EKEllie J. Kyung

Ellie J. Kyung

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ellie J. Kyung has a total of 24 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Blinded versus Unblinded Review: A Field Study on the Equity of Peer-Review Processes

    by Timothy J. Pleskac, Ellie J. Kyung, Gretchen B. Chapman, Oleg Urminsky

    Blinded review is often proposed as a solution to inequities in merit-based evaluations. We tested this claim in a high-stakes field experiment comparing single-blind review (reviewer identities withheld) and double-blind review (reviewer and author identities withheld) for 530 academic conference submissions. On reliability, both processes showed moderate reliability, leading to only 40% overlap among top-rated submissions. This low overlap indicates that differences between review systems partly stem from noise in the review system. In terms of fairness, our analysis indicates that single-blind reviews favored more senior coauthors and first authors who were PhD students or research scientists while disfavoring Asian (versus White) first authors. Meanwhile, male first authors received slightly higher ratings in both conditions (especially under double-blind), but the greater the proportion of male coauthors, the lower the review ratings were under double-blind (versus single-blind). On validity, we find that for submissions accepted and presented as a talk, neither review ratings nor author characteristics predicted talk quality, number of questions, or attendance. However, review ratings from both systems similarly predicted judged poster presentation quality (for a subset of submissions) and eventual publication (for all submissions). These findings highlight limitations in both single- and double-blind approaches. Double-blind review is not a cure-all for inequities that might arise via single-blind. Yet, including author identities does not appear to enhance reliability or validity enough to justify the risks this information poses in potentially advantaging a particular individual or group. This paper was accepted by Jack Soll, behavioral economics and data analysis. Funding: This work was supported by the National Science Foundation [Grant SBE 1824259]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.01646 .


  • Open Access

    Blinded versus unblinded review: A field study comparing the equity of peer review

    by Timothy Joseph Pleskac, Ellie Kyung, Gretchen B. Chapman, Oleg Urminsky

    Blinded review is often proposed as a solution to inequities in merit-based evaluations. We tested this claim in a high-stakes field experiment comparing single-blind review (reviewer identities withheld) and double-blind review (reviewer and author identities withheld) for 530 academic conference submissions. On reliability, both processes showed moderate reliability, leading to only 40\% overlap among top-rated submissions. This low overlap indicates that differences between review systems partly stem from noise in the review system. In terms of fairness, our analysis indicates that single-blind reviews favored more senior coauthors and first authors who were Ph.D. students or research scientists, while disfavoring Asian (vs. White) first authors. Meanwhile, male first authors received slightly higher ratings in both conditions (especially under double-blind), but the greater the proportion of male coauthors, the lower the review ratings were under double-blind (vs. single-blind). On validity, we find that for submissions accepted and presented as a talk, neither review ratings nor author characteristics predicted talk quality, number of questions, or attendance. However, review ratings from both systems similarly predicted judged poster presentation quality (for a subset of submissions) and eventual publication (for all submissions). These findings highlight limitations in both single- and double-blind approaches. Double-blind review is not a cure-all for inequities that might arise via single-blind. Yet, including author identities does not appear to enhance reliability or validity enough to justify the risks this information poses in potentially advantaging a particular individual or group.

  • Open Access

    The Scale Orientation Effect: How the Spatial and Magnitude Orientations of Scales Shape Survey Responses

    by Ellie J Kyung, Manoj Thomas, Aradhna Krishna

    Technology makes it possible for researchers and practitioners to change the orientation of scales. Scales can be oriented horizontally from left to right with values increasing (0 left–10 right) or decreasing (10 left–0 right), or vertically from top to bottom with values increasing (0 top–10 bottom) or decreasing (10 top–0 bottom). Does the congruence of a scale’s spatial orientation (whether it is horizontal vs. vertical) and magnitude orientation (the direction in which values increase) influence survey responses? Seven preregistered experiments (N = 6,002) and four preregistered supplemental experiments (N = 5,674) examine the effect of a scale’s spatial–magnitude congruence on net promoter scores (NPSs). The results show that scales with incongruent spatial–magnitude orientations result in lower NPS ratings than those with congruent ones. We propose that horizontal and vertical spatial orientations are associated with a specific magnitude orientation in the mind. Horizontal scales feel more intuitive with increasing (0 left–10 right) versus decreasing (10 left–0 right) magnitude. Vertical scales feel more intuitive with decreasing (10 top–0 bottom) versus increasing (0 top–10 bottom) magnitudes. Researchers and managers designing surveys to measure consequential constructs such as NPS should not change the spatial–magnitude orientation of their scales across companies, survey platforms, or time—such changes can render their data less comparable.

