As per the emotion theory, these are based on accountable actions of some agent that Steve believes are right or wrong.
Principle: User should attend to me when I am talking with them. Principle: User should be cautious when it is appropriate to do so. Principle: I should be patient with the user. Principle: I should tutor my users well enough that they make adequate progress on the domain tasks.
The principle regarding user-attention can be triggered, for example, when Steve is demonstrating a procedure, and the virtual reality software reports that the user's head is oriented so that Steve is no longer in the user's (presumed) field of vision, Steve might get annoyed. That Steve could be wrong is not something we worry about because people are often wrong in their perceptions as well; wrong or right, it should be clear to the user that Steve cares.
Impatience with the user can work to call attention to a (perceived) lack on the user's part. Steve might also have conflicting emotions about this, wherein he feels that getting impatient with the user is wrong. Thus Steve might be impatient about a lack of caution on the user's part, but feel shame for getting annoyed. It is possible that in this way the user might quickly get the message about lack of caution, but have the criticism softened by Steve's remorse.