The present study investigated effects of rules on misbehavior of computer simulations. Five hundred seventy-six cells were arranged on a 24 × 24 matrix. Each cell had 1 of 2 states (meaning “behavior”) —“obeying” or “breaking” a rule— and one of various decision matrices expressing attitudes toward misbehavior. Decision matrices were composed of 2 × 2 variables of M11, M12, M21, and M22. For example, M11 stands for the degree of cells’ own satisfaction when both they and their neighbor cells (that is, cells adjoining them in a matrix) obey a rule. M12 represents the degree of satisfaction when they obey a rule while their neighbors break it. If M11 + M21 is higher than M12 + M22, it means cells want their neighbors to obey a rule (“obeying-orientated”), and if M11 + M21 is lower than M12 + M22, it means they want them to break it (“breaking-orientated”). The cells changed their own states according to 1 of 4 rules: A1) “Coexistence”: A cell changes its own state to make the sum of degrees of its own and a neighbor’s satisfaction highest; A2) “Altruism”: A cell changes its own state to make a neighbor’s satisfaction highest; A3) “Egoism”: A cell changes its own state to make its own satisfaction highest; and B) “Considering behavior”: A cell changes its own state to make its own satisfaction highest considering its neighbors’ states (behaviors). Rules A1 and A2 refer to one neighbor’s decision matrix (attitude), and rule B considers many neighbors’ states (behaviors). The simulation outputs were rule-breaking frequencies. The results showed that output increases when the number of “breaking-orientated” decision matrices is large for all 4 rules; especially, the “altruism” rule was the most sensitive to this number.