Suppose we can choose between A: adding one person with negative utility −100, versus B: adding thousand people, each with small positive utility +1. If the critical level was fixed at say +10, then situation A decreases social welfare with 100, whereas B decreases it with 900, so traditional critical level theory indeed implies a sadistic conclusion to choose A. However, variable critical level utilitarianism can avoid this: the one person in A can choose a very high critical level for him in A, the thousand people in B can set their critical levels in B at say +1. Then B gets chosen. In general, people can choose their critical levels such that they can steer away from the most counterintuitive conclusions. The critical levels can depend on the situation and the choice set, which gives the flexibility. You can also model this with game theory, as in my draft article: https://stijnbruers.files.wordpress.com/2018/02/variable-critical-level-utilitarianism-1.pdf
Suppose we can choose between A: adding one person with negative utility −100, versus B: adding thousand people, each with small positive utility +1. If the critical level was fixed at say +10, then situation A decreases social welfare with 100, whereas B decreases it with 900, so traditional critical level theory indeed implies a sadistic conclusion to choose A. However, variable critical level utilitarianism can avoid this: the one person in A can choose a very high critical level for him in A, the thousand people in B can set their critical levels in B at say +1. Then B gets chosen. In general, people can choose their critical levels such that they can steer away from the most counterintuitive conclusions. The critical levels can depend on the situation and the choice set, which gives the flexibility. You can also model this with game theory, as in my draft article: https://stijnbruers.files.wordpress.com/2018/02/variable-critical-level-utilitarianism-1.pdf