Designing effective security service contracts presents a critical challenge for both Managed Security Service Providers (MSSPs) and their clients, driven by heterogeneous client risks, interdependent security externalities, and post-adoption client behaviors. Employing a game-theoretic framework, this research investigates an MSSP’s optimal contract strategy among three regimes: a fixed compensation contract (F-Contract), a proportional compensation contract (R-Contract), and a menu-based contract regime (M-Contract) that offers both options simultaneously, allowing clients to self-select. The differences in compensation structures between the F-Contract and R-Contract directly shape client behavior, resulting in varying levels of post-adoption client negligence. The menu-based contract design in our setting introduces a richer mechanism than simply segmenting heterogeneous clients: clients’ strategic migration across contract options changes the aggregate security of the protected network and, consequently, the MSSP’s optimal market coverage and profitability. As security loss risk increases, high-valuation clients migrate from fixed to proportional compensation. When this migration reduces aggregate negligence, it offsets the intensified negative externalities and induces the MSSP to expand, rather than contract, its client base and profits. Crucially, this reversal cannot arise under either pure contract regime or in screening models without contract-dependent behavior. We further show that no single regime always maximizes market coverage, MSSP profitability, or social welfare, as the optimal choice depends on the security loss risk and the negligence gap between the F- and R-Contracts. Specifically, when both factors are moderate, the M-Contract yields a “win-win-win” equilibrium that simultaneously advances the interests of the MSSP, clients, and social planners, requiring no external policy intervention. However, we also find that an MSSP’s privately optimal contract does not always inherently maximize social welfare. To address this, we characterize the conditions that call for targeted regulatory or contractual interventions to realign private and social incentives.
eHealth behavioral interventions have transformed how individuals manage their health and modify their lifestyles. Despite their growing popularity, many users gradually reduce or discontinue their participation over time. To understand this disengagement, we extend the Expectation-Confirmation Theory (ECT) by modeling user engagement as a dynamic learning process, where evolving perceptions of intervention effectiveness are shaped by ongoing experience. Leveraging a hierarchical Bayesian learning framework, we analyze users’ perception updates and how such dynamics influences engagement decisions. Our empirical results show that individuals’ learning performance appears to be lower for interventions with ambiguous instructions or those focused on short-term health outcomes. These interventions tend to generate noisier feedback that may hinder users’ ability to form accurate perceptions of intervention effectiveness, which may ultimately reduce sustained engagement. Given that eHealth behavioral interventions typically possess credence characteristics, where effectiveness may be difficult for users to evaluate directly, the need for clear, accessible information becomes especially critical. To help improve user learning and engagement, we propose several denoising strategies and evaluate them through counterfactual simulations. Our work extends ECT into healthcare settings and provides actionable insights for designing more supportive Health IT systems that foster informed decision-making and sustain user engagement.
Generative AI holds great promise for reshaping software development. We examine the impact of GitHub Copilot, a generative AI pair programmer, on collaborative open-source software (OSS) development, where developers voluntarily contribute and collaborate on projects. Using GitHub’s proprietary Copilot usage data combined with public OSS data from GitHub, we find that Copilot increases project-level code contributions by expanding developer participation and increasing code contributed by individual developers. However, these gains lead to increase in time to coordinate, review and integrate code. Overall, Copilot still increases the amount of code that is integrated in a timely manner, but the benefits are unevenly distributed. Core developers (highly active contributors) gain more from Copilot, whereas contributions from peripheral developers (occasional contributors) require relatively more coordination time. These findings suggest that as AI increasingly reduces the cost of producing code, coordination and integration may become more important constraints in software development. Organizations and OSS communities should therefore complement AI adoption with governance structures and coordination processes that facilitate code integration and support peripheral developers. More broadly, as AI increasingly automates routine coding activities, organizational processes and team structures may become critical determinants of the returns to generative AI adoption.
Physical retailers have long lacked the real-time behavioral visibility that online platforms enjoy through clickstream data. This research addresses that gap by introducing a video analytics framework that transforms in-store security camera footage into a rich, structured behavioral record: an "offline clickstream." Using computer vision and deep learning techniques, including person re-identification, trajectory reconstruction, pose estimation, and vision-language models, the system extracts moment-by-moment signals of shopper intent: how customers move through the store, how they interact with products, and how their body language evolves during a visit. A transformer-based prediction model trained on these signals achieves dramatically better purchase prediction accuracy than conventional demographic or contextual benchmarks alone: improving predictive performance by up to 79% on key metrics. Beyond prediction, the framework supports five real-time targeting policies; simulations show that a persuadability-based policy yields a 13.1% profit lift over no targeting. For retailers and policymakers, this research offers a scalable, privacy-conscious blueprint for bridging the capability gap between physical and digital commerce, enabling timely, personalized interventions that improve customer experience and store profitability.
Digital experimentation faces a sample size planning problem: the typically small treatment effects require a very large number of participants to be reliably detected, and yet the large number of simultaneous experiments being conducted by many platforms means that the number of available users to participate is often limited. Meanwhile, common practice often involves rules of thumb such as setting experiments to run for two weeks, which can result in substantial over- or undersampling, depending on unknown parameters of the distribution of the focal outcome variable. To address this problem, we propose a novel framework and method for sequentially sampled experiments with multiple treatment arms, where the objective is to determine the best treatment arm and whether that arm provides a practically significant improvement over the current business-as-usual platform condition. Our approach is unique among sequentially sampled best-arm identification methods in allowing experimenters to specify separate error rates for different failure conditions and separate effect-size margins for selection and testing tasks. We provide extensive conceptual and empirical comparisons with a suite of comparison methods that can serve as a useful guide to practitioners. Our method is freely available in the R package seqbest.
