Category Archives: Working Papers

From Enthusiasm to Reality: Evaluating Generative AI’s Role in Modern Journalism

Zhang, Xiaoke, Myunghwan Lee, Mi Zhou, Gene Moo Lee “From Enthusiasm to Reality: Evaluating Generative AI’s Role in Modern Journalism”, Work-in-progress.

  • Presentations: UBC (2024)

Generative AI (GenAI), initially greeted with enthusiasm for its potential for content creation, encounters challenges when applied in professional settings such as journalism. These challenges, including the generation of inaccurate outputs, inconsistencies, and a reduction in human accountability, may pose conflicts with the core journalistic values of accuracy, transparency, and credibility. Our research investigates the impact of GenAI in news media leveraging a unique empirical setting when a major news outlet in South Korea launched a GenAI-powered news editor to assist its journalists in news production in December 2023. Our preliminary analysis of 196,288 news articles published between June 2023 and April 2024 suggests that GenAI adoption has not led to a significant increase in productivity, indicating persistent challenges in effectively integrating GenAI into journalistic workflows. Our study seeks to further explore this phenomenon by addressing two primary questions. First, we will conduct a survival analysis to identify effective GenAI strategies that lead to consistent GenAI use and positive outcomes in news production. Second, we will examine the impact of GenAI on the overall media news output (e.g., local vs. global; factual vs. opinion news) and discuss its broader implications in ideology formation (e.g., polarization). This research will contribute to the nascent literature on GenAI’s impact on digital platforms by providing a nuanced understanding of the phenomenon.

Unlocking the Impact of Machine Learning on Organizational Learning: Evidence from US Public Firms

Lee, Myunghwan, Timo Sturm, Gene Moo Lee “Unlocking the Impact of Machine Learning on Organizational Learning: Evidence from US Public Firms”, Work-in-Progress.

  • Presentations: KrAIS Summer 2024

Organizational learning is a critical core process that fundamentally controls organizations’ innovation and thus affects organizational performance and long-term survival. Due to their ability to learn, recent research has recognized the far-reaching influence of machine learning (ML) systems’ contributions to organizational learning. So far, however, the emerging discourse on the role of ML in organizational learning has remained largely theoretical, offering helpful initial insights but inconclusive predictions about ML’s impact. To resolve existing conflicts by adding empirical evidence, we explore the real-world innovations of 265 ML and 700 non-ML organizations from 2006 to 2019. Based on a novel ML measure constructed from ML patents, publications, and transaction datasets, we find that ML primarily contributes to shifting organizational learning toward exploration. Our results further show that ML’s influence depends on external environmental factors: ML’s effect increases with higher levels of market concentration and competitors’ strategic orientation towards ML. Lastly, we find that organizations using ML tend to survive longer due to increased performance and more balanced innovation. To the best of our knowledge, this is the first large-scale empirical study of the impact of ML on organizational learning outcomes, contributing to rethinking organizational learning in the era of ML.

Unpacking the AI Blackbox:  The Impact of AI Strategies on Firm Performance with a Dual Lens on Product and Process Orientation

Park, Jaecheol, Myunghwan Lee, J. Frank Li, Gene Moo Lee “Unpacking the AI Blackbox:  The Impact of AI Strategies on Firm Performance with a Dual Lens on Product and Process Orientation”, Work-in-Progress.

  • Presentations: UBC (2024), INFORMS (2024)

Artificial intelligence (AI) technologies have become increasingly pervasive and hold great potential for large-scale economic impact. Aligned with this trend, numerous studies explore the adoption and use of AI technologies on firm performance. However, they predominantly focus on AI as an input (e.g., labor/job posting), neglecting to consider the strategic deployment of AI in business operations. Thus, it is crucial to understand how” and “where” to use AI to achieve business value. In this paper, we examine how firms’ strategic AI orientation affects firm performance with a dual-lens on product and process orientation. We measure AI orientation by employing LLM to assess it from strategy descriptions in Form 10-K filings between 2015 and 2022. Our findings show that 13% of firms have an AI orientation, 7% have an AI product orientation, and 3% have an AI process orientation. Additionally, we will provide some preliminary results on the impact of AI orientation on firm performance. 

Anatomy of Phishing Tactics and Susceptibility

Bera, Debalina, Gene Moo Lee, Dan J. Kim “Anatomy of Phishing Tactics and Susceptibility: An Investigation of the Dynamics of Phishing Tactics and Contextual Traits in Susceptibility,” Working Paper.

