In my career as an educational game designer, I mainly focus on the problem of getting a set of learning objectives that are met through gameplay. This approach works very well for teaching skills, but I don't think it's as strong an approach for teaching knowledge. There are just some things you can't learn by doing. Like the capitals of all the states, or the multiplication tables, or other things that are really more "memorization" than they are "learning".

Science Magazine has posted a video introduction to their special "Education and Technology" issue.

I couldn't agree more about the need to have real data telling educators (and developers) what does and does not work, and that we need a better distribution model for educational games.

However, It's always a little disingenuous to talk about World of Warcraft and The Sims when trying to make the point that video games can reach millions of players. WoW and The Sims are outliers - the most successful games out there.