When "Likers'' Go Private: Engagement With Reputationally Risky Content on X

TL;DR

The study leverages X's shift to private likes to causally evaluate whether hiding liker identities alters engagement with high-reputational-risk content. It combines two Difference-in-Differences designs at the platform level with a within-subject user survey to triangulate mechanisms, using $n=154{,}122$ posts from $1{,}068$ accounts and $N=203$ participants. Across analyses, there is no detectable platform-level increase in likes for controversial content ($\mathit{effect}=0.01$, $p=0.564$), though self-reported willingness to like such content rises modestly under private visibility in the survey, indicating an intention–behavior gap. The results imply that visibility changes alone are unlikely to meaningfully shift sensitive engagement; effective interventions may require additional levers like algorithmic exposure control, timing of feedback, or audience segmentation to meaningfully alter engagement with reputationally risky content.

Abstract

In June 2024, X/Twitter changed likes' visibility from public to private, offering a rare, platform-level opportunity to study how the visibility of engagement signals affects users' behavior. Here, we investigate whether hiding liker identities increases the number of likes received by high-reputational-risk content, content for which public endorsement may carry high social or reputational costs due to its topic (e.g., politics) or the account context in which it appears (e.g., partisan accounts). To this end, we conduct two complementary studies: 1) a Difference-in-Differences analysis of engagement with 154,122 posts by 1068 accounts before and after the policy change. 2) a within-subject survey experiment with 203 X users on participants' self-reported willingness to like different kinds of content. We find no detectable platform-level increase in likes for high-reputational-risk content (Study 1). This finding is robust for both between-group comparison of high- versus low-reputational-risk accounts and within-group comparison across engagement types (i.e., likes vs. reposts). Additionally, while participants in the survey experiment report modest increases in willingness to like high-reputational-risk content under private versus public visibility, these increases do not lead to significant changes in the group-level average likelihood of liking posts (Study 2). Taken together, our results suggest that hiding likes produces a limited behavioral response at the platform level. This may be caused by a gap between user intention and behavior, or by engagement driven by a narrow set of high-usage or automated accounts.

Figures (8)

• Figure 1: Research overview. $\mathbb{X}$ made public likes private on June 12, 2024. After the change, the identities of likers are visible only to themselves and the post authors. To evaluate the behavioral impact of this change, we estimate (1) the additional changes in likes received by high-reputational-risk accounts after the design change, relative to the pre-change period and compared to low-reputational-risk accounts (between-group); and (2) the additional changes in likes to high-reputational-risk accounts after the design change, relative to the pre-change period and compared to reposts on the same accounts (within-group).
• Figure 2: Analysis of weekly effects of making likes private on the number of likes. (a) The weekly averages of likes received by high-reputational-risk and low-reputational-risk posts. (b) The weekly averages of likes and reposts for high-reputational-risk posts. (c) The estimated weekly (and aggregated) effects of private visibility on likes to high-reputational-risk posts relative to low-reputational-risk posts. (d) The estimated weekly (and aggregated) effects of private like visibility on the number of likes for high-reputational-risk posts relative to the number of reposts. The error bars represent 95% CIs. The grey dash lines in (c) and (d) represent equivalence bounds.
• Figure 3: The estimated aggregated effects across account categories in Study 1. The error bars represent 95% CIs.
• Figure 4: Reported channels through which participants first became aware of $\mathbb{X}$'s change to private like visibility. Error bars represent 95% CIs.
• Figure 5: Participants’ self-rated likelihood of liking posts on high-reputational-risk and low-reputational-risk topics if their likes were public or private ($N=$ 203.0). (a) Mean likelihood of liking (0–100 scale) for each topic when likes were public or private (group means). (b) Mean within-subject change ($\Delta$) in likelihood of liking between private and public conditions, plotted for all participants and for the subset who were aware of the platform’s policy change through news or posts. Positive values indicate greater liking when likes were private. The error bars represent 95% CIs.