Technology Management and Information Systems Seminars

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. 

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.

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.

 InnoVAE: Generative AI for Understanding Patents and Innovation.

 Abstract:  

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.

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.