Generalizability

Random selection from a population is what earns the right to generalize to that population. Without random selection, you can only generalize to a population of individuals similar to those actually studied. For example, a well-run randomized experiment on 80 student volunteers at one university supports a cause-and-effect claim for people like those volunteers, but not a claim about all college students. That is why the two randomizations get judged separately: assignment supports causation, and selection supports generalization.