Problem definition: We examine moviegoers’ choices between consuming content via legal theatrical channels and illegal piracy channels. We focus on how two important factors affect this choice: the picture quality of piracy sources (a function of studio security investments) and the costs associated with legal channels, including consumer transportation costs (a function of screening volumes) and ticket prices. Methodology/results: We formulate a structural model and conduct counterfactual simulations. Our findings indicate that consumers’ choice between legal and illegal channels is significantly influenced by the quality of pirated sources and the costs of legal consumption. The emergence of high-quality piracy sources in the first week of a movie’s theatrical release leads to a 7.9% reduction in theatrical revenue during the first eight weeks of release, as compared with a scenario where only low-quality pirated sources are available. We also find that high-quality piracy sources pose a greater threat to the sales of smaller movies than of blockbusters. Managerial implications: Our work provides a rare insight into the operational decisions faced by movie studios around the theatrical delivery of movies. Our counterfactual simulations explore the potential for movie studios to manipulate the supply, cost, and quality associated with legal content to mitigate piracy and increase profit. Our results indicate that the potential for cost reductions is limited, as changes in ticket prices or screen volume have minimal impact on legal consumption. However, a moderate improvement in the value or quality of the theatrical experience (e.g., via one-time investment by theaters in upgrading equipment or technology) can effectively offset the impact of high-quality pirated content. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0407 .

This paper theoretically studies financial lenders’ strategic decisions in providing preapproval checking tools, and shows such decisions could be asymmetric.

Despite lower accuracy in poor neighborhoods, Zillow’s automated home value estimates provide greater benefits to these housing markets by reducing more uncertainty.

This paper shows that personalized product rankings, although improving search relevance, can unintentionally enable AI pricing algorithms to raise prices and reduce consumer welfare.

Prior work on human-algorithmic bias has seen difficulty in empirically identifying the underlying mechanisms of bias because in a typical "one-time" decision-making scenario, different mechanisms generate the same patterns of observable decisions. In this study, leveraging a unique repeat decision-making setting in a high-stakes microlending context, we aim to uncover the underlying source, evolution dynamics, and associated impacts of bias. We first develop a structural econometric model of the decision dynamics to understand the source and evolution of bias in human evaluators in microloan granting. We find that both preference-based and belief-based biases exist in human decisions and are in favor of female applicants. Our counterfactual simulations show that the elimination of either of the two biases improves the fairness in financial resource allocation as well as the platform profits. The profit improvement mainly stems from the increased approval probability for male borrowers, especially those who would eventually pay back loans. Furthermore, to examine how human biases evolve when being inherited by machine learning (ML) algorithms, we train state-of-the-art ML algorithms for default risk prediction on both real-world data sets with human biases encoded within and counterfactual data sets with human biases partially or fully removed. We find that even fairnessunaware ML algorithms can reduce bias in human decisions. Interestingly, although removing both types of human bias from the training data can further improve ML fairness, the fairness-enhancing effects vary significantly between new and repeat applicants. Based on our findings, we discuss how to reduce decision bias most effectively in a human-ML pipeline.

FinTech lending (e.g., micro-lending) has played a significant role in facilitating financial inclusion. It has reduced processing times and costs, enhanced the user experience, and made it possible for people to obtain loans who may not have qualified for credit from traditional lenders. However, there are concerns about the potentially biased algorithmic decision-making during loan screening. Machine learning algorithms used to evaluate credit quality can be influenced by representation bias in the training data, as we only have access to the default outcome labels of approved loan applications, for which the borrowers' socioeconomic characteristics are better than those of rejected ones. In this case, the model trained on the labeled data performs well on the historically approved population, but does not generalize well to borrowers of low socioeconomic background. In this paper, we investigate the problem of representation bias in loan screening for a real-world FinTech lending platform. We propose a new Transformer-based sequential loan screening model with self-supervised contrastive learning and domain adaptation to tackle this challenging issue. We use contrastive learning to train our feature extractor on unapproved (unlabeled) loan applications and use domain adaptation to generalize the performance of our label predictor. We demonstrate the effectiveness of our model through extensive experimentation in the real-world micro-lending setting. Our results show that our model significantly promotes the inclusiveness of funding decisions, while also improving loan screening accuracy and profit by 7.10% and 8.95%, respectively. We also show that incorporating the test data into contrastive learning and domain adaptation and labeling a small ratio of test data can further boost model performance.

