Robust Recourse for Binary Allocation Problems
TL;DR: We present the problem of counterfactual explanation and robust recourse for (sequential) binary allocation problems under distribution of the allocation parameters: resources, utility function and population.
Abstract: We present the problem of algorithmic recourse for the setting of binary allocation problems. In this setting, the optimal allocation does not depend only on the prediction model and the individual's features, but also on the current available resources, decision maker's objective and other individuals currently applying for the resource.
Specifically, we focus on 0-1 knapsack problems and in particular the use case of lending.
We first provide a method for generating counterfactual explanations and then address the problem of recourse invalidation due to changes in allocation variables. Finally, we empirically compare our method with perturbation-robust recourse and show that our method can provide higher validity at a lower cost.
Submission Track: Full Paper Track
Application Domain: None of the above / Not applicable
Clarify Domain: The presented use case is within finance, but allocation problems can be relevant to many domains.
Survey Question 1: Automated decision-making systems are used in many applications that affect people's lives. The literature mainly provide explanations and recourse with respect to a prediction model, while the decisions are often made as allocation problems due to limited resources (such as a budget constraint when approving loans). Hence, we provide a method for providing explanations and recourse for allocation problems.
Survey Question 2: Our goal is to provide users with agency regarding automated decisions that affect their lives. Thus, it is important to produce an explanation and algorithmic recourse which is reliable (with respect to the entire decision-making process) and valid with high probability with respect to possible changes in the future.
Survey Question 3: We assume access to such methods, but encapsulate it in our pipeline such that the choice is left for the decision maker.
Submission Number: 83
Loading