1. Trusting trustworthy people increases efficiency. The less checking and verification you have to do, the less redundant work there is and the more easily things go.

2. As the number of trustworthy people in a group increases, the value of choosing to trust a random person increases. The odds get greater that they will be trustworthy, letting you reap the greater efficiency of trusting them.

3. As the number of trusting people in a group increases, the relative value of malice increases. The more likely a con artist is to be trusted, the more likely they are to be successful at their goal.

So there's this weird trust graph where as you approach 100% trustworthiness, you also approach 100% exploitability. The group gradually turns itself into a honeypot. So the stable equilibrium point is somewhere less of 100%.