In high-stakes decision settings such as loan screening, predictive models often underperform for applicant groups that are underrepresented in the training data. We propose a data-centric learning framework that uses contrastive learning and domain adaptation to improve model generalization under selective labels. Specifically, we focus on micro-lending, where traditional machine learning algorithms for credit evaluation are trained primarily on approved loan applications and therefore observe default outcomes only for a selective subset of borrowers. Because these labeled samples tend to overrepresent borrowers with more favorable socioeconomic characteristics, standard models often generalize poorly to underrepresented applicants. To address this problem, we introduce a Transformer-based loan screening model that integrates self-supervised contrastive learning and domain adaptation. The model uses contrastive learning to train the feature extractor on unapproved (unlabeled) loan applications and uses domain adaptation to improve the transferability of the label predictor across labeled and unlabeled populations. We evaluate our approach on a real-world micro-lending dataset and demonstrate its effectiveness. The results show that our approach improves predictive performance for underrepresented borrowers while simultaneously increasing overall loan screening accuracy and lender profit by 7.10% and 8.95%, respectively. Additionally, we find that incorporating test data and labeling a small fraction of it further improves model performance.

Should firms that apply machine learning algorithms in their decision making make their algorithms transparent to the users they affect? Despite the growing calls for algorithmic transparency, most firms keep their algorithms opaque , citing potential gaming by users that may negatively affect the algorithm’s predictive power. In this paper, we develop an analytical model to compare firm and user surplus with and without algorithmic transparency in the presence of strategic users and present novel insights. We identify a broad set of conditions under which making the algorithm transparent actually benefits the firm. We show that, in some cases, even the predictive power of the algorithm can increase if the firm makes the algorithm transparent. By contrast, users may not always be better off under algorithmic transparency. These results hold even when the predictive power of the opaque algorithm comes largely from correlational features and the cost for users to improve them is minimal. We show that these insights are robust under several extensions of the main model. Overall, our results show that firms should not always view manipulation by users as bad. Rather, they should use algorithmic transparency as a lever to motivate users to invest in more desirable features. This paper was accepted by D. J. Wu, information systems. Supplemental Material: The online appendix is available at https://doi.org/10.1287/mnsc.2022.4475 .

Online social networks are increasingly being used to conduct commercial activities, and many online social networking platforms allow users to sell products to their online connections. Although extensive research has been conducted on the interactions among buyers within a social network, interactions among sellers have rarely been explored. Using seller data from a company that sells on a major online social networking platform in China, we empirically examine how sellers’ efforts and sales performance are affected by the efforts and sales performance of other sellers they are connected to (i.e., their inviters and invitees) and the commissions they themselves have received. We find evidence for social influence and competition effects in the “inviter-to-invitee” direction and sellers’ free-riding behavior driven by the commissions they receive from their invitees’ sales. These results extend the social network literature that has largely focused on connected buyers (or users) to connected sellers and offer implications for social networking platforms to promote seller participation.

In this paper, we empirically examine the impact of performance feedback on the outcome of crowdsourcing contests. We develop a dynamic structural model to capture the economic processes that drive contest participants’ behavior and estimate the model using a detailed data set about real online logo design contests. Our rich model captures key features of the crowdsourcing context, including a large participant pool; entries by new participants throughout the contest; exploitation (revision of previous submissions) and exploration (radically novel submissions) behaviors by contest incumbents; and the participants’ strategic choice among these entry, exploration, and exploitation decisions in a dynamic game. Using counterfactual simulations, we compare the outcome of crowdsourcing contests under alternative feedback disclosure policies and award levels. Our simulation results suggest that, despite its prevalence on many platforms, the full feedback policy (providing feedback throughout the contest) may not be optimal. The late feedback policy (providing feedback only in the second half of the contest) leads to a better overall contest outcome. This paper was accepted by Gabriel Weintraub, revenue management and market analytics department.

