Cognitive dissonance theory was a response to the rational man theory in economics, and so there was a ready-made reason for it to present itself as a proof that man was an irrational actor. Much of the popular literature on cognitive dissonance tries to make exactly this point.
But a middle position between the two might be to point out that, for us, “to be rational” means to have some sort of concord between belief and action, or between our general principles and the concrete applications in which we find them. By nature and blind necessity we will tend to this concord and consistency. The given in human life is that our lives must be truth preserving, and there must be a sort of logical validity between ideals and situations, beliefs and actions. But this does not mean that the most reasonable of two options will be chosen, or even that it will be chosen given enough time and a large enough sample set. If you punish pickpocketing by hanging, then the only possible response of a rational actor would be to utterly stop pickpocketing. This doesn’t happen. And so rational-agent theories probably don’t have the right account of rationality, but they are right to insist that our actions happen for reasons, and even that they are consistently checked by evidence in one way or another; and the pop-cognitive dissonance writers have to give up the rhetoric of human beings having no rational motivations, since the consistency we seek to establish in eliminating dissonance is sought precisely as truth-preserving. It is a properly rational consistency.
To repeat the example from logic: human motivations and doctrines necessarily tend to validity, but they tend to soundness only contingently and are fixed to truth or error in this further sense by additional factors. One such factor is, of course, the evidence for our beliefs and their outcomes, but this is primarily material and accidental. The more salient factors are our own habits of action and belief are good or evil, i.e. whether we have virtue or vice. That said, much of life lacks the determination of fixed habit and so is characterized by the waffling and conflicted interior state that translators of Aristotle are stuck calling “incontinence”, where a belief is now forcing us to change behavior and later the same behavior forces us to change belief.
The basic description of action then seems to trace all the way back to Aristotle: a human action can always be analyzed into causally prior principles, one of which is a general rule and the other a belief that links this general rule to an action to be concretely done. So long as the causally relevant factors of my action are, say, “all food should be healthy” and “of my options, X is healthy and Y isn’t” then I’ll find myself eating a big bowl of X. If I find myself eating Y, it’s either because I’m working from a different first or second premise, not because I’ve given up acting for reasons altogether. The premises in question here are at all levels of consciousness – we are usually working from how the action feels, if at all. But without these premises we lose any sense of what “dissonance” or “consonance” could mean.