DISC Research Seminar Series

Problem definition: This study explores the impact of initial and in-process waiting times in asynchronous telemedicine on subsequent online and offline visits, and the moderating effects of the consultation fee. Methodology/results: We focus on three measures of waiting time in asynchronous telemedicine: initial waiting time (the duration from a patient’s initial request to admission into an online consultation session), average in-process waiting time (the average time between a patient’s question and the doctor’s response during the session), and the variability of in-process waiting times, which together capture both the initial and ongoing responsiveness to patients during asynchronous care delivery. We use 42,111 patients’ online and offline consultation records from a primarily text-based, asynchronous telemedicine platform affiliated with a top-ranked hospital system (February 2021-April 2024). Our results show that patients with a longer (above median) average in-process waiting time (≥0.75 hours) have 14.53%, 16.47%, and 13.41% lower odds for subsequent all visits, online visits, and outpatient visits in the next 30 days, respectively. Patients with a higher (above median) variability of in-process waiting times (≥0.40 hours) have 9.70% and 12.72% lower odds of subsequent all visits and offline outpatient visits in the next 30 days, respectively. Surprisingly, the initial waiting time shows no significant effect. Results remain consistent when considering whether the subsequent visits are with the same doctor. Finally, the consultation fee negatively moderates the relationship between in-process waiting time and subsequent visits. Managerial implications: Average in-process waiting time in asynchronous telemedicine has the most significant impact on both subsequent online and offline patient visits among the three waiting time metrics. The findings highlight reducing in-process waiting in asynchronous telemedicine as a viable means of enhancing patient engagement and ensuring continuity of care across channels.

TBD

Retailers are shifting from brick-and-mortar to omnichannel systems. Ship-from-store (SFS) is a prominent yet risky integration because its value depends on product attributes and local shopping frictions: the same program can expand demand for some SKUs while cannibalizing store sales for others. Using transaction and inventory-order data from a large pharmacy chain, we document SKU-level patterns and analytically model how store-only customers and omni-customers choose between visiting the store and ordering online, incorporating shortened online waiting time under SFS and different cross-selling profits by channel. The model’s equilibrium links product characteristics and local geography to demand pooling, in-store fill rates, and profit, yielding testable predictions that separate market expansion from cannibalization and inform SKU selection and inventory targets. Testing these predictions across pre- and post-SFS periods, we find: (i) stores with higher store visit costs realize larger post-SFS profit gains (ii) SKUs with lower baseline in-store availability see larger improvements and (iii) among cannibalization-prone SKUs, order profitability still rises when post-SFS inventory targets are sufficiently high. Together, the model and evidence explain when SFS raises profit and why effects vary across products and locations.

Rising hospitalization rates and costs underscore the need to better understand how health outcomes are shaped by individuals’ social determinants of health, such as lifestyle and socio-economic factors. We propose a novel framework that leverages large-scale, population-level consumer location data to continuously capture granular individual behavior. By combining unsupervised learning with sequential deep learning models, our approach characterizes lifestyle patterns and quantifies their association with future hospitalizations. Applied to 45 million location records from a major U.S. metropolitan area, the framework uncovers diverse lifestyle patterns and their health implications. Our results show that lifestyle choices are stronger predictors of hospitalization risk than socio-economic status, community context, or healthcare accessibility. Importantly, individuals from lower-income communities or with limited access to healthcare can still maintain healthy lifestyles, highlighting opportunities to reduce health disparities through patient self-management. We also find that individuals with busy, varying work routines and limited gym visits are twice more likely to be hospitalized within a year compared to the average. Notably, the regularity of healthy behaviors, rather than their total duration, emerges as a key factor in reducing future hospitalization risk.

The advance of artificial intelligence (AI) technologies has enabled chatbots to be emotionally responsive. In customer service, a critical component of employee emotional responsiveness is to express empathy, which also serves as a popular service recovery strategy. Still, the impact of chatbot-expressed empathy is less clear. In this research, we investigate how and why empathy expressed by a chatbot after service failure may influence customers’ service evaluations differently compared to that expressed by a human. Extending the psychological reactance literature, we propose that an empathic chatbot may trigger customer reactance due to the unique nature of AI chatbots (vs. humans). We further propose that the triggered reactance can undermine perceptions of a chatbot’s competence and warmth, ultimately hurting service evaluations. We conducted three laboratory experiments to test these predictions. Our theoretical framework and findings illuminate the more nuanced and distinct role of empathy expressed by AI chatbots, and they suggest a critical need to be aware of user reactance when AI chatbot developers design empathic chatbots and when companies consider deploying such bots in customer service. Co-authors: Elizabeth Han and Han Zhang

A substantial proportion of purchases in the fashion industry are returned, representing around $890 billion in merchandise annually.  This study investigates how different discount strategies influence net sales, and in particular, return rates. Our analysis focuses on three key factors influencing the effectiveness of discount strategies in enhancing net sales: customers’ uncertainty about a product’s post-purchase value, the post-purchase leverage effect of bundle discounts, and customer inattention to pricing details. Collaborating with one of Turkiye’s largest fashion retailers, we use structural estimation to evaluate bundle discounts in comparison to per-item discounts. Our findings reveal that net sales under bundle discounts can outperform per-item discounts by 15.61% when the post-purchase leverage effect is utilized effectively. However, bundle discounts may underperform when this leverage is absent or poorly implemented. Additionally, we find that customers often exhibit inattention to pricing details, resulting in an overestimation of a product bundle’s value at the time of purchase, which subsequently contributes to higher return rates.

