Behavioral Game Design

Every computer game is designed around the same central element: the player. While the hardware and software for games may change, the psychology underlying how players learn and react to the game is a constant. The study of the mind has actually come up with quite a few findings that can inform game design, but most of these have been published in scientific journals and other esoteric formats inaccessible to designers. Ironically, many of these discoveries used simple computer games as tools to explore how people learn and act under different conditions.

The techniques that I'll discuss in this article generally fall under the heading of behavioral psychology. Best known for the work done on animals in the field, behavioral psychology focuses on experiments and observable actions. One hallmark of behavioral research is that most of the major experimental discoveries are species-independent and can be found in anything from birds to fish to humans. What behavioral psychologists look for (and what will be our focus here) are general "rules" for learning and for how minds respond to their environment. Because of the species- and context-free nature of these rules, they can easily be applied to novel domains such as computer game design. Unlike game theory, which stresses how a player should react to a situation, this article will focus on how they really do react to certain stereotypical conditions.

What is being offered here is not a blueprint for perfect games, it is a primer to some of the basic ways people react to different patterns of rewards. Every computer game is implicitly asking its players to react in certain ways. Psychology can offer a framework and a vocabulary for understanding what we are already telling our players.

Contingencies and Schedules

The concrete translation of "What are we asking of our players?" is "What are our contingencies?" A contingency is a rule or set of rules governing when rewards are given out. The anecdote about this discovery (as passed to me by one of his students) is that one day B. F. Skinner ran low on the small food pellets he gave the rats in his experiments. Rather than risk running out and having to stop work for the day, he began to provide the pellets every tenth time the rats pressed the lever instead of every time. Experimenting with different regimens of reward, he found that they produced markedly different patterns of response. From this was born a new area of psychology, and one that has some strong implications for game design.

	Some common terms in behavioral psychology as they apply to game design considerations: Reinforcer: An outcome or result, generally used to refer to a reward. Examples: an experience point, winning a level, a bigger gun. Contingency: A rule or set of rules governing when reinforcers are given. Also referred to as a schedule of reinforcement. Examples: a level every 1,000 experience points, a bonus level that is only available if you kill a certain opponent. Response: An action on the part of the player that can fulfill the contingency. This could be killing a monster, visiting an area of the game board, or using a special ability.

The contingencies in computer games are more complex, but the analogy is clear enough. For example, players in an RPG earn experience points to gain levels or collect bonus items to gain extra lives. In an arcade-style game, power-ups appear at random intervals, or only when certain conditions are met. As in any contingency, there are actions on the part of the participant which provide a reward under specific circumstances. This is not to say that players are the same as rats, but that there are general rules of learning which apply equally to both.

Ratios and Intervals

There are essentially two fundamental sorts of contingencies, ratios and intervals. Ratio schedules provide rewards after a certain number of actions have been completed. For example, a player might gain an extra life after killing 20 opponents. This would be called a "fixed ratio" schedule, because the same number of kills is required every time. Other types of ratios will be discussed later.

One of the most common contingencies found in games, fixed ratio schedules typically produce a very distinct pattern in the participant. First there is a long pause, then a steady burst of activity as fast as possible until a reward is given. This makes sense when one considers that the very first action never brings a reward, so there is little incentive to make that first kill. Once participants decide to go for the reward, they act as fast as they can to bring the reward quickly.

The distinct pause shown under a fixed ratio schedule can be a real issue for game designers. Having a period of time where there is little incentive to play the game can lead to the player walking away. Additionally, the length of the pause is a function of the size of the ratio (the number of actions required), so the more actions required the longer the pause. This means that if the ratio increases over time, such as the increasing number of experience points required to gain a level in Dungeons & Dragons, so does the pause. Eventually, the pause can become infinite, and the player simply decides it's not worth it and walks away.

	As in any contingency, there are actions on the part of the participant which provide a reward under specific circumstances.

On the plus side, during the pause other, less rewarding activities often come to the fore. For example, if players know it will take them a long time to gain their next level, they might take the time to test a new tactic or try out different aspects of the game.

There are also "variable ratio" schedules, in which a specific number of actions are required, but that number changes every time. A player might be required to shoot down approximately 20 enemy fighters to gain an extra ship, but the precise number is randomly generated each time. It's important to note that the player does not know how many actions are required this time, just the average number from previous experience.

Under variable ratio schedules, participants typically respond with a steady flow of activity at a reasonably high rate. While not quite as high a rate as the burst under a fixed ratio schedule, it is more consistent and lacks the pausing that can cause trouble. Since it's possible (though unlikely) that the player can gain a life for shooting down only one enemy, there's always a reason to go hunting.

	Killing opponents to gain experience points and gain levels is one example of a ratio contingency.

In general, variable ratio schedules produce the highest overall rates of activity of all the schedules that I'll discuss here. This doesn't necessarily mean they're the best, but if what you're looking for is a high and constant rate of play, you want a variable ratio contingency.

On the other side of the coin there are interval schedules. Instead of providing a reward after a certain number of actions, interval schedules provide a reward after a certain amount of time has passed. In a "fixed interval" schedule, the first response after a set period of time produces a reward. For example, the game might introduce a power-up into the playing field 30 minutes after the player collected the last one.

Participants usually respond to fixed interval contingencies by pausing for a while after a reward and then gradually responding faster and faster until another reward is given. In our power-up example, the player would concentrate on other parts of the game and return later to see if the new power-up had appeared. If it hadn't, the player would wander off again. Gradually the checks would become more frequent as the proper time approached, until at about the right time the player is sitting there waiting for it.

As in the fixed ratio, there is a pause that can cause problems for a game designer. Unlike the fixed ratio, there is no sharp transition to a high rate of activity. Instead, there is gradual increase as the appropriate time approaches. The pause remains, a period where player motivation is low.

There are also "variable interval" schedules, where the period of time involved changes after each reward. A counterpart to the variable ratio schedules, these also produce a steady, continuous level of activity, although at a slower pace. As in the variable ratio schedule, there is always a reason to be active. The power-up mentioned in the earlier example could reappear immediately after being collected or an hour later. The motivation is evenly spread out over time, so there are no low points where the players' attention might wander. The activity is lower than in a variable ratio schedule because the appearance is not dependent on activity. If the player looks for the power-up 1,000 times during the interval, it will appear no faster. Experiments have shown that we are very good at determining which consequences are the results of our own actions and which are not.

These are the basic building blocks, but this is by no means an exhaustive list. Each contingency is an arrangement of time, activity, and reward, and there are an infinite number of ways these elements can be combined to produce the pattern of activity you want from your players.