When we are talking about an agent, we mean a virtual, independent entity, commonly used for user assistance or interaction. In order to make an agent more accessible for its users, many have attempted to make the agent more human-like; most of us will remember the (ever so annoying) paperclip in a certain office product, which had a pair of eyes so we would be able to identify with it more. The accessibility however seems to depend much more on another factor, namely the behavior the agent is showing. Once this behaviour is out of context, repetitive, too predictable or not matching the user’s experience, it will soon be considered annoying, boring or simply not useful. If an agent is showing (semi) intelligent behaviour however, it can constantly enrich the user’s knowledge by matching its experience and become endlessly more interesting and entertaining by showing intelligent and variable but yet decisive behaviour.
Video games are becoming increasingly realistic. Recent new game genres, like MMRPGs (Massive Multi player Role Playing Games) usually contain NPCs (Non Playing Characters) (characters controlled by the computer), which are showing a very limited behaviour repeatedly in an endless loop. A more dynamic, intelligent behaviour of these agents could give a new dimension to these games. Those of you who have ever played a strategy game (for example Command & Conquer) will agree that what’s referred to as “computer AI” in these games is so limited that the computer is only able to beat a human player by producing impossible numbers of units. A nicely coordinated attack or what we would refer to as an intelligent move (like luring out the enemy, or finding an alternative approach route to a strategic enemy building) is behaviour we are not yet seeing, but which could greatly improve the gaming experience and would possible even make the game a good training instrument for military or emergency services.