Archive: August 26th, 2009

The Sims 3

[Games,Research] (08.26.09, 3:24 pm)

Generally, my point of interest is in comparison to other Sims games. My interest here is not the way it plays exactly, but the changes in the simulation. These changes are primarily the new systems of character traits and moodlets. These are promising from the perspective of useful character simulation.

I don’t mean to suggest that I want to adopt the mechanics and model that are used in The Sims 3, but rather, I mean to illustrate that these mechanics solve a major problem in simulating agents that are believable as characters. Sims characters have always been only somewhat believable. The characters are dolls: they are not characters, but rather they are objects from which we can interpret characters. The work here lies in the player’s imagination.

My problem with The Sims has always been that it is very difficult to create fundamentally different characters. In both the Sims 1 and 2, characters were differentiated in that they had several numeric stats. These are: sloppy/neat, shy/outgoing, lazy/active, serious/playful, and grouchy/nice. The Sims 2 expanded this by adding another layer on top of this model, which is aspirations and fears. It was thus possible to create several conceptually interesting characters, such as a very shy person who aspires to popularity. However, with an intention to create a specific character, or create characters referenced from fiction, the system of statistics is awkward.

Furthermore, the statistical model itself is difficult to apply to other domains. The example I tend to consider in these situations is Jane Austen’s Pride and Prejudice. Considering the main cast of characters, we can come up with values for them in the sloppy/neat sorts of parameters. However, we cannot distinguish exactly what it is that makes Mr. Collins horrible in these statistics. A possibility would be to extend the statistical model, and for instance, have numeric statistics that represent the spectrum of differences between the characters: for instance, there might be an axis of gentlemanliness, one of stuffiness, or rakishness, but these are clumsy and artificial. What’s worse is that these parametric systems create dangerous middle zones: what does it mean to be halfway between rakish and gentlemanly? Most frustratingly, it conceals and muddles the complex relationship between the inner and surface lives and natures of the characters.

While the mechanics present in The Sims 3 does not provide a solution to this problem, it offers a pleasing shift in perspective. The trait system offers several traits that sims can possess. Each sim may have 5 traits, plus a lifetime aspiration. Traits may be things indicated by the earlier sliding-value system. For instance, a character may be neat or a slob. But the trait system also distinguishes between several more subtle social qualities. A character could be flirty, friendly, or charismatic. Each trait represents something slightly different. These give more detail and subtlety to characterization. There are also other sorts of social traits, which guide how the characters might act in social contexts. My favorite of these is “inappropriate” which causes (permits, really) sims to do things that are socially inappropriate.

The traits system also defines how sims respond to the world around them. In a simple example: a sim who is neat will get more pleasure out of being clean than a slob, but a slob will not mind dirty surroundings as much as a neat sim. These differences are manifested mechanically through moodlets. Moodlets last some period of time, and usually have a positive or negative effect on a single mood variable. The character’s overall mood is thus the sum of a base value and all of the sim’s moodlets. What is more, an NPC might respond favorably or unfavorably to the actions of a player controlled sim. The NPC behavior is controlled by their own traits and moodlets, which are usually opaque to the player. This creates a very simplified resemblance of an inner-life.

One could imagine a more robust authoring system whereby it is possible to author new types of traits, moodlets, and all of these other details, which could be authored modularly. So, with the earlier example, it would be possible to create a Mr. Darcy by having traits of “gentleman,” “proud,” and “shy.” Austen has a trend wherein in each of her books there is a “talky” character, which could be manifested as a trait. Similar sorts of traits could be attributed to every character. This solves the issue of ambiguous in-between values, and also opens up a method for thinking about the characters not just in Austen’s works, but in other character-driven story worlds. As I mentioned earlier, I don’t think that the Sims 3 mechanics are exactly what I want to use, as, after all, Mr. Darcy changes considerably over the course of Pride and Prejudice, and the system as described is not capable of representing complex relationships, but it is a good system to keep in mind.

Reading Info:
Author/EditorElectronic Arts
TitleThe Sims 3
Tagsai, games, simulation, social simulation
LookupGoogle Scholar, Google Books, Amazon