Based on a talk given a few years back at the Austin Game Developers Conference and then distilled into a wonderful book published by Paraglyph Press, Raph gave wonderful insight into what he considers to be the fundamental nature of games and the reason why we play them, and their importance to human survival (!)
His central thesis that games are delivery vectors for cognitive schemata. Games express patterns, and our biology predisposes us toward the consumption of patterns - toward pattern matching, in essence - such that games become highly effective learning and teaching tools. Chutes n' Ladders turns out to be an excellent means of teaching non-Euclidean geometry! As that example highlights, though, games teach almost covertly, by applying a "dressing" to the core gameplay mechanic that helps us engage in a narrative or premise that masks the fact that we're learning.
The problem? The dressing matters. Quake II isn't much different than the various "punch the monkey" banner ads in terms of locating a point on a 2D plane and clicking as the core game mechanic, but the elaborate narrative and visual immersiveness of the former makes it far more compelling and meaningful than the latter. Yet as gamers and game developers, we often see through these dressings - and we often dismiss them as "just dressing," especially when the dressing may be socially unpalatable or otherwise. The classic example here is Grand Theft Auto 3's enablement - not as an explicit codification but as emergent gamer-supplied behavior - of the player to receive sexual favors from a prostitute, with a health bonus, and then shoot or run her over and retrieve the cash payment for services. "It's just a game!" we rant at the people who say this material is dangerous and salacious.
Maybe it isn't.
For our artform to mature, the cognitive schemata that games represent must be taken seriously. We must regard games as media - as communicative channels for the transmission of socio-cultural memes and messages, as much as we regard them as entertainment and pastime. The difference between entertainment and art, according to Koster, lies in intensity or how seriously the fundamental concept is being taken.
He also spoke on predictability, in that games always have a single right answer, and that being a fundamental flaw and part of why they have/can not yet be taken seriously as a medium for art. Art allows for contemplation and interpretation by the beholder; it doesn't possess just one correct answer. "Games will never be mature as long as the designers create them with complete answers to their own puzzles in mind." Of course, there will be a class of players who prefer to tackle problems they already know how to solve - and that's fine. But games need to expand beyond those boundaries of shooting things and moving things and explore economic questions such as whether stock options in a dot-com constitute a viable payment/investment strategy and other complex, open-ended questions.
In essence, posits Koster, all games are "serious" - or at least should be. Games are the only medium that teach formal abstract systems; don't let them just be tic-tac-toe (noughts and crosses)!