Pandemic is a cooperative board game for 2 to 4 players, in which the players are tasked with getting four different diseases that plague the world under control. Players move around a world map, treat diseases, build research stations and ultimately have to find cures for the four diseases. While playing the game players constantly have to weigh their options: If they spread out they have an easier time treating diseases, but finding cures often requires multiple players to be collocated.
We are currently developing AI agents that play Pandemic with a human player. These agents observe the actions of the human player, and determine which goal they are likely to be pursuing. Using this information the agent can form its own plan to augment the human player’s. The goal is for the agent to collaborate with the human player with whatever task they are currently focusing on, but also avoid duplicating efforts. In this paper we discuss in more detail what makes Pandemic an interesting challenge for AI research, and how we are approaching the game.