There are two different questions here. And it is important to mark how different they are. There is one school of thought that says 'games teach'. The idea being that students are more likely to engage with games than with standard pedagogical methods such as text books, so they will have more incentive to learn.
A game like Civilization can teach history or SimCity can teach city planning. There seems to be some evidence that this is in fact true. So games can be a good teaching platform. And it is most useful when you have particular learning objectives you want to impart.
But for us, the second part of the formula is this: players learn. This is where our work on imagination is the most central. What players are doing in these environments is engaging in what [psychologist] John Dewey called 'productive inquiry'. They are using the resources they have at hand to learn how to do the things they want to do.
As my colleague John Seely Brown talks about this, they are inverting the traditional learning hierarchy. Where students used to learn about things in order to learn how to be things -- [for example, learning] math, design and engineering in order to become an architect -- they are now using these new learning environments to be what they want to be and falling back to traditional notions of 'learning about' when they get stuck. For example, you just start building and when you need to learn something to help solve a problem, you go read up on it.
It is a whole new way of thinking about learning, where practice and engagement with the world comes first and knowledge and learning in the older sense is a tool for problem solving, rather than the basis for an identity or professional behaviour.
You mentioned the prevalence of team bonding and social skills in your paper. How much of these social skills translate to the real world, and how do you feel about the 'socially-inept gamer' stereotype?
This is a difficult question to answer, because it is hard to know what the "real world" is. We need to stop thinking about these domains as separate. When you are playing with a group of guildmates and you learn to read their social cues, you are not learning to read World of Warcraft social cues, you are learning to read social cues period.
No one asks the question the other way around. How much of what we know about people's real world behaviour translates inside the game? All of it does! If someone in the game says 'Stop talking about how much you dislike Bob, it is getting on my nerves,' no one pauses to say, 'I understand he is annoyed in the real world but how seriously should I take that in the game space, how much of his real world annoyance is transferring into the virtual world?"
So why should we think of it any differently working the other way? From our perspective, the things you are learning in the space of MMOs you are learning in the virtual and physical worlds at the same time.