Phishing is a deceptive tactic to create a front of apparent credibility to fraudulently acquire sensitive personal or financial information from an unsuspecting user or espionage system by infiltrating malware or crimeware. Despite automated technological solutions and training interventions, recent phishing statistics show that specifically few phishing tactics are increasing users’ phishing susceptibility (PS). Further, assessing the moderating role of phishing contextual traits in the relationship between phishing tactics and PS indicates the importance of their trait differences. Based on theoretical postulation, employing a sequential mixed method design, and using two sets of data (simulated phishing penetration testing results and scenario-based experiments), we examine the effect of phishing tactics along with the moderating role of individual phishing contextual traits on PS. This study extends the theoretical boundary relevant to phishing tactics and provides practical guidance to identify the most dangerous phishing tactics that increase PS and phishing contextual traits that help to combat phishing attacks.

The Effect of Mobile Device Management on Work-from-home Productivity: Insights from U.S. Public Firms

Park, Jaecheol, Myunghwan Lee, Gene Moo Lee “The Effect of Mobile Device Management on Work-from-home Productivity: Insights from U.S. Public Firms”, Work-in-Progress.

  • Presentations: UBC 2023, MSISR 2023, KrAIS 2023, WeB 2023, AOM 2024
  • Best Paper Nomination at WeB 2023

The use of mobile IT, providing employees with accessibility, flexibility, and connectivity, has become increasingly vital for businesses, especially for work-from-home during the COVID-19 pandemic. However, despite its prevalence and importance in the industry, the business value of mobile device management (MDM) and its role in establishing digital resilience remain underexplored in the literature. To address this research gap, our study examines the effect of MDM on a firm’s resilience to the pandemic. Drawing on the resource-based view (RBV), we find that firms with MDM have better financial performance during the pandemic, demonstrating greater resilience to the shock. Additionally, we explore the moderating role of external and internal factors, revealing that firms with high environmental munificence or those with low IT capabilities experience greater resilience effects from MDM. Furthermore, we observe heterogeneous effects across industries that firms in industry sectors demanding greater mobility have a greater resilience effect from MDM. This study contributes to the information systems literature by emphasizing the business value of MDM and its crucial role in building digital resilience.

Disrupt with AI: The Impact of Deep Learning Capabilities on Exploratory Innovation

Lee, Myunghwan, Victor Cui, Gene Moo Lee. “Disrupt with AI: The Impact of Deep Learning Capabilities on Exploratory Innovation”, AOM 2023

Given the importance of exploratory innovation in fostering firms’ sustainable competitive advantages, firms often depend on technological assets or inter-firm relationships to pursue exploration. Regarded as a general-purpose technology, deep learning (DL)-based artificial intelligence (AI) can be an exploratory innovation-seeking instrument for firms in searching unexplored resources and thereby broadening their boundary. Drawing on the theories of organizational learning and path dependence, we hypothesize the impact of a firm’s DL capabilities on exploratory innovation and how DL capabilities interact with conventional pathbreaking activities such as technical assets and inter-firm relationships. Our empirical investigations, based on a novel DL capabilities measure constructed from comprehensive datasets on AI conferences and patents, show that DL capabilities have positive impacts on exploratory innovation. The results also show that extant technological assets (i.e., structured data management capabilities) and inter-firm relationships remedy the constraints on a firm’s innovation-seeking behaviors and that these path-breaking activities negatively moderate the positive impact of DL capabilities on exploratory innovation. To our knowledge, this is the first large-scale empirical study to investigate how DL affects exploratory innovation, contributing to the emerging literature on AI and innovation.

How Does AI-Generated Voice Affect Online Video Creation? Evidence from TikTok

Zhang, Xiaoke, Mi Zhou, Gene Moo Lee How Does AI-Generated Voice Affect Online Video Creation? Evidence from TikTok”, Working Paper.

  • Presentations: INFORMS DS (2022), UBC (2022), WITS (2022), Yonsei (2023), POSTECH (2023), ISMS MKSC (2023), CSWIM (2023), KrAIS Summer (2023), Dalhousie (2023), CIST (2023), Temple (2024), Santa Clara U (2024), Wisconsin Milwaukee (2024)
  • Best Student Paper Nomination at CIST 2023
  • API sponsored by Ensemble Data
  • SSRN version: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4676705

The rising demand for online video content has fostered one of the fastest-growing markets as evidenced by the growing popularity of platforms like TikTok. In response to the challenges of video creation, these platforms are increasingly incorporating artificial intelligence (AI) to support creators in their video creation process. However, little is known about how AI integration influences online content creation. Our paper aims to address this gap by investigating the impact of AI-generated voice on video creators’ productivity and creative patterns. Using a comprehensive dataset of 554,252 videos from 4,691 TikTok creators, we conduct multimodal analyses of the video data to detect the adoption of AI voice and to quantify video characteristics. We then estimate the adoption effects using a stacked difference-in-differences model coupled with propensity score matching. Our results suggest that AI voice adoption significantly increases creator productivity. Moreover, we find that the use of AI voice enhances video novelty across image, audio, and text modalities, suggesting its role in reducing workload on routine tasks and fostering creative exploration. Lastly, our study also uncovers a disinhibition effect, where creators tend to conceal their identities with the AI voice and exert more negative sentiments because of diminished social image concerns. Our paper provides the first empirical evidence of how AI reshapes online video creation, providing important implications for creators, platforms, and policymakers in the creator economy.

