In order to move forward from our catchbob environment, I need a more complex task, where there should be a mapping between a problem space and the space visited by the participants. A good example would be a cluedo-like task where people need to gather cues all around the campus (finding the weapons used for a crime). Here the mapping is simple: a cue found at a certain location...inference...this cue shows that a specific behavior has occured in this place...inference...However, Pierre argues that this is not semantically complex enough...