VCVincent E. Castillo

Vincent E. Castillo

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Vincent E. Castillo has a total of 15 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Do Humans and GAI See Eye to Eye? Implications of LLM Scoring Volatility in Supplier Evaluations

    by Finnegan A. McKinley, Anne E. Dohmen, Vincent E. Castillo

    This study compares Generative Artificial Intelligence (GAI) to human procurement professionals on supplier evaluation tasks. Using Structural Topic Modeling (STM) on 123 government supplier bids from 31 projects solicited by the State of Ohio between January 2023 and December 2024, we compare evaluations from three reasoning models (o3, Grok‐3‐Mini, DeepSeek R1‐0528) against human evaluators. Adopting a signaling theory perspective, we find asymmetry in signal processing between GAI and human evaluators. GAI demonstrates high consistency and strong human alignment when evaluating compliance signals (e.g., technical specifications), which makes it suitable for qualification screening. However, GAI exhibits high scoring volatility with competitive signals (e.g., value‐add propositions), indicating that human judgment remains critical for assessing differentiation. We also find that the number of bidders influences signal composition, with compliance signals more prevalent in less competitive solicitations. The findings suggest a two‐stage evaluation framework where GAI handles compliance screening and humans focus on competitive assessment. GAI scoring volatility serves as a canary‐in‐the‐mine to identify when human oversight is necessary.


  • Open Access

    SCM-Arena: A Behavioral Benchmark for LLM Decision-Making in Supply Chains

    by Abhinav Hasija, Vincent E. Castillo

    SCM-Arena defines a reproducible behavioral evaluation protocol for LLM decision policies in delayed-feedback supply chain systems. Version 1 is frozen as of January 31, 2026 and comprises 75 models, 144 conditions per model, 5 replications per condition, and over 53,000 total episodes. Unlike knowledge-based benchmarks that test isolated question-answering, SCM-Arena places models in the role of ordering agents across a four-tier supply chain simulation derived from the Beer Game with varying demand patterns, information visibility, decision memory, and prompt framing conditions. Performance is assessed through cost ratio relative to a human baseline, demand amplification (bullwhip effect), and behavioral descriptors of ordering coordination and decision complexity. Aggregated results are publicly accessible at scm-arena.com. This paper describes the protocol design, metrics, and platform.

  • Open Access

    Last Mile Delivery Capacity Planning With Two‐Sided Uncertainty

    by Vincent E. Castillo, Marc E. Posner, Anníbal C. Sodero, Walter Zinn

    Crowdsourced delivery (CD) is increasingly combined with private delivery (PD) to respond to last mile delivery (LMD) demand uncertainty. As independent contractors, CD drivers self‐schedule, which simultaneously creates labor supply uncertainty. Consequently, such LMD capacity planning requires addressing both uncertainties concurrently, a duality we term “two‐sided uncertainty.” In this setting, two key questions emerge: what is the optimal level of PD capacity when CD is available? How does two‐sided uncertainty impact the PD capacity decision? To that end, we adapt a two‐stage newsvendor model to explore the impact of two‐sided uncertainty on optimal PD capacity when CD is available. We combine Monte Carlo simulation and optimization in empirically grounded experiments using data from a Brazilian prepared meals company. We find a positive relationship between two‐sided uncertainty and PD capacity. We also reveal nuance in the relationship by identifying moderating effects due to correlation in driver availability across CD price tiers and failed delivery costs. The results also indicate that optimal delivery capacity typically combines PD and CD sources rather than relying on a single source. We show that managerial decisions based solely on lowest unit delivery cost while assuming away two‐sided uncertainty can lead to suboptimal decisions.


  • Improving Driver Engagement in Delivery and Rideshare Services

    by Lance W. Saunders, Vincent E. Castillo, William J. Rose, Anne E. Dohmen, John E. Bell

    Due to driver autonomy, delivery and rideshare platforms in the sharing economy (SE) face uncertainty in service capacity planning. This study investigates how the alignment between the technical and social subsystems within the overall socio‐technical system of SE platforms influences driver engagement. Drawing on socio‐technical systems theory, we conceptualize SE platforms as systems defined by the organizational design constructs of centralization, formalization, and complexity. Using a discrete choice experiment with 100 experienced SE drivers, we examine how platform design attributes affect drivers' utility and engagement probability. Our findings reveal that drivers value flexibility, preferring platforms that offer direct task allocation and single‐task acceptance features. Contrary to expectations, drivers generally favor base pay–oriented remuneration over promotion‐oriented structures. Individual driver characteristics such as SE income dependency and multi‐apping tendencies significantly influence platform preferences. The findings contribute to logistics research by empirically establishing the relationship between platform structure and driver utility and by extending the application of organizational design theory to digital labor platforms. Our results provide insights into how firms can reduce capacity uncertainty by engaging SE drivers within their networks, highlighting the importance of aligning platform design with the preferences of a heterogeneous driver population.


