I'd like to add one: using consistent metaphors. A user of software is constantly trying to form a mental model of how this ethereal, formless thing behaves. A state machine. What can and cannot happen, what will and will not happen after a given action, what can and cannot happen once we're in a different state. The shakier and less scrutable and/or reliable this mental model, the more anxiety is felt.
As programmers we're partly insulated from this effect. We may not know the exact inner-workings of a piece of software we didn't write, but we know some general things about software and the way it does and doesn't behave that soften the huge void of scary unknowns. This helps us form our mental model.
Physical metaphors of objects, continuity, permanence, locality, persistence, independence, are often used in GUIs for this reason. If I click a tab and then click back to the previous one, I expect to return to the same state I was in. If I change the text in one field, I expect that unrelated fields won't be impacted by that. Etc. This is a good starting point. Desktop platforms and then mobile platforms have built additional semi-consistent UX expectations on top of those largely physical intuitions. This helps too. But your application needs to go beyond that: it needs to present a simplified model of its internal state-machine to the user, and then it needs to hold to that. That mental model, once formed by the user, needs to have predictive power about the way the system behaves under different circumstances.