Technology Management and Information Systems Seminars


Date Speaker Affiliation Presentation Room
06.12.22 Yaniv Dover HUJI
The Role of Emotion Shocks in Online Community Dynamics
Online communities generate marketing value for both firms and consumers, and their sustainability relies on continuous user engagement. In an environment in which competition for engagement is fierce, digital platforms seemingly favour the use of negative emotion to engage users. Yet, the impact of emotion shocks on the fabric of social interactions in online communities remains unclear. In this study, we assess the impact of negative emotional shocks on user engagement in online communities, leveraging a natural experiment on Particularly, we measure the effects of emotion shocks associated with outcomes of basketball games. We track the volume and network structure of user engagement in basketball communities around game days, and estimate causal effects of emotion shocks using difference-in-difference models. We find that negative emotional shocks decrease user engagement, retention, and the activation of new users.
Furthermore, the structure of discussion networks changes, and user-generated content expresses more social disconnection. These effects are strongest for unexpected negative events, suggesting that expectations management can mitigate emotional disruptions. The effects also vary depending on the relative position of users in their social networks. The results provide insight into the complex social consequences of exposing users to emotions on community dynamics.
27.12.22 Ayelet Israeli Harvard

BEAT Unintended Bias in Personalized Policies 
An inherent risk of algorithmic personalization is disproportionate targeting of individuals from certain groups (or demographic characteristics such as gender or race), even when the decision maker does not intend to discriminate based on those “protected” attributes. This unintended discrimination is often caused by underlying correlations in the data between protected attributes and other observed characteristics used by the algorithm (or machine learning (ML) tool) to create predictions and target individuals optimally. Because these correlations are hidden in high dimensional data, removing protected attributes from the database does not solve the discrimination problem; instead, removing those attributes often exacerbates the problem by making it undetectable and, in some cases, even increases the bias generated by the algorithm. We propose BEAT (Bias-Eliminating Adapted Trees) to address these issues. This approach allows decision makers to target individuals based on differences in their predicted behavior—hence capturing value from personalization—while ensuring a balanced allocation of resources across individuals, guaranteeing both group and individual fairness. Essentially, the method only extracts heterogeneity in the data that is unrelated to protected attributes. To do so, we build on the General Random Forest (GRF) framework (Athey et al., (2019)) and develop a targeting allocation that is “balanced” with respect to protected attributes. We validate BEAT using simulations and an online experiment with N=3,146 participants. This approach can be applied to any type of allocation decision that is based on prediction algorithms, such as medical treatments, hiring decisions, product recommendations, or dynamic pricing

03.01.22 Neta Livneh TAU    
17.01.23 Roee levy TAU

Demand for Online News, Inertia, and Misperceptions
There has been growing concern that individuals consume low-quality, extreme, news on social media. We test whether such news choices may stem from misperceptions and whether simple interventions can decrease the slant and increase the quality of people’s news diets on social media. In a survey experiment, we first elicit perceptions regarding the slant and quality of news and then randomize participants into three groups: a group provided with information on media slant, a group provided with information on media quality, and a group not provided with any information. After the intervention, we offer all participants an option to readjust their news portfolio by following or unfollowing news outlets on Facebook. We find that individuals suffer from misperceptions, especially regarding the quality of outlets. Correcting these misperceptions increases the quality of the outlets individuals choose to follow. However, providing information on slant does not have a significant effect on news choices. Still, we find that simply providing individuals an option to readjust their news choices results in a more moderate news portfolio.


19.01.23 Raveesh Mayya NYU Delaying Informed Consent: An Empirical Investigation of Mobile Apps' Upgrade Decisions
We study apps' decisions to upgrade to Android 6.0, which restricts their ability to seek blanket permissions to sensitive user information at install-time, instead requiring them to request à la carte permissions at runtime. Such a shift in Android’s permission-seeking policy comes in the wake of apps’ overreaching and users’ proactive measures to protect their sensitive information. Mobile apps on Android had a choice of upgrading to Android 6.0 anytime over a three-year window instead of being forced to upgrade immediately. Given the choice of upgrading to version 6.0, that provides mobile apps with the latest platform features or staying with an earlier version that provides them with better access to user information, our study seeks to examine the upgrade decisions of apps and the outcomes of such decisions. We assemble and analyze a unique panel dataset comprising 13,691 of the most popular apps for 24 months and find that apps that traditionally seek more runtime permissions than those required for the app’s functionality, delay upgrading. Specifically, we find that such upgrade delays are more likely by overreaching apps that seek to serve targeted advertisements in-app. Even those apps that have an iOS equivalent (iOS had à la carte permission regime since 2012) demonstrate similar delaying behavior, indicating that the reasons for delay are strategic and not operational. More importantly, we find that such upgrade delays come at a cost to apps in terms of marketplace outcomes such as rating and popularity. We discuss the implications of our findings for app-providers and platform-operators.


