FUNDAMENTALS OF GAME DESIGN, SECOND EDITION
Simulating Individuals
Many CMSs simulate the behavior of a group of people (or ants in the case of SimAnt) within an environment managed by the player. Games such as the original SimCity, which handle a large number of people, model behavior statistically rather than keeping separate values for each person. However, you might want to simulate the actions of particular individuals that the player can see moving around, as the modern versions of SimCity do. This will make your game a good deal more entertaining because the player can take an interest in the actions and progress of specific people. It appeals to a voyeuristic impulse and makes the consequences of the player's decisions seem more personal. It's particularly effective when the player can actually see unhappy people packing up and leaving.
Modeling individuals rather than statistical aggregates adds considerably to your design job. You will need to create a behavioral model and determine what aspects of the individual's condition the player will be trying to optimize. For example, many such games include a single-valued variable that tracks a character's degree of happiness or unhappiness and a set of needs that the simulated character desires to fulfill. Fulfillment may come as the result of the character's autonomous action (driving from home to work fulfills the need to get to work) or from action taken by the player (building a school fulfills the characters' need for educational opportunity). If a need goes unfulfilled, either through a problem that arises within the simulation (traffic jams prevent the person from getting to work) or because the player fails to act (no school has been built), there should be a negative consequence of some kind (the simulated person becomes unhappy).
Modeling individuals relieves you of the job of creating a statistical model because the behavior of the individuals collectively provides the statistics, but balancing such a game is a more intricate task. You will probably discover emergent behaviors; that is, unanticipated consequences of design decisions. Some of these will be fascinating and almost seem like intelligence, but others will clearly be degenerate: simulated people locked in a tight behavioral loop, for example, only ever doing one or two things because your needs mechanism isn't balanced properly.
Behavioral modeling is too big a subject for us to address comprehensively here. Consult the references for further reading.
