Fair Allocation through Selective Information Acquisition

Strategic screening can reduce unfair disparities in resource allocation. The paper demonstrates that when decision-makers selectively gather more information about certain applicants before making allocation decisions, they can reduce group-based disparities—particularly benefiting groups like those without traditional credit histories who typically face information disadvantages in lending decisions.

Targeting "marginal" candidates for additional screening is most efficient. Rather than screening everyone or screening randomly, the algorithm focuses on candidates whose decisions could plausibly change with better information. This targeted approach maximizes the value gained from limited screening budgets, making fair allocation more practical for resource-constrained institutions.

The approach improves both fairness and overall utility. Testing on real-world datasets showed that this selective information acquisition strategy simultaneously improved social welfare outcomes while addressing disparities, suggesting that fairness and efficiency need not be in direct conflict in allocation decisions.

Information gaps are a root cause of allocation disparities. The paper highlights that disparities across groups often stem from unequal information quality rather than actual differences in qualifications. By strategically filling these information gaps, institutions can make more equitable decisions without necessarily lowering standards.

The solution is computationally practical for real institutions. The algorithm reduces the problem to a series of linear programs that can be solved efficiently, making it feasible for banks and other institutions to implement in actual lending and resource allocation decisions.