This is liable to be a multi-part post, because it seems like there is a lot of ground to cover. I have a lot to say about The Sims. It bears noting first that the game I am most familiar with is not the original, but rather its sequel The Sims 2, which is also broadened by a variety of expansion packs. Generally, when I talk about The Sims, I am thinking of the whole franchise. As a game it is extremely notable, but also relevant to my research because of several factors:

1. It is a people simulator. Its subject matter is mundane and domestic, about people going about their lives in a day to day manner.
2. The AI that controls the Sims is ferociously dumb. The Sims act according to a hill climbing algorithm in relation to the world around them. The technical implementation of this is called “smart terrain” and all the logic is encoded within the objects with which the Sims interact.
3. The game is exceptionally evocative. The setting and characters are believable in a way that makes them engaging and fun to play with.
4. It broke the gender barrier: More than half the players of the Sims are female. This is especially true of the machinima community. (A citation would be helpful here)
5. The Sims has a powerful modular architecture that enables it to be modded easily, and is supplementable through expansion packs.
6. The Sims is the best selling PC game of all time.

All of these factors make the game extremely significant in the landscape of PC games everywhere, especially given the immense popularity and uniqueness of The Sims franchise. I want to explain here why it is so relevant to the work I am doing now.

The first and foremost reason why The Sims is relevant is because of the AI. The Sims themselves are believable, but not realistic. Characters have a state that is built around specific domain models (relationships, needs, “skills”, etc) rather than propositional models. What is fascinating about this is the sheer lack of material that could have been encoded into the model, what is more, the game works better because of this absence. Consider some of the things normally important to AI that have been left out. Beliefs and world knowledge is one example. Sims are totally autonomous. They can behave fine in one space just as well as another. Instead, all of the real intelligence is representational. The Sims are not believable because of an accurate or realistic model, but rather because of an evocative representational model.

There is an intricate balance in the game between simulation and representation. But, before it is possible to explore that in more detail, it is necessary to examine what is actually being simulated. The Sims is a people simulator. It represents domestic life, but as is the case with all Maxis games, it is a specific flavor of domestic life. The original game of the Sims presented a model of a materialistic suburban life. However, this model has been expanded in subsequent expansion packs. These expansions expand the complexity of the underlying model by introducing new logical elements. The Sims 2: Seasons is about how weather affects people’s moods and lives. The Sims 2: Free Time explores the social and personal dimensions of interests and hobbies. It will require a careful analysis to examine exactly what is being modeled, but some insight can be gained by looking into how the modeling works.

The original game of The Sims was heavily influenced by the logic of Maslow’s heirarchy of needs. The implementation in The Sims places each element: needs, relationships, skills in terms of sliding numeric values. This approach of representing things numerically can be considered The Sims modeling strategy. This approach is effective but also leaves out a great amount of detail. The strategy can convey simple relationships (between entities and concepts) in a domain. Notably, this representation strategy cannot represent complex relationships where there is a formal structure between entities. An example of this sort of complexity is in human relationships when there is internal complexity and occasional self contradictions. Another example of this flaw can be seen in the single/dual axis representations of morality in games (Fable, D&D).

The work that I am doing uses a very different modeling strategy. My strategy is much more symbolically oriented and structural, because my goal is to represent the sorts of relationships that are impossible with numerical sliders. Specifically, human behavior and relationships are modeled using Goffman’s framework revolving around roles and performance, which are fuzzy and complex. The real issue at steak is how to increase logical complexity without undermining representational power.