Search

Monitoring is ubiquitous in the gig economy wherein the workforce is geographically dispersed. However, workers are often reluctant to be monitored because of privacy concerns, resulting in a hidden economic cost for employers as workers demand higher wages for monitored jobs. We investigate how three common dimensions of monitoring affect workers’ willingness to accept monitored jobs through online experiments on two gig economy platforms. The three dimensions of monitoring are intensity (how much information is collected), transparency (whether the monitoring policy is disclosed to workers), and control (whether workers can remove sensitive information). We find that, as the monitoring intensity increases, workers become less willing to accept monitoring because of elevated privacy concerns. Furthermore, being transparent about the monitoring policy increases workers’ willingness to accept monitoring only when the monitoring intensity is low. Interestingly, providing control over high-intensity monitoring does not significantly reduce workers’ privacy concerns, rendering this well-intentioned policy ineffective. Finally, females are more willing to accept monitored jobs than males as they perceive higher payment protection from monitoring and have lower privacy concerns. On average, the hourly wage compensation required for gig workers to accept monitoring is 1.6∼1.8 dollars, which amounts to roughly 28.6%∼37.5% of their average hourly wage.

This paper provides a novel theoretical angle and robust empirical evidence demonstrating that the auction duration and the item description length are two essential auction design parameters that can function as screening mechanisms for bidder quality on online service outsourcing platforms. These outsourcing platforms use buyer-determined reverse auctions to find service providers. Using data from a major online outsourcing platform, we examine the effects of the auction duration and the item description length on both bidder entry (i.e., the number of bids and bidder quality) and contract outcomes (i.e., whether a project is contracted and the buyer’s expected utility from the winning bid) based upon the project- and bidder-level analyses. We find that auctions with longer durations and item descriptions attract more bids, but they also attract disproportionately more low-quality bidders, creating a double whammy of higher evaluation costs and adverse selection for buyers. This, in turn, leads to less successful contracting as well as lower buyer utility. Our research highlights the screening role of the auction duration and item description length for buyers on online service outsourcing platforms: by shortening auction durations and item descriptions, buyers can expect higher quality bidders, increase contracting probability, and enhance utility.

Artificial intelligence (AI) systems have become increasingly popular in many areas. Nevertheless, AI technologies are still in their developing stages, and many issues need to be addressed. Among those, the reliability of AI systems needs to be demonstrated so that the AI systems can be used with confidence by the general public. In this paper, we provide statistical perspectives on the reliability of AI systems. Different from other considerations, the reliability of AI systems focuses on the time dimension. That is, the system can perform its designed functionality for the intended period. We introduce a so-called SMART statistical framework for AI reliability research, which includes five components: Structure of the system, Metrics of reliability, Analysis of failure causes, Reliability assessment, and Test planning. We review traditional methods in reliability data analysis and software reliability, and discuss how those existing methods can be transformed for reliability modeling and assessment of AI systems. We also describe recent developments in modeling and analysis of AI reliability and outline statistical research challenges in this area, including out-of-distribution detection, the effect of the training set, adversarial attacks, model accuracy, and uncertainty quantification, and discuss how those topics can be related to AI reliability, with illustrative examples. Finally, we discuss data collection and test planning for AI reliability assessment and how to improve system designs for higher AI reliability. The paper closes with some concluding remarks., 40 pages

We theorize peer awards’ effects on the volume and novelty of creative user-generated content (UGC) produced at online platform communities. We then test our hypotheses via a randomized field experiment on Reddit, wherein we randomly and anonymously assigned Reddit’s Gold Award to 905 users’ posts over a two-month period. We find that peer awards induced recipients to make longer, more frequent posts and that these effects were particularly pronounced among newer community members. Further, we show that recipients were causally influenced to engage in greater (lesser) exploitation (exploration) behavior, producing content that exhibited significantly greater textual similarity to their own past (awarded) content. However, because the effects were most pronounced among new community members, who also produce content that, in general, is systematically more novel than that of established members to begin with, this process yields a desirable outcome: larger volumes of generally novel UGC for the community. This paper was accepted by Chris Forman, information systems.