25.01.23 Alexander Rossmann Reutlingen University Caught in Disruption: Status quo of digital transformation and current research topics for global OEMs in the automotive industry
The business model of global OEMs in the automotive industry is increasingly being challenged by digital technologies. Mobility-as-a-service, car sharing, self-driving cars and connected vehicles are attacking the business model of incumbant OEMs. However, such innovations are also leading to new business opportunities. Common to these business potentials is the focus on software and data. Therefore, value creation will be based extensively on core competencies in the area of software and data. The question is whether disruptive solutions in this area will be driven by established OEMs or technology giants and startups from outside the industry. Disruptions for OEMs like Mercedes, Volkswagen and Porsche can be observed in various functional areas and along the entire value chain. For instance, innovations in the area of Industry 4.0 point to how vehicles will be produced in smart factories and how the data produced in production might be used for customer experience management. The vehicles themselves will be sold more through online marketing and eCommerce and the whole customer experience journey is getting more and more digital. In the case of autonomous vehicles, completely new business models will emerge at the so-called third place (What will drivers do when they are not driving?). Mobility-as-a-service platforms might be seen as the final stage in the digital transformation process. The speech provides insights into the research agenda of global OEMs in the automotive industry in various focus areas across product development, production, marketing and sales. In addition, the speech will provide a detailed insight into some selected research projects, publications and planned initiatives.
7.2.23 Galit Shmueli National Tsing Hua University How to "Improve" Prediction Using Behavior Modification
Many internet platforms that collect behavioral big data use it to predict user behavior for internal purposes and for their business customers (e.g., advertisers, insurers, security forces, governments, political consulting firms) who utilize the predictions for personalization, targeting, and other decision-making. Improving predictive accuracy is therefore extremely valuable. Data science researchers design algorithms, models, and approaches to improve prediction. Prediction is also improved with larger and richer data. Beyond improving algorithms and data, platforms can stealthily achieve better prediction accuracy by "pushing" users' behaviors towards their predicted values, using behavior modification techniques, thereby demonstrating more certain predictions. Such apparent "improved" prediction can unintentionally result from employing reinforcement learning algorithms that combine prediction and behavior modification. This strategy is absent from the machine learning and statistics literature. Investigating its properties requires integrating causal with predictive notation. To this end, we incorporate Pearl's causal do(.) operator into the predictive vocabulary. We then decompose the expected prediction error given behavior modification, and identify the components impacting predictive power. Our derivation elucidates implications of such behavior modification to data scientists, platforms, their customers, and the humans whose behavior is manipulated. Behavior modification can make users' behavior more predictable and even more homogeneous; yet this apparent predictability might not generalize when customers use predictions in practice. Outcomes pushed towards their predictions can be at odds with customers' intentions, and harmful to manipulated users.
28.2.23 Shir Etgar Columbia and TAU Does the Type of Device Influence Cognitive Processes?
 Given the prevalence of smartphones and computers in use today, the question arises whether these devices have a similar or different influence on our cognitive processes (e.g., reasoning, decision-making, learning, etc.). Previous studies found specific behavioral differences between smartphone use and computer use. However, those findings have always been examined and explained independently from one another. The current research aims to explore whether there is a general mechanism underlying all of these different findings. Relying on the dual-process theory, I hypothesize that smartphone use is associated with lower levels of deliberative cognitive processes than computer use. This hypothesis was examined by exploring the effect of the type of device on different features deriving from the dual-process theory, such as cognitive functioning, processing speed, and biased thinking. Various methods were applied to demonstrate the effect, including Bayesian multilevel meta-analysis, correlational study, and an experimental approach. The talk will review their findings and suggest an underlying mechanism for this effect.
Recanati 430.
14.3.23 Danit Ein-Gar TAU

The Donor's Choice Dilemma 

We hold the assumption that "any help is better than no help", but do people follow this assumption when they need to choose between two needy recipients? 

In ongoing research, we first find that prospective donors say they want choice. When given a choice between two similarly deserving recipients, they experience a moral conflict. Choosing to help one recipient seems unfair toward the unchosen recipient. Due to this moral conflict between the wish to help and the wish to do so in a fair manner many prospective donors prefer not to donate to any recipient (“opt-out”) acting against the intuition that "any help is better than no help".  We identify the types of choice-sets that yield the highest opt-out rates. We then focus on one particularly interesting case which attenuates the opt-out rates – a case where donors need to choose between helping a boy or helping a girl. Overall, opt-out rates decline suggesting that gender can help mitigate the moral conflict of choosing one over the other. Interestingly, we also find that in a western culture (US), participants chose to help a girl over a boy, in support of the western stereotype that women are needier than men. However, in an eastern culture (Chinese), the effect is reversed, and donors prefer to help the boy over the girl in support of male favoritism in eastern societies. 

