1) Write a 1-2 paragraph summary of the assigned reading. (6 pts)

An Experiment in Automatic Game Design, by Julian Togelius and Juergen Schmidhuber, writes about how computational intelligence (CI) can be used to evolve the rules of a game. While there have been studies on how computational intelligence techniques can be used in game design to develop new behaviors and environments, there have not been studies that show how CI can be used to generate the game itself. This paper begins by introducing the concept of measuring fun and how, in games, what makes it fun is the challenge of the player experience. If a game is too easy to play or if the outcome can be predicted, it is not fun for the users to play. They talked about theories of entertainment can be divided into two categories: static and dynamic. Static theories are ones that do not require the player to learn while playing, whereas dynamic theories adapt to the game over time. In the section “Game Engine and Rule Space Definition,” they laid out constraints or axioms for a game with evolving rules capabilities, as they noted that “it would not be possible to evolve a game without any constraints at all.” They then gave examples of different games that could be generated using the same set of axioms, which I thought was very cool. They continuously returned to this idea of play = learn, which was cool. They also noted that for many of the games they evolved they were not considered very good games based on the way they defined good games earlier int the paper. At the end, they noted how their overall purpose in this paper was to encourage people to explore this currently very unexplored area of game design research and how CI-generated games may be the hack to creating unpredictable games that will shock game critics.

2) What, if anything, did you find exciting or inspiring about the reading? What do you feel the authors did well? Were there any points that you agreed with? (2 pts)

I was very intrigued by Koster’s informal theory of fun in game design and how it relates to learning. He argued that “playing is learning, and learning is fun.” He believed that games that are easy to start playing but continuously get more difficult as one continues are examples of fun games. I loved the idea that this can be summed of as “games that take a minute to learn, but a lifetime to master.” This made me think about chess and how it definitely falls into this category. I also thought the idea of dynamic theories of fun sounded cool, and I liked how these theories were able to adapt to the game over time. I thought that the authors overall did a good job of explaining the different ways they laid out the constraints for generating games and I was interested by what they made “changeable.” I agreed with a lot of their overall arguments and reasons for writing this paper.

3) Was there anything that was confusing or unclear? Were there any points that you disagreed with? (2 pts)

I did not quite understand how fitness played a role here. They mentioned how “fitness is measured indirectly,” and I didn’t really know what this meant. I also don’t understand how they actually GENERATED these games and what role CI played here. If they were just using random number generators, that doesn’t feel very intelligent to me. I would like to understand the programs and algorithms they wrote to actually develop inputs for the different axioms and constraints they put forth.