In addition, we require a theory of how to act on relevant streetlevel knowledge. I presume here that this theory of decision is a simple commonsense learning theory. You may start with such limited information about me that you can only estimate the likelihood that the typical person in my position would be trustworthy with respect to what you might entrust to me. You might even have such limited information about me that you can only assess from your past experience whether cooperation has paid off in similar circumstances. Suppose it did, and now you risk cooperating with me. (This is not to say you choose to trust me. Rather, gambling on me seems to you the rational thing to do.) You either gain or lose from your cooperation, and this experience is added to your evidence on trustworthiness for future occasions. If I am somehow a new kind of person in your experience, your initial estimate may be unstable, and my behavior might tilt your assessment heavily for or against my kind in future encounters.

Russell Hardin, Trust and Trustworthiness, loc. 536