Essay on Digital Platforms and AI

Date of Award

Fall 10-1-2021

Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Shin, Jiwoong


The first chapter is a joint work with Jiwoong Shin, Soheil Ghili, and Jaehwan Kim. In this chapter, we demonstrate the impact of the gig economy on product quality in seemingly unrelated local industries through the labor market. Our empirical context is the quality of service for restaurants in the city of Austin and we examine how they were impacted by the {exogenous exit and re-entry of rideshare platforms, Uber and Lyft into the city due to regulatory changes. We leverage these exogenous shocks and combine them with sentiment-analyzed data from Yelp reviews that capture how customers assess the quality of service at each restaurant. We show that, compared to control cities, customers in Austin become more negative about service quality when Uber and Lyft are present in the city. Additionally, we use rich data on employee turnover and wages to demonstrate that, compared to the control cities, service staff turnover indeed increases in Austin when Uber and Lyft are present. We also conduct several additional studies and robustness checks that are all congruent with our hypothesis that Uber and Lyft lower the quality of service in Austin restaurants by raising the turnover of their staff. Together, these results suggest significant ramifications of the gig economy on the broader industries through the labor market. The second essay, a joint work with K.Sudhir and Kosuke Uetake, studies whether and how a private communication channel can affect decision-making of players in an online freelancing platform. Online platforms that facilitate exchanges and trade through matching have proliferated in recent years. Though these platforms allow for information provision by participants, information asymmetry remains a significant impediment in facilitating matches between participants. This paper investigates whether the platform can enhance efficiency by providing private communication channels to exchange tailored information between participants to reduce information asymmetry. As such communications tend to be ``cheap talk'' in that messages are neither verifiable nor differentially costly. Hence, whether such a communication channel can improve efficiency is an empirical question. We take advantage of a natural experiment on an online labor market platform, which introduced a new channel of communication between service buyers and freelancers to answer the question. We find that the communication channel does improve efficiency in long-term projects. Specifically, both short-term and long-term projects have higher matching probabilities, but the benefits of the communication channel seem to be greater for long-term projects in terms of more communications, higher contract probabilities. Interestingly, the ``cheap talk'' channel also reduces the need for ``costly signaling'' by service buyers to post a higher project price. Because most online labor platforms charge percentage fees from a contract price, our finding gives additional managerial insight that the platform should contemplate the trade-off between increased quantity and reduced price. In the third chapter, a joint work with Jin Kim and Minkyung Kim, we study human learning from Artificial Intelligence. Although Artificial Intelligence (AI) is expected to outperform humans in many domains of decision-making, the process by which AI arrives at its superior decisions is often hidden and too complex for humans to fully grasp. As a result, humans may find it difficult to learn from AI, and accordingly, our knowledge about whether and how humans learn from AI is also limited. In this paper, we aim to expand our understanding by examining human decision-making in the board game Go. Our analysis of 1.3 million move decisions made by professional Go players suggests that people learned to make decisions like AI after they observe {reasoning processes of AI, rather than mere {actions of AI. Follow-up analyses compared the decision quality of two groups of players: those who had access to AI programs and those who did not. In line with the initial results, decision quality significantly improved for the players with AI access after they gained access to {reasoning processes of AI, but not for the players without AI access. Our results demonstrate that humans can learn from AI even in a complex domain where the computation process of AI is also complicated.

This document is currently not available here.