Recanati 430.
18.4.23 Chen Elyashar TAU

Should a Chatbot Show it Cares? Toward Optimal Chatbot Design via Emotion Recognition and Sentiment Analysis 

Recent advances in natural language processing (NLP) have led to substantial improvements
in chatbots, autonomous interactive software agents that are based on NLP capabilities. This
has led consumer-facing firms to increasingly employ chatbots in customer service
interactions. We seek to understand whether and how chatbots of e-commerce websites should
incorporate sentiment into customer-service interactions. We first analyze customer service
interactions on Twitter between customers and human representatives to identify different
types of responses, their sentiment, and resulting satisfaction levels. We then proceed to run an
online experiment in which participants interact with a chatbot that uses either caring or neutral
language. We find that, while, in most cases an empathic chatbot response increases customer
satisfaction (at least weakly), there exist other cases, where an empathic response decreases
user satisfaction. Thus, a caring chatbot is not always optimal. This work will inform chatbot
design in customer service settings, improving customer satisfaction with chatbot interactions
without relying on any personal or historical user data.

Recanati 105
9.5.23 Itay P. Fainmesser, Johns Hopkins University Digital Privacy

We study the incentives of a digital business to collect and protect users’ data. The users’ data the business collects improve the service it provides to consumers, but they may also be accessed, at a cost, by strategic third parties in a way that harms users, imposing endogenous users’ privacy costs. We characterize how the revenue model of the business shapes its optimal data strategy: collection and protection of users’ data. A business with a more data-driven revenue model will collect more users’ data and provide more data protection than a similar business that is more usage-driven. Consequently, if users have small direct benefit from data collection, then more usage- riven businesses generate larger consumer surplus than their more data-driven counterparts (the reverse holds if users have large direct benefit from data collection). Relative to the socially desired data strategy, the business may over- or under-collect users’ data and may over- or under-protect it. Restoring efficiency requires a two-pronged regulatory policy, covering both data collection and data protection; one such policy combines a minimal data protection requirement with a tax proportional to the amount of collected data. We finally show that existing regulation in the US, which focuses only on data protection, may even harm consumer surplus and overall welfare.

Recanati 105
Dokyun "DK" Lee Questrom School of Business, Boston University

 InnoVAE: Generative AI for Understanding Patents and Innovation.


A lack of interpretability limits the use of common unsupervised learning techniques (e.g., PCA, t-SNE) in contexts where they are meant to augment managerial decision-making. We develop a generative deep learning model based on a Variational AutoEncoder (“InnoVAE”) that converts unstructured patent text into an interpretable, spatial representation of innovation (“Innovation Space”). After validating the internal consistency of the model, we apply it to three decades of computing system patents to show that our approach can be used to construct economically interpretable measures—at scale—that characterize a firm’s IP portfolio from the text of its patents, such as whether a patent is a breakthrough innovation, the volume of intellectual property enclosed by a portfolio of patents, or the density of patents at a point in Innovation Space. We show that for explaining innovation outcomes, these interpretable, engineered features have explanatory power that augments and often surpasses the structured patent variables that have informed the very large and influential literature on patents and innovation. Our findings illustrate the potential of using generative methods on unstructured data to guide managerial decision-making.

23.5.23 Jan Recker  University of Hamburg We study how control has been enacted on the repair aftermarket of the Apple iPhone, the largest aftermarket for smartphones. We analyze how over a period of thirteen years Apple and independent repair service providers used different ways (through physical, regulatory, or algorithmic instruments) to influence each other’s possibilities for controlling inputs, processes, and outputs of smartphone repair. We show how the increasing use of algorithmic instruments for enacting control, made possible through emerging functionality for tethering, encryption, and temporary binding implemented in the iPhone itself, were shaped by, and shaped, the actions of Apple and the independent repair service providers. Our research provides several contributions to our understanding of control, the dialectics in enacting control, and the increasing use of algorithmic instruments for enacting control on product aftermarkets. Our study also provides practical implications regarding the regulation of smartphone aftermarkets. Recanati 105
4.6.23 Alex Tuzhilin Stern School of Business, NYU

Consumer Preference Exploration with Unexpected Recommender Systems

One of the key issues with recommender systems constitutes the filter bubble phenomenon when consumers are presented mostly with familiar and repeated types of recommendations which isolates them from the less familiar world of broader choices and options. To address this problem, this talk presents a novel approach to providing unexpected recommendations that surprise consumers by significantly deviating from their typical expectations. In particular, the unexpectedness objective is introduced in this talk using certain deep learning methods and then is subsequently incorporated into the utility function in a personalized manner that captures heterogeneous consumer propensity to seek product variety. It will also be shown that it is desirable to provide more unexpected recommendations to variety-seekers, and vice versa. It will be shown that the proposed model significantly increases various business performance metrics vis-à-vis the currently used methods.

Recanati 405


Past Seminars:


 |  2022  | 


Register here
Tel Aviv University makes every effort to respect copyright. If you own copyright to the content contained
here and / or the use of such content is in your opinion infringing, Contact us as soon as possible >>