From the perspective of trust, societal dilemmas involve a Red Queen Effect. On one hand, defectors should evolve to be better able to fool cooperators. And on the other, cooperators should evolve to better recognize defectors. It's a race between the ability to deceive and the ability to detect deception.
There's a lot of research on detecting deception, and humans seem not to be very good at it. There are exceptions, and people can learn to be better at it—but in general, we can't tell liars from truth-tellers. Like the Lake Wobegon children who are all above average, most of us think we're much better at detecting deception than we actually are. We're better, but still not great, at predicting cooperators and defectors.1
This is surprising. The Red Queen Effect means both sides improve in order to stay in place, yet in this case, defectors have the upper hand. A possible reason is that we have developed another method for figuring out who to trust and who not to. We're a social species, and in our evolutionary past we interacted with the same people over and over again. We don't have to be that good at predicting bad behavior, because we're really good at detecting it after the fact—and using reputation to punish it.2
In fact, our brains have specially evolved to deal with cheating after the fact. Perhaps the most vivid demonstration of this can be seen with the Wason Selection Task. The test compares people's ability to solve a generic logical ...