In combinatorial allocation problems, indivisible objects have to be assigned to selfish agents. A standard assumption for those problems is that agents have quasilinear utility functions. However, in many environments either money cannot be exchanged or agents cannot be assumed to maximize payoff. We focus on two specific non-quasilinear environments. First, we analyze a course allocation problem where students have preferences over schedules and report on a large-scale course assignment application at the TU Munich. Second, we study non-quasilinear utility functions as they have been reported for display ad auctions, and propose a truthful randomized approximation mechanism.
«
In combinatorial allocation problems, indivisible objects have to be assigned to selfish agents. A standard assumption for those problems is that agents have quasilinear utility functions. However, in many environments either money cannot be exchanged or agents cannot be assumed to maximize payoff. We focus on two specific non-quasilinear environments. First, we analyze a course allocation problem where students have preferences over schedules and report on a large-scale course assignment applic...
»