本文精选了信息系统领域国际顶刊《Information Systems Research》近期发表的论文，提供信息系统研究领域最新的学术动态。

Identification of Causal Mechanisms from Randomized Experiments: A Framework for Endogenous Mediation Analysis

原刊和作者：

Information Systems Research Volume34, Issue1

Jing Peng （University of Connecticut）

Abstract

Experimental research in the business disciplines often focuses on the overall treatment effect and the heterogeneity therein. Whereas this type of research allows us to understand the strength and direction of the treatment effect under different conditions, it does not directly speak to the generative mechanisms, namely, why and how the effect arises. A standard procedure to identify the mechanisms underlying a treatment effect is mediation analysis, but extant mediation analysis frameworks either have no causal interpretation or require the mediators to be unconfounded. Because mediators are posttreatment variables that typically cannot be preassigned beforehand, the endogeneity of mediators remains a serious concern even in randomized experiments. In response to this issue, we present a flexible endogenous mediation analysis framework that still has causal interpretation when the mediator is endogenous. We then discuss the identification conditions for different types of endogenous mediators, including unobserved or partially observed ones, under this framework. We show that endogenous mediation models can be parametrically identified without an instrumental variable when the generating process of the mediator is nonlinear. We further examine how the identification strengths of these models vary with a series of factors, including the level of endogeneity, the goodness of fit of the mediator model, the percentage of observed mediator values, and the misspecification of the error terms. Finally, we provide guidelines on when and how to use endogenous mediation analysis and discuss implications for experimental design and empirical research. We offer an R package that implements the proposed endogenous mediation models.

Link: https://doi.org/10.1287/isre.2022.1113

The Path to Hedonic Information System Use Addiction: A Process Model in the Context of Social Networking Sites

原刊和作者：

Information Systems Research Volume34, Issue1

lsaac Vaghefia （City University of New York）

Bogdan Negoita （HEC Montreal）

Liette Lapointe （McGill University）

Abstract

This study answers the call for a longitudinal view of addiction to hedonic information systems (IS) use by proposing a process model of its development, in the context of social networking site use. Through inductive and iterative analyses of primary data collected via interviews and surveys, and secondary data in the form of narrative accounts, we explain the process of addiction development via three phases associated with nominal, compulsive, and addicted use. In each phase, combinations of salient individual needs, affordances, technology features, IS use behaviors, and control mechanism outcomes (successful or unsuccessful) influence an individual’s trajectory toward hedonic IS use addiction. Drawing on cybernetic theory, we explain the role of users’ control mechanisms. We show how deficiencies related to the sensing, comparing, or regulating act, in conjunction with salient affordances, influence the development of addiction. The findings extend variance-based research on IS use addiction. They carry implications for research, users, technology providers, and policy makers in relation to hedonic IS use addiction.

Link: https://doi.org/10.1287/isre.2022.1109

Are You What You Tweet? The Impact of Sentiment on Digital News Consumption and Social Media Sharing

原刊和作者：

Information Systems Research Volume34, Issue1

Hyelim Oh （Sogang University）

Khim-Yong Goh （National University of Singapore）

Tuan Q. Phan （University of Hong Kong）

Abstract

Although social media has helped online newspapers by allowing users to organically share articles, some have argued that it has cannibalized and hurt newspapers through reduced readership and diminished agenda-setting power. Motivated by these two opposing effects, it is critical to understand what affects the duality between sharing news articles on social media and reading the articles on news websites. Using rich clickstream data on online news readership obtained from an English-language newspaper in an Asian country and social media data collected from Twitter, we focus on article sentiment as a key news content attribute and find a differential effect of sentiment on readership and sharing behaviors across the news site and third-party social media platform. Our results show that people are likely to read news articles with negative sentiment on the news site, but they tend to share articles with positive sentiment on Twitter. Specifically, a one-unit increase in content sentiment is associated with a 10.86% or 273-unit decrease in news site page views but a 17.0% or 2.10-unit increase in Twitter sharing volume. Upon decomposition of news article sentiment, we also find a contrasting positive author sentiment effect and a negative news topic valence effect on news readership. To uncover the underlying mechanism of the findings, we test the key intuitions from prior self-presentation literature. We find that an increase in a Twitter user’s followers (i.e., audience size) leads to an increase in the Twitter user’s propensity to share positive-sentiment news articles. Our findings on the role of sentiment on content consumption and sharing affirm the coopetitive but complementary relationship between news websites and social media platforms. Our results also guide publishers to better craft their news content and manage social media presence to improve audience engagement and readership outcomes while preserving the agenda-setting ability of news media.

Link: https://doi.org/10.1287/isre.2022.1112

sDTM: A Supervised Bayesian Deep Topic Model for Text Analytics

原刊和作者：

Information Systems Research Volume34, Issue1

Yi Yang （Hong Kong University of Science and Technology）

Kunpeng Zhang （University of Maryland）

Yangyang Fan （Hong Kong Polytechnic University）

Abstract

Topic modeling methods such as latent Dirichlet allocation (LDA) are powerful tools for analyzing massive amounts of textual data. They have been used extensively in information systems (IS) and business discipline research to identify latent topics for data exploration and as a feature engineering mechanism to derive new variables for analyses. However, existing topic modeling approaches are mostly unsupervised and only leverage textual data, while ignoring additional useful metadata often associated with text, such as star ratings in customer reviews or categories of posts in online forums. As a result, the identified topics and variables derived based on the learned topic model may not be accurate, which could lead to incorrect estimations that affect subsequent empirical analysis and to inferior performance on predictive tasks. In this study, we propose a novel supervised deep topic modeling approach called sDTM, which combines a neural variational autoencoder model and a recurrent neural network. sDTM leverages the auxiliary data associated with text to enhance the topic modeling capability. We conduct empirical case studies and predictive analytics on an online consumer review data set and an online knowledge community data set. Experimental results show that in comparison with benchmark methods, sDTM can enhance both the empirical estimation and predictive performance. sDTM makes methodological contributions to the IS literature and has direct relevance for research using text analytics.

Link: https://doi.org/10.1287/isre.2022.1124

Augmenting Password Strength Meter Design Using the Elaboration Likelihood Model: Evidence from Randomized Experiments

Information Systems Research Volume34, Issue1

Warut Khern-am-nuai （McGill University）

Matthew J. Hashim （University of Arizona）

Alain Pinsonneault （McGill University）

Weining Yang （ByteDance Inc）

Ninghui Li （Purdue University）

Abstract

Password-based authentication is the most commonly used method for gaining access to secured systems. Unfortunately, empirical evidence highlights the fact that most passwords are significantly weak, and encouraging users to create stronger passwords is a significant challenge. In this research, we propose a theoretically augmented password strength meter design that is guided by the elaboration likelihood model of persuasion (ELM). We evaluate our design by leveraging three independent and complementary methods: a survey-based experiment using students to evaluate the saliency of our conceptual design (proof of concept), a controlled laboratory experiment conducted on Amazon Mechanical Turk to test the effectiveness of the proposed design (proof of value), and a randomized field experiment conducted in collaboration with an online forum in Asia to establish proof of use. In each study, we observe the changes in users’ behavior in response to our proposed password strength meter. We find that the ELM-augmented password strength meter is significantly effective at addressing the challenges of password-based authentication. Users exposed to this strength meter are more likely to change their passwords, leading to a new password that is significantly stronger. Our findings suggest that the proposed design of augmented password strength meters is an effective method for promoting secure password behavior among end users.

Link: https://doi.org/10.1287/isre.2022.1125