(The following is armchair psychological speculation based on anecdotal evidence. If anyone can respond with relevant science, that'd be awesome, but otherwise responses in kind are quite welcome.)
I'm puzzled regarding the motivational effects of negative emotion, particularly shame, guilt, and regret--I'll just say 'regret' going forward, though there could be important differences between them. In particular, I've observed people being oddly unmotivated to avoid doing things that will very predictably cause them unhappiness, in a way that seems to go beyond garden variety akrasia. It recently occurred to me that homo hypocritus theories of regret predict this result.
In contrast to the naive theory, under which regret is inflicted by a brain on itself to teach it to change its behavior, the homo hypocritus theory holds that regret exist to convince other observers that its behavior will change. This allows someone to continually do antisocial things while convincing others that they won't do so in the future. Look at it from the perspective of a gene-selfish brain designer: self-inflicted regret, unlike externally-inflicted injuries, doesn't carry any intrinsic harms, so there's no reason to behave in a way that avoids it.
On closer inspection, neither theory makes a whole lot of sense on its own. Regret has distinctive displays that would be pointless if they were only for internal consumption. And there would be no reason for others to think regret would lead to changed behavior if this were never the case. So a revised theory combines the two as follows:
Naive 2.0) Regret originally evolved to act both as impetus for an individual to change behavior and signal to others that such a correction was taking place.
Homo Hypocritus 2.0) Sometimes brains exploit this by feeling regret and sending the signal while somehow blocking the behavior update--this is advantageous if the regretted action only harms the agent via others' disapproval and the emotional display allays that disapproval.
This could vary by individual, by situation, or both. And it comes with the standard evo-psych disclaimer that people with HH brains aren't faking their suffering. Rather, this might explain behavior we'd call 'compulsive' or 'self-destructive'--it could be that the compulsion to do regrettable things isn't extra-strong, but rather that the brain's motivation to avoid those behaviors is blocked. In many cases these individuals, subject to constant cycles of action and regret, would be the primary victims of their brain's cynical adaptation.
So what testable predictions would this yield? (We can worry about how to test these ethically later).
* There should be people and/or situations where concrete external punishment would be much more motivating than regret even if the latter causes much more suffering.
* If we could arrange for people to experience regret inside an MRI machine, we might observe variations in how much lasting change occurs, and observe those whose brains change more change their behavior more. This might also correlate with life outcomes or general tendency to do regrettable things.
* At least some humans should have innate defenses against evolutionarily-hypocritical but personally-sincere regret.
* There should be distinct and identifiable modes of compulsive/self-destructive behavior that only occur with behaviors whose harmful results are mediated by other humans' reactions, at least as far as the savannah-brain can tell.