I found a quite interesting paper called Not Polyannas: High Generalized Trust Predicts Lie Detection Ability by two researchers at the University of Toronto.
It’s about an experiment where participants first make a web survey that measures their generalized trust, ie if they generally trust other people.
People vary in their trusting propensities: Some, who assume that others are generally trustworthy, make themselves vulnerable to their counterparts until evidence challenges their trustworthiness assumptions; others assume that people are generally untrustworthy and act accordingly until their counterparts demonstrate their trustworthiness gradually over time.
The paper says the following about the reasons behind the observed result:
Yamagishi (2001) has theorized that generalized trust is a form of social intelligence that can be highly adaptive, counter to game theory’s predictions. His model suggests that high trusters, who take more social risks and are, therefore, more vulnerable to exploitation, obtain more differentiating social data and learn more, e.g., ‘‘Ah, this is what someone who will deceive me does.’’ In contrast, by defending themselves from possible exploitation, low trusters seem to be suspicious of everyone: They send signals that limit the development of potentially beneficial relationships and, therefore, in the absence of differentiating social data, they learn less about distinguishing trustworthy from untrustworthy others. Thus, by defending themselves from the costs associated with exploitation, low trusters can incur potentially massive opportunity costs.
…three potential adaptive explanations for a positive relationship between generalized trust and social intelligence . First, high generalized trust drives social risk taking, and the possibility of exploitation pushes high trusters to invest in learning how to identify people who are not trustworthy. Low trusters need no such skills because a social posture of defensiveness is a reliable (if costly) exploitation prophylactic. Second, advanced sensitivity to trustworthiness cues reduces a person’s vulnerability to detrimental consequences. Those who are less sensitive are better off assuming that unknown others are generally untrustworthy, leading to less generalized trust among the less socially intelligent. Assuming that people are liars prevents a person from being duped. In contrast, being effectively sensitive makes it safe to assume that others generally tell the truth because this sensitivity will help detect a lie before a person falls victim to it.