Machine learning (ML) algorithms used by financial lenders in their screening processes are hidden from the consumers who are affected by their decisions leading many consumers to make sub-optimal decisions when seeking credit. Despite increasing calls for greater transparency, only a few lenders provide personalized approval odds to consumers (e.g. via financial intermediaries like Credit Karma or pre-approval tools). We investigate how competition among algorithmic lenders affects their decisions to provide approval odds to consumers. We show that competitive pressures between lenders can undermine the disclosure incentives. Lenders use asymmetric disclosure of approval odds strategically to soften the competition when their algorithms are fairly accurate. The asymmetric disclosure of approval odds endogenously creates product differentiation and allows lenders to focus on different segments of consumers softening the competition on the interest rates. We find that consumer surplus is highest when both lenders provide approval odds and lowest when neither provides approval odds. However, our analysis also shows that any policy that mandates all lenders to provide personalized approval odds to consumers may not necessarily improve consumer surplus.

We develop a structural econometric model to capture the decision dynamics of human evaluators on an online micro-lending platform, and estimate the model parameters using a real-world dataset. We find two types of biases in gender, preference-based bias and belief-based bias, are present in human evaluators' decisions. Both types of biases are in favor of female applicants. Through counterfactual simulations, we quantify the effect of gender bias on loan granting outcomes and the welfare of the company and the borrowers. Our results imply that both the existence of the preference-based bias and that of the belief-based bias reduce the company's profits. When the preference-based bias is removed, the company earns more profits. When the belief-based bias is removed, the company's profits also increase. Both increases result from raising the approval probability for borrowers, especially male borrowers, who eventually pay back loans. For borrowers, the elimination of either bias decreases the gender gap of the true positive rates in the credit risk evaluation. We also train machine learning algorithms on both the real-world data and the data from the counterfactual simulations. We compare the decisions made by those algorithms to see how evaluators' biases are inherited by the algorithms and reflected in machine-based decisions. We find that machine learning algorithms can mitigate both the preference-based bias and the belief-based bias.

Automated pricing comes in two forms - rule-based (e.g., targeting or undercutting the lowest price, etc) and artificial intelligence (AI) powered algorithms (e.g., reinforcement learning (RL) based). While rule-based pricing is the most widely used automated pricing strategy today, many retailers have increasingly adopting pricing algorithms powered by AI. Q-learning algorithm (a specific type of RL algorithm) is particularly appealing for pricing because it autonomously learns an optimal pricing policy and can adapt to any evolution in competitors' pricing strategy and market environment. It is commonly believed that the Q-learning algorithm has a significant advantage over simple rule-based pricing algorithms; therefore, in a competitive environment, most firms should adopt Q-learning based pricing algorithms if their competitors are using such algorithms. However, through extensive pricing experiments in a workhorse oligopoly model of repeated price competition, we show that a firm's best response to its competitor's Q-learning based algorithms is to use simple rule-based pricing algorithms. We find that when a Q-learning algorithm competes against a rule-based pricing algorithm, higher prices are sustained in the market in comparison to when multiple Q-learning algorithms compete against each other. The high prices are sustained because the rule-based algorithm introduces stationarity into the repeated price competition, which allows the Q-learning algorithm to more effectively search for the optimal policy benefiting both sellers. Further, the experimental phase where the Q-learning algorithm learns the optimal pricing policy is significantly shorter when it competes against a rule-based pricing algorithm in comparison to when it competes against another Q-learning algorithm. Our results are robust to alternative modeling assumptions on market structure, algorithm type, number of players, etc.

Big data and machine learning (ML) algorithms are key drivers of many fintech innovations. While it may be obvious that replacing humans with machine would increase efficiency, it is not clear whether and where machines can make better decisions than humans. We answer this question in the context of crowd lending, where decisions are traditionally made by a crowd of investors. Using data from Prosper.com, we show that a reasonably sophisticated ML algorithm predicts listing default probability more accurately than crowd investors. The dominance of the machine over the crowd is more pronounced for highly risky listings. We then use the machine to make investment decisions, and find that the machine benefits not only the lenders but also the borrowers. When machine prediction is used to select loans, it leads to a higher rate of return for investors and more funding opportunities for borrowers with few alternative funding options. We also find suggestive evidence that the machine is biased in gender and race even when it does not use gender and race information as input. We propose a general and effective "debasing" method that can be applied to any prediction focused ML applications, and demonstrate its use in our context. We show that the debiased ML algorithm, which suffers from lower prediction accuracy, still leads to better investment decisions compared with the crowd. These results indicate that ML can help crowd lending platforms better fulfill the promise of providing access to financial resources to otherwise underserved individuals and ensure fairness in the allocation of these resources.

