beliefs. We surveyed a sample of UK residents about their beliefs on the different COVID-19 vaccines before and after receiving their jab. Before vaccination, individuals exhibit similar prior beliefs and stated preferences about the different vaccines. After vaccination, however, they update their beliefs overly optimistically about the safety and effectiveness of the vaccine they received, state that they would have chosen it if they could, and have distorted memories about their past beliefs. These results cannot be explained by conventional experience effects. At the aggregated level, they show that random assignment to a health treatment predicts a polarization of opinions about its quality. At the individual level, these findings provide evidence in line with the predictions of motivated beliefs and over-inference from weak signals in a real-world health setting.