If the amount of violence is proportional to the fraction of a population that is under high stress, then three security strategies become obvious: removing the stressed people from the population, reducing stress on the people who are experiencing most of it, and reducing the stress on the entire population.
The most humane way to remove the stressed people corresponds to what I earlier called “isolation” (the flip-side, separating the population from the stressed people, is what I called “retreat”). The most inhumane way to do this is what I called “offense,” with “defense” somewhere in-between. Reducing the stress on the stressed people corresponds to what I called “alliance.” Reducing stress on the entire population is a critical requirement for what I referred to as “assimilation.” It should be noted that a trivial but historically viable alternative to “removing” stressed people is to keep populations artificially small (less than 250), so that statistically (under most conditions) the number of murderers is less than one.
Offense, defense, isolation and retreat can now be analytically shown to be the most inadequate solutions to the problem of violence. They simply delay the onset of violence until either conditions become more stressful or another generation is born, members of which having a low tolerance for the existing stress.
Reducing stress on individuals is only slightly better, requiring a complex social structure that must shift with conditions and necessarily increase stress on other members of the population (who must interact directly with the affected people). It is the fallback position when the stress on the entire population cannot be effectively reduced.