arXiv subjects

Scott Rodilitz

Publications and source records attributed to Scott Rodilitz.

5 recordsLinked to original sources

arXiv2022

Online Algorithms for Matching Platforms with Multi-Channel Traffic

Two-sided platforms rely on their recommendation algorithms to help visitors successfully find a match. However, on platforms such as VolunteerMatch (VM) -- which has facilitated millions of connections between volunteers and nonprofits -- a sizable fraction of website traffic arrives directly to a nonprofit's volunteering page via an external link, thus bypassing the platform's recommendation algorithm. We study how such platforms should account for this external traffic in the design of their recommendation algorithms, given the goal of maximizing successful matches. We model the platform's problem as a special case of online matching, where (using VM terminology) volunteers arrive sequentially and probabilistically match with one opportunity, each of which has finite need for volunteers. In our framework, external traffic is interested only in their targeted opportunity; by contrast, internal traffic may be interested in many opportunities, and the platform's online algorithm selects which opportunity to recommend. In evaluating different algorithms, we parameterize the competitive ratio based on the amount of external traffic. After demonstrating the shortcomings of a commonly-used algorithm that is optimal in the absence of external traffic, we propose a new algorithm -- Adaptive Capacity (AC) -- which accounts for matches differently based on whether they originate from internal or external traffic. We provide a lower bound on AC's competitive ratio that is increasing in the amount of external traffic and that is close to (and, in some regimes, exactly matches) the parameterized upper bound we establish on the competitive ratio of any online algorithm. We complement our theoretical results with a numerical study motivated by VM data that demonstrates the strong performance of AC and furthers our understanding of the difference between AC and other commonly-used algorithms.

cs.GT↗

arXiv2021

Fair Dynamic Rationing

We study the allocative challenges that governmental and nonprofit organizations face when tasked with equitable and efficient rationing of a social good among agents whose needs (demands) realize sequentially and are possibly correlated. As one example, early in the COVID-19 pandemic, the Federal Emergency Management Agency faced overwhelming, temporally scattered, a priori uncertain, and correlated demands for medical supplies from different states. In such contexts, social planners aim to maximize the minimum fill rate across sequentially arriving agents, where each agent's fill rate is determined by an irrevocable, one-time allocation. For an arbitrarily correlated sequence of demands, we establish upper bounds on the expected minimum fill rate (ex-post fairness) and the minimum expected fill rate (ex-ante fairness) achievable by any policy. Our upper bounds are parameterized by the number of agents and the expected demand-to-supply ratio, yet we design a simple adaptive policy called projected proportional allocation (PPA) that simultaneously achieves matching lower bounds for both objectives (ex-post and ex-ante fairness), for any set of parameters. Our PPA policy is transparent and easy to implement, as it does not rely on distributional information beyond the first conditional moments. Despite its simplicity, we demonstrate that the PPA policy provides significant improvement over the canonical class of non-adaptive target-fill-rate policies. We complement our theoretical developments with a numerical study motivated by the rationing of COVID-19 medical supplies based on a standard SEIR modeling approach that is commonly used to forecast pandemic trajectories. In such a setting, our PPA policy significantly outperforms its theoretical guarantee as well as the optimal target-fill-rate policy.

cs.GT↗

arXiv2020

Commitment on Volunteer Crowdsourcing Platforms: Implications for Growth and Engagement

Volunteer crowdsourcing platforms match volunteers with tasks which are often recurring. To ensure completion of such tasks, platforms frequently use a lever known as "adoption," which amounts to a commitment by the volunteer to repeatedly perform the task. Despite reducing match uncertainty, high levels of adoption can decrease the probability of forming new matches, which in turn can suppress growth. We study how platforms should manage this trade-off. Our research is motivated by a collaboration with Food Rescue U.S. (FRUS), a volunteer-based food recovery organization active in over 30 locations. For platforms such as FRUS, success crucially depends on volunteer engagement. Consequently, effectively utilizing non-monetary levers, such as adoption, is critical. Motivated by the volunteer management literature and our analysis of FRUS data, we develop a model for two-sided markets which repeatedly match volunteers with tasks. Our model incorporates match uncertainty as well as the negative impact of failing to match on future engagement. We study the platform's optimal policy for setting the adoption level to maximize the total discounted number of matches. We fully characterize the optimal myopic policy and show that it takes a simple form: depending on volunteer characteristics and market thickness, either allow for full adoption or disallow adoption. In the long run, we show that such a policy is either optimal or achieves a constant-factor approximation. Our finding is robust to incorporating heterogeneity in volunteer behavior. Our work sheds light on how two-sided platforms need to carefully control the double-edged impacts that commitment levers have on growth and engagement. A one-size-fits-all solution may not be effective, as the optimal design crucially depends on the characteristics of the volunteer population.

cs.GT↗

arXiv2020

Online Policies for Efficient Volunteer Crowdsourcing

Nonprofit crowdsourcing platforms such as food recovery organizations rely on volunteers to perform time-sensitive tasks. To encourage volunteers to complete a task, platforms use nudging mechanisms to notify a subset of volunteers with the hope that at least one of them responds positively. However, since excessive notifications may reduce volunteer engagement, the platform faces a trade-off between notifying more volunteers for the current task and saving them for future ones. Motivated by these applications, we introduce the online volunteer notification problem, a generalization of online stochastic bipartite matching where tasks arrive following a known time-varying distribution over task types. Upon arrival of a task, the platform notifies a subset of volunteers with the objective of minimizing the number of missed tasks. To capture each volunteer's adverse reaction to excessive notifications, we assume that a notification triggers a random period of inactivity, during which she will ignore all notifications. However, if a volunteer is active and notified, she will perform the task with a given pair-specific match probability that captures her preference for the task. We develop an online randomized policy that achieves a constant-factor guarantee close to the upper bound we establish for the performance of any online policy. Our policy as well as hardness results are parameterized by the minimum discrete hazard rate of the inter-activity time distribution. The design of our policy relies on sparsifying an ex-ante feasible solution by solving a sequence of dynamic programs. Further, in collaboration with Food Rescue U.S., a volunteer-based food recovery platform, we demonstrate the effectiveness of our policy by testing them on the platform's data from various locations across the U.S.

cs.DS↗

arXiv2020

Snapshot Models of Undocumented Immigration

The Mexican Migration Project (MMP) is a study that includes samples of undocumented Mexican immigrants to the United States after their return to Mexico. Of particular interest are the departure and return dates of a sampled migrant's most recent sojourn in the United States, and the total number of such journeys undertaken by that migrant household, for these data enable the construction of data-driven undocumented immigration models. However, such data are subject to an extreme physical bias, for to be included in such a sample, a migrant must have returned to Mexico by the time of the survey, excluding those undocumented immigrants still in the US. In our analysis, we account for this bias by jointly modeling trip timing and duration to produce the likelihood of observing the data in such a "snapshot" sample. Our analysis characterizes undocumented migration flows including single visit migrants, repeat visitors, and "retirement" from circular migration. Starting with 1987, we apply our models to 30 annual random snapshot surveys of returned undocumented Mexican migrants accounting for undocumented Mexican migration from 1980-2016. Contrary to published estimates based on these same data, our results imply migrants remain in the US much longer than previously estimated based on analysis that ignored the physical snapshot bias. Scaling to population quantities, we produce lower bounds on the total number of undocumented immigrants that are much larger than conventional estimates based on US-based census-linked surveys, and broadly consistent with the estimates reported by Fazel-Zarandi, Feinstein and Kaplan (2018).

physics.soc-ph↗