The global objectives of the game are established by design (for instance, destroy all aliens to prevent invasion, build and manage a city). These are neutral objectives that form the essence of the game. The game is often designed to assist the players in achieving these objectives—for example, using a mission briefing, tutorial, or inline help.
In many cases, however, the player has much freedom in the way the objectives are achieved (for instance, choosing paths and weapons in first-person shooters). In fact, some games enable the player to roam free of explicit objectives (for instance, in large-town simulations such as Grand Theft Auto III). There is always some freedom in deciding what to do and how to do it. This is the core of the gameplay experience.
For human players, the personal interpretation of the game's objectives is often instinctive, as if driven by feelings. Intelligent abilities allow the players to achieve their personal objectives, but responding to their emotions is usually the top priority.
In the previous part, high-level moods were modeled (for instance, aggressive and complacent). Desires and motivations can be derived from these moods. The examples in Table 44.1 serve as a basis for reinforcement-learning AI, developed later in this part.
The personal objectives can be considered in two ways, each with varying levels of abstraction. These desires can be expressed as follows:
The criteria can be considered a particular way of implementing a fitness function. Again, this subject is discussed in more depth for weapon selection in Chapter 22.
In psychological terms, humans often use emotions as means for guiding and evaluating intelligent behavior. However, this process can be considered as a hierarchy instead; a higher-level component—sometimes known as a reflective process—could decide which criteria to comply with. The hierarchy is a way to interpret and refine goals in an intelligent manner. In classical AI, the layers usually correspond to planning algorithms, although reactive techniques would be equally appropriate. Emotions instead bypass this hierarchy by providing a suitable approximation, which proves more realistic rather than intelligent (and more efficient for game AI).