Certainly - you might want to read Rosalind Picard's "Affective Computing". The book makes a point that for really intelligent systems - those that are able to cope with (emotional) users in natural situations - you need to be able to recognize user affect and to adequately respond. However, since those are no trivial tasks, you might want to specify the functions of your system, and then judge if it has to work in a context were user emotions are of relevance. A system that is supposed to be a professional tool might be less exposed to situations in which user emotion is a relevant factor, whereas a consumer electronics system that works in every-day contexts might profit from its adaptability to user affect.
Intelligence is also about the ability to empathise, ie, to predict ones internal feelings or beliefs and manipulate it, for example, a monkey who discovers food keeps cool and as if he saw nothing so not to alert others ... Not all animals have this ability ... So rather than make machines simulate emotions its far more important for machines to understand contextual emotions