Social: Theories and Models - Foundations for Designing User-Centered Systems - page 252

Table 9.1 A typical prisoner's dilemma payoff matrix

Prisoner 1 Prisoner 2

Cooperate
(with prisoner 1)

Defect
(on prisoner 1)

Cooperate (with prisoner 2) Both prisoners get 1 year

in jail

10 years for prisoner 1
Prisoner 2 goes free

Defect (on prisoner 2) 10 years for prisoner 2

Prisoner 1 goes free

Both prisoners get 8 years in jail

against the other. These two decisions or strategies are usually called cooperate
(with the other prisoner) or defect (on the other prisoner). The payoff if the two
prisoners cooperate with each other and plead not-guilty is that they both go to
prison for 1 year. If one prisoner defects and testifies against the other, they walk
free but the other prisoner serves 10 years. If both prisoners defect, they share the
full blame and get 8 years each.
Where there are costs and benefits involved in carrying out shared tasks, the
costs should be kept as low as possible (not necessarily minimized), and the
benefits should be kept as high as possible (not necessarily maximized). In any
trade-off the benefits should be designed to outweigh the costs. If you think of
online file sharing and email, for example, these have both radically reduced the
costs of sharing information, sometimes with unanticipated effects, such as too
much sharing.
Using a payoff matrix approach can be helpful in representing tasks where
social interaction is involved, even if there is no dilemma involved. Thus a matrix
can be used to show the payoff for asking questions in class, for responding to
emails, and so on. The payoff matrix approach can be applied to situations that
involved more than two strategies, more payoffs, and more players, although it
becomes harder to draw the matrix as the numbers involved increase. You should
also note that the players do not need to have equivalent payoffs: for example, the
payoffs for teachers and for students in a class are different.
Note that some matrices do not have a single best choice (when there is a stable
choice that maximizes the payoff for both parties, it is known as a Nash equilib-
rium point). If you have played Paper Rock Scissors, you know that paper covers
rock, that rock breaks scissors, and that scissors cuts paper—there is not a choice
that beats all the others. Similarly, if you play video games, you can create payoff
matrices for different pieces playing against each other, for example, in Command
and Conquer, for tanks vs tanks, tanks vs infantry, and tanks vs planes, etc. In each
case, there may not be a dominant decision or choice that is better than all others.
Axelrod ( 1984 ) studied what happens when the PD game is played multiple
times, also called an iterated Prisoner's Dilemma game. When the game is played
repeatedly there is a chance for cooperation to emerge. If you defect early, your
colleague may stop trusting you. If you cooperate longer, some opponents may not
defect. If the game is ongoing, the promise of cooperation with a positive payoff