Assuring that operators will be able to perform their activities even though the interactive system exhibits failures is one of the main issues to address when designing and implementing interactive systems in safety critical contexts. The zero-defect approaches (usually based on formal methods) aim at guaranteeing that the interactive system will be defect free. While this has been proven a good mean for detecting and removing faults and bugs at development time, natural faults (such as bit-flips due to radiations) are beyond their reach. One of the way to tackle this kind of issue is to propose redundant user interfaces offering multiple ways for the user to perform operations. When one of the interaction mean is failing, the operator can select another functional one. However, to avoid errors and increase learnability, it is important to ensure that the various user interfaces are “similar” at presentation and interaction levels. This paper investigates this relation between dependability and similarity for fault-tolerant interactive systems.