  • Open Access

    Search or Scroll: How Credibility Versus Likability Premiums Shape Consumer Following Decisions

    by Edith Shalev, Coby Morvinski, Meyrav Shoham, Ellie J Kyung

    Consumers’ choices of whom to follow on digital platforms shape their informational landscape. In an era where the credibility of informational sources is critical, this research examines two key questions: (1) When do consumers prioritize communicator credibility over likability in their decisions to follow? and (2) How do multiple credibility and likability cues interact to influence these decisions? Analyzing four large datasets from popular following-enabled platforms, we find consumers’ orientation towards content consumption—goal-directed (“search”) versus experiential (“scroll”)—is key. Communicator credibility drives following on search-driven platforms (Yelp, Goodreads), while likability drives following on scroll-driven ones (Twitter/X, Instagram). Aggregate communicator sentiment across multiple posts serves as a cross-platform indicator of credibility and likability, and its effect on follower-count differs by platform type. On scroll-driven platforms, communicators with positive aggregate sentiment benefit from a likeability premium, attracting the most followers; this preference for positivity is mitigated by the presence of alternative communicator likeability cues (e.g., using sociable language). On search-driven platforms, communicators with mixed aggregate sentiment benefit from a credibility premium, attracting the most followers; this preference for mixed sentiment is mitigated in the presence of alternative credibility cues (e.g., Yelp’s “Elite” badge). Implications for consumer protection and platform design are discussed.


  • Open Access

    Optimizing Animation Speed: Convex Effects on Perceived Waiting Time and Digital Customer Experience

    by Yu Ding, Ellie J Kyung

    Incidental waits are an unavoidable element of the consumer experience on any digital platform. Firms typically utilize no animation or single-speed, repeated animations during these waits. How might the use of such animations and their visual qualities influence perceived waiting time? Can simple design changes by managers influence the customer experience? Although prior research suggests a linear relationship in which faster animations reduce perceived waiting time, we find a convex relationship between animation speed and time perception: moderate-speed animations minimize perceived waiting time compared to no, slow-moving, or fast-moving animations (experiments 1a–1c). We refer to this effect as the convex effect of animation speed. This effect occurs when people use animation speed to infer wait time (experiment 2a) and because moderate-speed animations draw more attention than static images or faster animations (experiment 2b). Animations that introduce dual-attention elements (experiment 3a) or atypical animations (experiment 3b) shift attention away from movement speed, attenuating this effect. Finally, this effect extends to website click-to-landing rates (experiment 4), conversion rates (experiment 5), and product evaluations during mobile shopping (experiment 6). These findings highlight the practical value of optimizing animation speed in user interface design to enhance both customer experience and business outcomes.


  • AI-induced indifference: Unfair AI reduces prosociality

    by Raina Zexuan Zhang, Ellie J. Kyung, Chiara Longoni, Luca Cian, Kellen Mrkva


  • Open Access

    AI-induced Indifference: Unfair AI Reduces Prosociality

    by Raina Zexuan Zhang, Ellie Kyung, Chiara Longoni, Luca Cian, Kellen Mrkva

    The growing prevalence of artificial intelligence (AI) in our lives has brought the impact of AI-based decisions on human judgments to the forefront of academic scholarship and public debate. Despite a massive growth in research related to receptivity towards AI, little is known about how interactions between humans and AI shape subsequent interactions among humans. We explore this question in the context of unfair decisions determined by AI versus humans and focus on the spillover effects of such interactions on prosociality. Four experiments (combined N = 2425) show that receiving an unfair allocation by an AI (vs. a human) actor leads to a lower rate of prosocial punishment in a subsequent decision—an effect we call AI-induced indifference. In Experiment 1, after receiving an unfair monetary allocation by an AI (versus a human) actor, people are less likely to altruistically sanction a human actor at a personal cost in a subsequent, unrelated decision. Experiments 2a and 2b provide evidence of the underlying mechanism: People blame AI actors less than their human counterparts for unfair behavior, decreasing their desire to subsequently sanction injustice. In an incentive-compatible design, Experiment 3 shows that AI-induced indifference manifests even when the interaction with AI occurs in a different domain than the human-to-human interaction. These findings directly address the spillover effect of human-AI interaction on human-to-human interactions and suggest that interacting with unfair AI may desensitize people to the behavior of unfair human actors. Implications for future research are discussed.


  • Top Cited Papers

    Algorithmic Transference: People Overgeneralize Failures of AI in the Government

    by Chiara Longoni, Luca Cian, Ellie J. Kyung

    Artificial intelligence (AI) is pervading the government and transforming how public services are provided to consumers across policy areas spanning allocation of government benefits, law enforcement, risk monitoring, and the provision of services. Despite technological improvements, AI systems are fallible and may err. How do consumers respond when learning of AI failures? In 13 preregistered studies (N = 3,724) across a range of policy areas, the authors show that algorithmic failures are generalized more broadly than human failures. This effect is termed “algorithmic transference” as it is an inferential process that generalizes (i.e., transfers) information about one member of a group to another member of that same group. Rather than reflecting generalized algorithm aversion, algorithmic transference is rooted in social categorization: it stems from how people perceive a group of AI systems versus a group of humans. Because AI systems are perceived as more homogeneous than people, failure information about one AI algorithm is transferred to another algorithm to a greater extent than failure information about a person is transferred to another person. Capturing AI's impact on consumers and societies, these results show how the premature or mismanaged deployment of faulty AI technologies may undermine the very institutions that AI systems are meant to modernize.