  • Open Access

    Designing technology for on‐demand delivery: The effect of customer tipping on crowdsourced driver behavior and last mile performance

    by Vincent E. Castillo, Diane A. Mollenkopf, John E. Bell, Terry L. Esper

    Crowdsourcing technology platforms specializing in on‐demand last mile delivery face a novel problem—heightened agent independence increases uncertainty in last mile delivery and has the potential to undermine operational performance. Uncertainty emanates from drivers' competing interests, or opportunity costs. Rather than increase remuneration, some platforms have designed the technology to allow customers to tip, but the subsequent effects on driver behavior and last mile performance have not yet been studied. We explore this feature of on‐demand delivery technology design—customer tipping—and its implications for crowdsourced delivery performance. Using netnography and secondary data sources, we empirically ground an adaptive, multi‐agent hybrid simulation to propose how customer tipping impacts service performance through context‐contingent driver behavior. Our results generally indicate that tipping as a technology design feature mitigates uncertainty in the crowdsourced delivery fleet and reduces fulfillment times and unit delivery costs. We also find evidence suggesting that the impact of tipping on driver behavior is geography‐contingent, with tipping having unexpectedly detrimental effects in high population density areas relative to low‐density areas. We conclude by making a series of theoretical propositions and managerial implications about designing technology to facilitate crowdsourced delivery.


  • Journeys, Not Destinations: Theorizing a Process View of Supply Chain Integrity

    by Matthew A. Douglas, Diane A. Mollenkopf, Vincent E. Castillo, John E. Bell, Emily C. Dickey

    Integrity is considered an important corporate value. Yet recent global events have highlighted the challenges firms face at living up to their stated values, especially when extended supply chain partners are involved. The concept of Supply Chain Integrity (SCI) can help firms shift focus beyond internal corporate integrity, toward supply chain integrity. Researchers and managers will benefit from an understanding of the SCI concept toward implementing SCI to better align supply chain partners with stated corporate values. This research fully develops and empirically grounds the firm-level, inter-firm-oriented SCI concept. The thematic analysis of six firms’ archival and website content elaborated empirical descriptions of SCI themes and enabled the development of a process model for SCI, presenting a novel view of the underlying process by which firms can assess, develop, and maintain SCI across their supply chains. We propose the SCI model as an evolutionary process to improve a firm’s supply chain sustainability, rather than a dichotomous end state where firms either “have” integrity or they don’t. The SCI model could be used as a tool to help leaders create necessary change to better align values and supporting statements with culture, while influencing and affecting stakeholders across the supply chain. This is particularly important in today’s world, where business leaders must consider all stakeholders and address important stakeholder-driven issues such as supply chain sustainability, resilience, and security, which are now at the forefront in the ever-changing environment.


  • Hybrid last mile delivery fleets with crowdsourcing: A systems view of managing the cost‐service trade‐off

    by Vincent E. Castillo, John E. Bell, Diane A. Mollenkopf, Theodore P. Stank

    Retail firms now offer same‐day delivery via hybrid fleets that augment a privately owned delivery fleet with crowdsourced assets. As research on crowdsourced delivery continues to grow, it has implicitly assumed deliveries being conducted with an entirely crowdsourced fleet. In this study, we adopt a sociotechnical systems perspective to consider the phenomenon of combining privately owned delivery vehicles with crowdsourced assets into a hybrid fleet. In this exploratory effort, we examine how elements of the hybrid delivery system, namely, driver autonomy, compensation, fleet size, fleet mix, and demand intensity affect cost and service in last mile delivery. Using a multimethod simulation combining agent‐based and discrete event techniques with home delivery data from a major US retail pharmacy, we show how the emergent outcomes of this hybrid delivery system are a consequence of the constituent elements’ interdependence. Our experiments suggest a convex relationship between crowdsourced driver compensation and cost performance, such that low and high compensation amounts may actually increase unit delivery costs relative to a median compensation level, and a nonlinear negative relationship between compensation and fulfillment time. We find that these effects are moderated by fleet design (in terms of size and vehicle type mix) and order arrival rate intensity.