Trust, Lies, and Social Intelligence

We can’t all be good at that.

Posted Feb 16, 2013

He said he would betray me, but he was lying.

~ Anonymous

Social behavior is governed by two powerful motives. One motive is to get along and belong. Social isolation is a dangerous thing. The other motive is to enhance and protect one’s relative status within the group (cf. Bakan's, 1966 distinction between communion and agency).. Social life is challenging because these two motives are in conflict. Any increase in the satisfaction of one motive puts a brake on the other. To achieve overall satisfaction, an individual must balance the two motives just right, and do so without solving several differential equations simultaneously. Things are even trickier because a person is surrounded by others who are trying to do the same thing. Any of their efforts to get along have a positive effect on the individual, whereas their attempts to gain status have a negative effect.

Once again, the much-abused prisoner’s dilemma captures the drama. In the minimal case, two individuals “play” by each choosing to either cooperate or to defect. The payoffs are ranked such that unilateral defection, which enhances the defector, is best, followed by mutual cooperation, which allows both to get along, followed by mutual defection, which protects both from exploitation, and finally unilateral cooperation, which suckers the player who wanted to get along.

When the interaction is not anonymous and when repeated play is likely, players can try to figure out the intentions of the others. Research says that prosocial individuals (trustors and cooperators) do this better than doubters and defectors (Brosig, 2002; Carter & Weber, 2010). This result makes sense. There are very few unconditional prosocials, or altruists. Most individuals who have prosocial values are willing to trust and cooperate unless there is reason to believe that the other person might betray them. These prosocials are conditional cooperators. Research also says that most people project their own values and intentions onto others (Krueger, 2007). That is, until there is evidence to the contrary, prosocials expect their partners to cooperate with them.

Expectations about the behavior of others are hypotheses. The task of social intelligence (or “social intelligence gathering”) is to see if there is any evidence regarding these hypotheses. A socially intelligent person will seek to detect lies or signals of impending betrayal particularly if this person already has a suspicion that it might be so. A person intending to cooperate but expecting defection is most vulnerable. A timely confirmation of this expectation permits a change in own behavior to avoid exploitation.

In contrast, individuals intending to defect are self-regarding. They need not ask what the other person might do. They accept the comfort of classic game theory, knowing that they will be better off defecting than cooperating regardless of what the other person does. They may hope that the other person cooperates so that their own defection amounts to self-enhancement, but this is no incentive to gather evidence. Defectors need no social skills.

The stage is set. Let’s play. Suppose ½ of the players plan to cooperate and ½ plan to defect; 2/3 of the cooperators and 1/3 of the defectors expect cooperation. Suppose that there is no accuracy in these predictions. Consider the cooperators. 1/3 of them correctly predict cooperation (hits) and 1/3 incorrectly predict it (false positives); 1/6 correctly predict defection (correct rejections) and 1/6 incorrectly predict it (misses). The correlation phi between prediction and reality is 0.

Recall that we see social intelligence in the confirmation of a hypothesis that others plan harm. On this assumption, the 1/6 of the cooperators whose predictions turn out to be correct rejections, will strategically switch to defection. They are, after all, conditional cooperators. With this 1/6 of the original group of cooperators gone, we are now left with 4/5 who predict cooperation and 3/5 who will actually be paired with other cooperators. In this remaining set of cooperators, the probability of a hit and a false positive is each 2/5, and the probability of a miss is 1/5. Over this set of final cooperators, the correlation (phi) between prediction and reality is -.408. The final group of cooperators is less socially intelligent than the initial one.

What about the defectors? The new arrivals of cooperators-turned-defectors amount to 1/7 of the final group of defectors. The probabilities are 1/7 each for a hit and a false positive, 2/7 for a miss, and 3/7 for a correct rejection. With phi = .091, the defectors as a group have become smarter, but not by as much as the cooperators have become dumber. If we give up the assumption of social projection, the asymmetry becomes even stronger (phi = -.5 and .167 respectively for the final group of cooperators and defectors). Social projection is an empirical fact, however, and once again, it turns out to be a protection against things being worse than they are.

So who is more socially intelligent? Today’s thought experiment suggests that initial intention, expectation, presence of clues and sensitivity to them all matter. Once strategic behavior has occurred, patterns of accuracy and inaccuracy can emerge or change at the group level, while doing little to offer sustainable inferences about the social intelligence of individual group members. Among the final group of defectors, for example, most individuals did nothing; only the minority of former cooperators showed social intelligence.

Finally, we can ask what happens to the alignment of players. With projection, 58.3% of the players end with partners choosing the same strategy (phi = .169); without projection, the alignment (62.5%) is even stronger (because more individuals switched; phi = .258).

Bakan, D. (1966). The duality of human existence: An essay on psychology and religion. Oxford, England: Rand Mcnally.

Brosig, J. (2002). Identifying cooperative behavior: some experimental results in a prisoner’s dilemma. Journal of Economic Behavior & Organization, 47, 275-290.

Carter, N. L., & Weber, J. M. (2010). Not Pollyannas: Higher generalized trust predicts lie detection ability. Social Psychological and Personality Science, 1, 274-279.

Krueger, J. I. (2007). From social projection to social behaviour. European Review of Social Psychology, 18, 1-35.