Residents often rely on newspapers and television to gather hyperlocal news for community awareness and engagement. More recently, social media have emerged as an increasingly important source of hyperlocal news. Thus far, the literature on using social media to create desirable societal benefits, such as civic awareness and engagement, is still in its infancy. One key challenge in this research stream is to timely and accurately distill information from noisy social media data streams to community members. In this work, we develop SHEDR (social media–based hyperlocal event detection and recommendation), an end-to-end neural event detection and recommendation framework with a particular use case for Twitter to facilitate residents’ information seeking of hyperlocal events. The key model innovation in SHEDR lies in the design of the hyperlocal event detector and the event recommender. First, we harness the power of two popular deep neural network models, the convolutional neural network (CNN) and long short-term memory (LSTM), in a novel joint CNN-LSTM model to characterize spatiotemporal dependencies for capturing unusualness in a region of interest, which is classified as a hyperlocal event. Next, we develop a neural pairwise ranking algorithm for recommending detected hyperlocal events to residents based on their interests. To alleviate the sparsity issue and improve personalization, our algorithm incorporates several types of contextual information covering topic, social, and geographical proximities. We perform comprehensive evaluations based on two large-scale data sets comprising geotagged tweets covering Seattle and Chicago. We demonstrate the effectiveness of our framework in comparison with several state-of-the-art approaches. We show that our hyperlocal event detection and recommendation models consistently and significantly outperform other approaches in terms of precision, recall, and F-1 scores. Summary of Contribution: In this paper, we focus on a novel and important, yet largely underexplored application of computing—how to improve civic engagement in local neighborhoods via local news sharing and consumption based on social media feeds. To address this question, we propose two new computational and data-driven methods: (1) a deep learning–based hyperlocal event detection algorithm that scans spatially and temporally to detect hyperlocal events from geotagged Twitter feeds; and (2) A personalized deep learning–based hyperlocal event recommender system that systematically integrates several contextual cues such as topical, geographical, and social proximity to recommend the detected hyperlocal events to potential users. We conduct a series of experiments to examine our proposed models. The outcomes demonstrate that our algorithms are significantly better than the state-of-the-art models and can provide users with more relevant information about the local neighborhoods that they live in, which in turn may boost their community engagement.

Search and experience goods, as well as vertical and horizontal differentiation, are fundamental concepts of great importance to business operations and strategy. In our paper, we propose a set of theory-grounded data-driven measures that allow us to measure not only product type (search vs. experience and horizontal vs. vertical differentiation) but also sources of uncertainty and to what extent consumer reviews help resolve uncertainty. We used product rating data from Amazon.com to illustrate the relative importance of fit in driving product utility and the importance of search for determining fit for each product category at Amazon. Our results also show that, whereas ratings based on verified purchasers are informative of objective product values, the current Amazon review system appears to have limited ability to resolve fit uncertainty. Industry practitioners could utilize our approaches to quantitatively measure product positioning to support marketing strategy for retailers and manufacturers, covering an expanded group of products.

This study examines the role of text-based direct messaging systems in online labor markets, which provide a communication channel between workers and employers, adding a personal touch to the exchange of online labor. We propose the effect of workers’ use of the direct messaging system on employers’ hiring decisions and conceptualize the information role of direct messaging. To empirically evaluate the information role of the direct messaging system, we leverage data on the direct messaging activities between workers and employers across more than 470,000 job applications on a leading online labor market. We report evidence that direct messaging with a prospective employer increases a worker’s probability of being hired by 8.9%. However, the degree to which workers benefit from direct messaging is heterogeneous, and the effect amplifies for workers approaching employers from a position of disadvantage (lacking tenure or fit with the job) and attenuates as more workers attempt to message the same prospective employer. The effects also depend on message content. In particular, we find that the benefits of direct messaging for workers depend a great deal on the politeness of the workers, and this “politeness effect” depends on several contextual factors. The beneficial effects are amplified for lower-status workers (i.e., workers lacking tenure and job fit) and workers who share a common language with the employer. At the same time, the beneficial effects weaken in the presence of typographical errors. These findings provide important insights into when and what to message to achieve favorable hiring outcomes in online employment settings.