In the first set of experiments, we adopt the model from Section 4.2 where agents For simplicity, we refer to the first setting as the "constrained agent" case, and the To begin, we verify our theoretical findings from Section 4.2. First we let the decision-maker lead and the agents follow, and then we switch the roles. In this section, we verify our findings on a model where the decision-maker's problem is the same In particular, the agents' risk R takes the form: R ( µ,θ) = λ 2 null µnull We remark that though this setup is conceptually very similar to that in Section 4.2 (increasing In our experiments we once again let the decision-maker lead and the agents follow, and then we switch the roles. In Figure 4 we empirically observe that there is gap between the decision-maker's risk at their Stackelberg equilibrium and at the agents', and that the decision-maker consistently achieves a lower Further, we note that agents consistently prefer leading, meaning that both the decision-maker and agents prefer if the order-of-play is flipped. Our empirical results suggest that the agents' equilibrium is a strictly better equilibrium in terms of the social cost (defined classically in game theory as the sum of the agents' and In Figure 5 we again observe that the proposed dynamics converge to the Stackelberg equilibria.