Jeff Noon "Vurt", "Pixel Juice" "Automated Alice" the VERY short stories by Frederik Brown ("Solipsist", "Sentinel") Grant Morrison "The Invisibles"

Helping developers understand how to build better games is exactly the point of this research.

Yes, the data is collected anonymously, it is impossible to uncover the identity of a player. Play sessions are assigned unique numbers so it's impossible also to recognize the same player sending data from different play sessions.

The way we set up the categorization system is a lot closer to the Five Factor Model, where people are scored along certain axes, than the MBTI. It's the difference between "type" and "traits"personality theories. The neural networks were used to find similarities between the profile of players

The article does not claim to offer universal categorization of players, but a localized categorization for THIS game.

the evaluation of players' style is an ongoing process, it's not done once and forever nailed. It dynamically adapts at checkpoints.

a game adapting to the player's style does not mean always reinforcing or giving in, once the game system is aware of patterns of behavior is up to designers to decide what to do: reinforce or frustrate

Contrarily to Bartle types, the clusters do not represent general universal behavioral patterns, but ad-hoc and localized clusters of players with similar behaviors. The results apply to TRU only as the gameplay metrics generating the data are based on mechanics specific to this game.

subversive behaviors do not just represent outliers; it can be very interesting to look at players that don't fit clusters, these are the players that invented rocket-jumping and bunny-hopping, so it might be really interesting to look at "monkey-chasers", "spastic insomniacs" and all profiles not fitting the big clusters as they might be early-adopters.