7 激励(incentive)对策及其应用在多级递阶决策问题中,上级(主导)决策人为了更好地控制下级的行为,常常不单是宣布一个固定的决策量 u_L,而是宣布一个如何根据下级的实际行动而灵活反应(例如奖惩)的策略。由于上级可能只拥有关于下级行动 u_F 的部份信息 z_L=η(u_F),因此这种策略的一般形式为配 u_L=γ_L(z_L)=γ_L〔η(u_F)〕。为了叙述简明起见,以下我们亦可简单地假定 z_L=u_F,只是这时 γ_L 只能在较窄的类,即可以写成 η(·)的复 7 incentive (incentive) and its application In multi-level hierarchical decision-making, in order to better control the behavior of the subordinate (superiors) decision-makers, often not only to declare a fixed decision volume u_L, but announced a how to The strategy of reacting flexibly (such as rewards and punishments) to actual actions of subordinates. Since the superordinate may only have partial information about subordinate action u_F, z_L = η (u_F), the general form of this strategy is u_L = γ_L (z_L) = γ_L [η (u_F)]. For the sake of simplicity, we can simply assume that z_L = u_F, except that γ_L can only be found in narrower classes, that is, it can be written as a complex of η (·)