If everyone decides to do what, to specifically refrain from voting, or to do a sober cost-benefit analysis of voting? If people voted rationally, they would put far more value on participating in a civic rite and feeling like their voice is heard, and very little value on the prospect of changing election outcomes. Perhaps many people already think this way, explaining why about half of eligible voters don't bother. If no one cared about rites and signaling, far fewer people would vote, and at some point an individual vote would have enough potential power to make it worthwhile to do the research and make the effort to vote, so the system would not collapse. Can you give an example of this outside of voting?Your one vote is arguably not worth the effort, yet if everyone decides that, it has great consequence.
Doing something for the public good often looks irrational on an individual basis.
The same if conditions apply; the specific answer depends on circumstances. Is your friend elderly? Do they socialize a lot? Do you embrace, or maintain distance? Do you spend a lot of time close together talking? In the kitchen or on a breezy porch? It is rational to do things that benefit the public good (at reasonable cost). How could this not be the case, if someone is not a predator or parasite? We are the public, we benefit from a healthy public.
Caplan is 49 years old and in apparent good health. There is little risk* that COVID-19 will affect his personal health, whatever his behavior. *EDIT: little risk, relative to other risks that he already accepts, such as driving a car or drinking sugary soda. How is it rational for him to buy and use gloves for grocery shopping, if not to reduce the risk of spreading disease to others? I'm not sure I understand what you mean by generalizing. His behavior is to become informed about the risks and make carefully calculated, emotionally neutral decisions that balance the positive and negative consequences of his choices. I would like to see that pattern generalized. His specific choices (on restaurants, or gloves) are tailored to his individual situation and preferences, and should not be generalized. Note that Caplan does not pretend to have all the answers. Admitting to confusion, he points out the value in relaxing behavior and restrictions to measure the risk that comes with a more open posture. A more effective and ethical approach might be paid voluntary human experimentation to improve our understanding of the risks.