This study explores public preferences between algorithmic and human decision-makers (DMs) in high-stakes contexts, how these preferences are impacted by performance metrics, and whether the public's evaluation of performance differs when considering algorithmic versus human DMs. Leveraging a conjoint experimental design, respondents (n = 9,000) chose between pairs of DM profiles in two scenarios: pre-trial release decisions and bank loan decisions. DM profiles varied on the DM’s type (human vs. algorithm) and on three metrics—defendant crime rate/loan default rate, false positive rate (FPR) among white defendants/applicants, and FPR among minority defendants/applicants—as well as an implicit (un)fairness metric defined by the absolute difference between the two FPRs. Controlling for performance, we observe a general tendency to favor human DMs, though this is driven by a subset of respondents who expect human DMs to perform better in the real world, and there is an analogous group with the opposite preference for algorithmic DMs. We also find that the relative importance of the four performance metrics remains consistent across DM type, suggesting that the public's preferences related to DM performance do not vary fundamentally between algorithmic and human DMs. Taken together, the results collectively suggest that people have very different beliefs about what type of DM (human or algorithmic) will deliver better performance and should be preferred, but they have similar desires in terms of what they want that performance to be regardless of DM type.

Motivated by our collaboration with a major refugee resettlement agency in the U.S., we study a dynamic matching problem where each new arrival (a refugee case) must be matched immediately and irrevocably to one of the static resources (a location with a fixed annual quota). In addition to consuming the static resource, each case requires post-allocation services from a server, such as a translator. Given the uncertainty in service time, a server may not be available at a given time, thus we refer to it as a dynamic resource. Upon matching, the case will wait to avail service in a first-come-first-serve manner. Bursty matching to a location may result in undesirable congestion at its corresponding server. Consequently, the central planner (the agency) faces a dynamic matching problem with an objective that combines the matching reward (captured by pair-specific employment outcomes) with the cost for congestion for dynamic resources and over-allocation for the static ones. Motivated by the observed fluctuations in the composition of refugee pools across the years, we aim to design algorithms that do not rely on distributional knowledge. We develop learning-based algorithms that are asymptotically optimal in certain regimes, easy to interpret, and computationally fast. Our design is based on learning the dual variables of the underlying optimization problem; however, the main challenge lies in the time-varying nature of the dual variables associated with dynamic resources. Our theoretical development brings together techniques from Lyapunov analysis, adversarial online learning, and stochastic optimization. On the application side, when tested on real data from our partner agency, our method outperforms existing ones, making it a viable candidate for replacing the current practice upon experimentation.

Ensuring that refugees and asylum seekers thrive (e.g., find employment) in their host countries is a profound humanitarian goal, and a primary driver of employment is the geographic location to which the refugee or asylum seeker is assigned. In the past few years, innovations in analytics have given rise to machine learning (ML) models that predict integration outcomes using personal characteristics. With these ML models, recent research has proposed and implemented algorithms that assign refugees and asylum seekers to geographic locations in a manner that maximizes the average employment. While these algorithms can have substantial overall positive impact (up to 50% increases in average employment rate compared with current practice), using data from two industry collaborators we show that the impact of these algorithms can vary widely across key subgroups based on country of origin, age, or educational background.

This study proposes two new dynamic assignment algorithms to match refugees and asylum seekers to geographic localities within a host country. The first, currently implemented in a multi-year pilot in Switzerland, seeks to maximize the average predicted employment level (or any measured outcome of interest) of refugees through a minimum-discord online assignment algorithm. Although the proposed algorithm achieves near-optimal expected employment compared to the hindsight-optimal solution (and improves upon the status quo procedure by about 40%), it results in a periodically imbalanced allocation to the localities over time. This leads to undesirable workload inefficiencies for resettlement resources and agents. To address this problem, the second algorithm balances the goal of improving refugee outcomes with the desire for an even allocation over time. The performance of the proposed methods is illustrated using real refugee resettlement data from a large resettlement agency in the United States. On this dataset, we find that the allocation balancing algorithm can achieve near-perfect balance over time with only a small loss in expected employment compared to the pure employment-maximizing algorithm. In addition, the allocation balancing algorithm offers a number of ancillary benefits compared to pure outcome-maximization, including robustness to unknown arrival flows and greater exploration.