Viddy well, O my brothers
This passage caught my eye, in an article describing the impact of the massively multiplayer online game:

Unlike single-player games, these virtual environments don't go into cryogenic suspension in your absence. Events transpire. Battles are fought. Rivalries flare. Alliances are formed. In a world that can't be flicked on and off, actions have lasting consequences, both narrative and social.
As players build ther characters, accruing strength and skill with experience, they also build reputations and relationships. Survival depends on whom you know, and how much they trust you, rather than simply on the weapons you're carrying. Success depends less upon combat skills than on social interactions.

So I offer some observations.

• Parents of children who have immersed themselves in Dungeons and Dragons fear that role-play games are corrupting them. Even those who dispute that role-play games corrupt children strongly agree that role-play can modify behavior.
  
• Children who misinterpret facial expressions tend towards violence and delinquency. They perceive threats where there are none, making them more prone to violence; they receive stress and dissatisfaction from human company. Prenatal exposure to alcohol is a culprit, though I suspect that other, non-organic causes exist, and under controlled circumstances it may be unlearned, even if organic in origin.
  
• The free market relies as much on one maintaining one's reputation, as on one having recourse to unfair dealers through the courts. Businessmen who deal unfairly or without goodwill get fewer and less valuable deals, even if they operate well within the boundaries of the law. To prosper, one must be trusted and one must trust.
  

Put these all together and what might they spell? Massively-parallel multiplayer games have evolved to the point that a player's actions have consequences that the player may not notice immediately. Among those consequences, a bad reputation. People who do not value, therefore do not cultivate, a good reputation will become life's losers.

By engaging in role-play on this scale, with real people at the millions of distant ends of the game, he will learn real-world values, but in a way that he can afford to lose a few times, come to understand why he lost, and not suffer (or cause) real violence or hardship. She would learn that by giving someone else enough rope to hang himself, she fastens the other end of the rope to herself.

The kid may also learn what many ordinary adults (including many therapists) have not: cultivating and protecting one's reputation by serving others is not tantamount to kissing ass, castration, or selling out. It demands bravery, singlemindedness, and an erotic drive; if more adolescents and teens saw this, they would perhaps be more enthusiastic about growing up and being responsible.

The "unpacking of the face," can make this game therapeutic, can intervene with those troubled juveniles who misinterpret the face. As a game, the kid will likely stay with it longer than he would with interacting with real people. As the game unfolds, the kid learns better to interpret others' emotions and intent, and to work towards longer-term goals; he or she might also benefit from a choreographed "replay" of the poorer decisions and the consequences that they precipitated.

A massively-parallel multiplayer approach is preferable to one manipulated by few people or only one person, even if that one person is a credentialed therapist, because it would better mirror the real world the kid will (re)join. Even some bad actors should be admitted, because we can't keep them out of the real world. Let the therapist moderate, not create, this world.

This is just a thought, of course, and I do not claim it as my own. Many libertarians would find the idea chilling, like the Lodovico Technique made warm and fuzzy by hiding it behind a StarWars videogame, and handing the joystick to the subject. But as a parent, I can sympathize with parents who somehow failed to intervene in the right way, at the right time.