In the course of this project we have developed a framework to both advance the state of the art of how to build believable agents (by incorporating social identity and anticipation), and allow social scientists to use the game associated with this framework to study the formation of partnership between agents (both human and artificial) in a virtual environment.
The proposed framework models a specific type of games, “team games” in which social identity assumes a more important role than in usual social dilemmas. In this type of games, social identity plays an important role since groups are not only formed but also have specific out-group competitive goals and, according to the previous experiment, it appear to have a significant impact upon cooperation rates.
It is, therefore, the purpose of this work that this framework allows designing environments where social identity and its impact upon the outcomes of problematic social settings can be studied.
The framework was implemented in multi-player game that takes place in a 3D environment, where both humans and virtual agents participate. After the story introduction, players begin the game, as their plane crashed on a deserted island. Since there is an active volcano at the centre of the island, they need to search for wood in order to build a raft and escape. As such, everyone agrees to collect wood on a daily basis so that the raft may be built. However, as each player finds out when exploring the island alone, gems can also be found scattered all over the island. The passengers are then faced with the dilemma of either helping everyone by collecting wood or gathering gems and thus become rich when they are saved: if everyone collects mainly gems then the raft will not be built in time and everyone will die when the volcano erupts.
Although this description explains the overall game, several variations are to be considered in order to study their impact on the behaviour of players.