Firms increasingly invest in new technologies to boost productivity, yet little is known about how well workers’ skills align with evolving production systems. We develop a new empirical methodology—leveraging large language models (LLM's) with matched data from worker resumes and firm job postings—to construct novel measures of skill mismatch across firms that utilize different information technologies (IT). We document three sets of findings. First, skill mismatch exhibits a U-shaped pattern as IT evolves: mismatch is high when new technologies are deployed, declines over time as firms and workers adjust, and rises again as technologies become obsolete. Second, mismatch is substantial not only for technical skills, but also for complementary non-technical skills such as managerial ability. Third, skill mismatch is inversely related to firm profitability and total factor productivity over the technology lifecycle. These patterns are consistent with theories of technological change in which skill adjustment lags drive heterogeneity in firm-level productivity. Our results highlight the central role of skill alignment between workers and firms in the diffusion of new technologies. Co-authors: Ashwini Agrawal, London School of Economics Daniel Kim, University of Waterloo

After briefly talking about Information System Research and some recent editorials, I will offer some thoughts about digitalization and its impacts, using the example of digital classrooms in schools. I will highlight the utopian and dystopian narratives on digitalization in the literature and argue that both narratives are incomplete, polarizing, and not particularly helpful. Indeed, Information Systems scholars may be in a unique position to offer a balanced perspective and positively influence digitalization's role in organizations and society. I will end my presentation with some implications for research and practice.

This paper studies a long-term power purchase agreement (PPA) between a firm and a new renewable energy generator. The firm must dynamically satisfy uncertain electricity demand beyond its existing energy sources, while wholesale electricity prices evolve stochastically over time. Upon signing a PPA, a new renewable facility becomes operational, and the firm owns its output for the contract duration. The new facility’s capacity is determined based on PPA terms. The firm dynamically chooses when to initiate the PPA and how much to pay to maximize its expected total discounted benefit. We show that the firm’s optimal timing follows a (time-dependent) threshold policy. Our results offer key insights for policymakers and renewable energy developers. We find that, contrary to common wisdom, reducing investment costs for renewable technologies can lead to smaller renewable capacity, output, and emissions savings when projects are developed under PPAs. This finding calls for caution in applying investment tax credits in such contexts. We also show that total renewable energy generation and emissions savings may decrease with higher site productivity. Therefore, restricting renewable facility development to most productive sites might be counterproductive under PPAs. We establish the robustness of our findings across a broad range of practical scenarios.

This paper examines how firms design their global supply networks in response to perceived trade policy effect uncertainty (TPEU). Drawing on organizational information processing theory and analyzing 33,807 firm-year observations from 2007 to 2022, we document a U-shaped relationship between TPEU and supply network interconnectedness (SNI). This pattern suggests that firms adopt distinct supply network strategies depending on the degree of perceived uncertainty. Specifically, under low TPEU, firms reduce SNI by relying on less connected but more trusted suppliers. In contrast, under high TPEU, firms increase SNI by establishing redundant ties to enhance flexibility. Furthermore, we find that this U-shaped relationship is attenuated for firms with greater digital capability or higher levels of organizational slack. Our results remain robust across a range of tests addressing potential endogeneity concerns.

Artificial Intelligence has seen tremendous successes over the years and is expected to lead to an era of augmented intelligence systems that can leverage both human and AI capabilities in a seamless manner. Yet, this is not without challenges and getting there requires a focus on intentional design. This talk presents an augmented intelligence principle regarding the effectiveness of such systems and uses this principle to highlight key issues including the search for optimal designs and how to do so. Drawing on this perspective we discuss complex systems simulations as a possible alternative and illustrate its value through examples.

Technology sharing to rivals and new-product introductions enabled by those technologies have often been observed across different industries. We develop a gametheoretic model to examine why a firm would voluntarily share its technology and help its rival develop a new product. We find that the cannibalization consideration in the rival’s multiproduct pricing imposes externality on the focal firm, which largely gives rise to its incentive to share technology, in addition to the potential benefit from the change in demand elasticity. Surprisingly, the rival does not always embrace the shared technology. In equilibrium, as long as the existing product valuation is not too high, the new product would be introduced when the new product’s valuation is neither low nor high. When the new product’s valuation is high, the excessive additional competition against the focal firm discourages it from sharing. When the new product’s valuation is low, either cannibalization does not emerge in the rival’s pricing, providing no incentive for the focal firm to share, or the rival could be harmed by a new product with mediocre valuation. We show that social welfare increases with the new-product introduction to a large extent except when the existing product’s valuation is high but the new product’s valuation is relatively low. The new-product introduction increases consumer surplus only when the existing product valuation is low. Compared to technology sharing to an independent third party, the focal firm is more likely to share its technology to the rival.

Keywords:
technology sharing, new-product development, cannibalization externality, spokes model