Figure 2.

Identity and liking ratings of 100 personality traits by 68 participants (on a 1–7 Likert scale). Darker blue indicates more frequent observations, which can be clearly seen to fall on the diagonal, indicating that subjects usually identify with the traits they like, and vice versa. Of the 100 traits each subject rated, an algorithm selected four subject-specific triplets of adjectives that best fit into the four corners of this “identification-by-liking” space. Black circle sizes are proportional to the number of adjectives that were selected by the algorithm. During the subsequent games, subjects were matched with others that either identified or didn't identify with the triplet of traits selected by the algorithm.