Problem definition : This paper studies the role of seekers’ problem specification in crowdsourcing contests for design problems. Academic/practical relevance : Platforms hosting design contests offer detailed guidance for seekers to specify their problems when launching a contest. Yet problem specification in such crowdsourcing contests is something the theoretical and empirical literature has largely overlooked. We aim to fill this gap by offering an empirically validated model to generate insights for the provision of information at contest launch. Methodology : We develop a game-theoretic model featuring different types of information (categorized as “conceptual objectives” or “execution guidelines”) in problem specifications and assess their impact on design processes and submission qualities. Real-world data are used to empirically test hypotheses and policy recommendations generated from the model, and a quasi-natural experiment provides further empirical validation. Results : We show theoretically and verify empirically that with more conceptual objectives disclosed in the problem specification, the number of participants in a contest eventually decreases; with more execution guidelines in the problem specification, the trial effort provision by each participant increases; and the best solution quality always increases with more execution guidelines but eventually decreases with more conceptual objectives. Managerial implications : To maximize the best solution quality in crowdsourced design problems, seekers should always provide more execution guidelines and only a moderate number of conceptual objectives.

Data protection and privacy rights afforded to consumers by General Data Protection Regulation (GDPR) have presented a challenge to the firms that depend on private consumer data for targeted pricing, product recommendations and other purposes. Specifically, GDPR affords consumers the right to not share any data or get past data erased and get a fresh start with the firm. Using a simple analytical model that captures the benefits of data sharing and the privacy concerns for the consumers, we show that GDPR can actually benefit firms. In particular, there are conditions under which both firm and consumer surplus as well as the amount of data collection increase under GDPR. The key insight is that GDPR serves to differentiate consumers based on their dis-utility for sharing data (privacy sensitivity). The optimal equilibrium strategy for the platform is to increase the data collection on those that have low privacy sensitivity, give them better recommendations but charge higher prices. At the same time, collect no or little data for highly price-sensitive consumers who exercise the GDPR right. It is optimal for the firm to provide poor product recommendations to users who exercise the GDPR rights to discourage them from exercising these rights. Overall, the results show that there are beneficial effects of GDPR to firms particularly in markets where consumers care about privacy a lot.

In this paper, we propose a two-step data-analytic approach to the promotion planning for mobile applications (apps). In the first step, we use historical sales data to estimate the app demand model and quantify the effect of price promotions on download volume. The estimation results reveal two interesting characteristics of the relationship between price promotion and download volume of mobile apps: (1) the magnitude of the direct immediate promotion effect is declining within a multiday promotion; and (2) due to the visibility effect (i.e., apps ranked high on the download chart are more visible to consumers), a price promotion also has an indirect effect on download volume by affecting app rank, and this effect can persist after the promotion ends. Based on the empirically estimated demand model, we propose a moving planning window heuristic to construct a promotion policy. Our heuristic promotion policy consists of shorter and more frequent promotions. We show that the proposed policy can increase the app lifetime revenue by around 10%.

Artificial intelligence (AI) and machine learning (ML) algorithms are widely used throughout our economy in making decisions that have far-reaching impacts on employment, education, access to credit, and other areas. Initially considered neutral and fair, ML algorithms have recently been found increasingly biased, creating and perpetuating structural inequalities in society. With the rising concerns about algorithmic bias, a growing body of literature attempts to understand and resolve the issue of algorithmic bias. In this tutorial, we discuss five important aspects of algorithmic bias. We start with its definition and the notions of fairness policy makers, practitioners, and academic researchers have used and proposed. Next, we note the challenges in identifying and detecting algorithmic bias given the observed decision outcome, and we describe methods for bias detection. We then explain the potential sources of algorithmic bias and review several bias-correction methods. Finally, we discuss how agents’ strategic behavior may lead to biased societal outcomes, even when the algorithm itself is unbiased. We conclude by discussing open questions and future research directions.