Ideas are Easy but Execution is Everything: Measuring the Impact of Stated AI Strategies and Capability on Firm Innovation Performance

Lee, Myunghwan, Gene Moo Lee (2022) “Ideas are Easy but Execution is Everything: Measuring the Impact of Stated AI Strategies and Capability on Firm Innovation Performance”Work-in-Progress.

Contrary to the promise that AI will transform various industries, there are conflicting views on the impact of AI on firm performance. We argue that existing AI capability measures have two major limitations, limiting our understanding of the impact of AI in business. First, existing measures on AI capability do not distinguish between stated strategies and actual AI implementations. To distinguish stated AI strategy and actual AI capability, we collect various AI-related data sources, including AI conferences (e.g., NeurIPS, ICML, ICLR), patent filings (USPTO), inter-firm transactions related to AI adoption (FactSet), and AI strategies stated in 10-K annual reports. Second, while prior studies identified successful AI implementation factors (e.g., data integrity and intelligence augmentation) in a general context, little is known about the relationship between AI capabilities and in-depth innovation performance. We draw on the neo-institutional theory to articulate the firm-level AI strategies and construct a fine-grained AI capability measure that captures the unique characteristics of AI-strategy. Using our newly proposed AI capability measure and a novel dataset, we will study the impact of AI on firm innovation, contributing to the nascent literature on managing AI.

Do Incentivized Reviews Poison the Well? Evidence from a Natural Experiment at Amazon.com

Park, Jaecheol, Arslan Aziz, Gene Moo Lee. “Do Incentivized Reviews Poison the Well? Evidence from a Natural Experiment at Amazon.comWorking Paper.

  • Presentations: UBC (2021), KrAIS (2021), WISE (2021), PACIS (2022), SCECR (2022), BU Platform (2022), CIST (2022), BIGS (2022)
  • Preliminary version in PACIS 2022 Proceedings

The rapid growth in e-commerce has led to a concomitant increase in consumers’ reliance on digital word-of-mouth to inform their choices. As such, there is an increasing incentive for sellers to solicit reviews for their products. The literature has examined the direct and indirect effects of incentivized reviews on subsequent organic reviews within consumers who received incentives. However, since incentivized reviews and reviewers are often only a small proportion of a review platform (only 1.2% in our sample), it is important to understand whether their presence and absence on the platform affect the organic reviews from other reviewers who have not received incentives, which are often in the majority. We theorize two underlying effects that incentivized reviews can generate on other organic reviews: the herding effect from imitating incentivized reviews and the disclosure effect from the increased trust or skepticism by explicit incentive disclosure statements. Those two effects make organic reviews either follow or deviate from incentivized reviews. Using Bidirectional Encoder Representations from Transformers (BERT) to identify incentivized reviews and a natural experiment caused by a policy change on Amazon.com in October 2016, we conduct difference-in-differences with propensity score matching analyses to identify the effects of banning incentivized reviews on organic reviews. Our results suggest the disclosure effects are salient: banning incentivized reviews has positive effects on organic reviews in terms of frequency, sentiment, length, image, and helpfulness. Moreover, we find that the presence of incentivized reviews has poisoned the well for organic reviews regardless of the incentivized review ratio and that the effect is heterogeneous to product quality uncertainty. Our findings contribute to the literature on online review and platform design and provide insights to platform managers.

Learning Faces to Predict Matching Probability in an Online Dating Market (ICIS 2022)

Kwon, Soonjae, Gene Moo Lee, Dongwon Lee, Sung-Hyuk Park (2024) “Digital Cupid: Empowering Generative AI for Fair and Efficient Matchmaking,” Working Paper.

  • Previous title: Learning Faces to Predict Matching Probability in an Online Dating Market
  • Presentations: DS (2021), AIMLBA (2021), WITS (2021), ICIS (2022)
  • Preliminary version in ICIS 2022 Proceedings
  • Based on an industry collaboration

With the increasing prevalence of online transactions, enhancing matching efficiency has emerged as a critical objective for most matching platforms. However, these efforts often lead to decreased fairness, making it challenging to balance these two elements. This study presents a novel generative AI-based approach to increase the platform’s efficiency and fairness simultaneously in the context of online dating. By developing a model that utilizes users’ multimodal features to predict individual preferences, we assess the impact of various matching algorithms on platform efficiency and fairness. Extensive simulations show that our fairness-aware algorithm significantly enhances both metrics, addressing conventional methods’ severe efficiency-fairness tradeoff issue. We also introduce a novel generative AI-based personalization technique that modifies users’ profile images in different directions according to their counterparts, further boosting efficiency without sacrificing fairness. Our matching framework can be applied to platforms with various objectives, contributing to all stakeholders in digital platforms.