On Frank's account, an emotion is a commitment device if and only if (a) it increase the probability of prosocial behavior and (b) it is a hard to fake signal (i.e. evolutionary biologists call them "handicaps").
Robert H. Frank
Consider an emotion like guilt. One can argue that guilt accomplishes (a) because it is reparative when moral transgressions occur (i.e. when one cheats). However, it does not look like (b) is satisfied. Specifically, psychologists deny that guilt has a unique facial expression. Averting one's eyes and lowering your shoulders can express embarrassment, shame, etc. Even Paul Ekman acknowledges the point,
As best as I can determine, shame doesn’t have its own signal. And neither does guilt. They’re very hard to reliably distinguish from the family of emotions: sadness, disappointment, grief, discouragement, and anguish. It’s not that I think shame and guilt are the same; it’s just that they are the same in signal.
Thus, it appears that guilt is not a commitment device. Question: How might Frank reply to this argument?