Within Algorithms
Do Clicks Show What People Want?
Clicks, likes and watch time can reward content people later judge to be unhealthy, misleading or simply not worth seeing.
On this page
- Revealed preferences versus reflective preferences
- Why curiosity and anger can look like satisfaction
- What healthier ranking objectives might measure
Page outline Jump by section
Introduction
The assumption behind many social media ranking systems is simple: if people click, watch, like, comment or share something, they must want more of it. Yet a growing body of research suggests that engagement and satisfaction are not the same thing. People often interact with content because it is surprising, infuriating, emotionally charged or difficult to ignore, even when they later judge it to be misleading, unhealthy or a poor use of their time. This gap matters for myths and misconceptions because engagement-based systems can end up rewarding the very content that users say they wish they saw less of. The result is a ranking problem: platforms can become highly effective at predicting behaviour while remaining surprisingly poor at predicting what users ultimately value. [OUP Academic]academic.oup.comuser engagement such as clicks, shares, and likes.Read moreOUP AcademicEngagement, user satisfaction, and the amplification of…by S Milli · 2025 · Cited by 244 — Social media ranking algorithms…
Do Clicks Show What People Want?
Revealed preferences versus reflective preferences
Researchers often distinguish between revealed preferences and reflective preferences. Revealed preferences are inferred from behaviour: what people click, watch or share. Reflective preferences are what people say they want after thinking about their goals, values and long-term interests.
Social media platforms largely rely on revealed preferences because they are easy to measure at scale. Every tap, pause, comment and share produces data. Reflective preferences are harder to capture because they require surveys, self-assessment or direct questions about what makes an experience worthwhile. [OUP Academic]academic.oup.comuser engagement such as clicks, shares, and likes.Read moreOUP AcademicEngagement, user satisfaction, and the amplification of…by S Milli · 2025 · Cited by 244 — Social media ranking algorithms…
A particularly relevant study examined the relationship between engagement-based ranking and user satisfaction on political content. Researchers conducted a preregistered audit of Twitter’s ranking system and found that engagement-based ranking amplified emotionally charged and hostile political posts. Crucially, users did not consistently prefer the content selected by the algorithm when asked about their actual preferences. The algorithm was effective at generating engagement but less effective at delivering content users considered desirable. [OUP Academic]academic.oup.comuser engagement such as clicks, shares, and likes.Read moreOUP AcademicEngagement, user satisfaction, and the amplification of…by S Milli · 2025 · Cited by 244 — Social media ranking algorithms…
More recent research on young adults’ news-consumption habits found a similar pattern. Participants frequently engaged with low-quality content despite reporting a preference for accurate, diverse and higher-quality information. When asked to design an ideal news feed, they created feeds that differed noticeably from the feeds implied by their engagement histories. [arXiv]arxiv.orgSource details in endnotes.
For discussions of myths and misconceptions, this distinction is important. A person’s behaviour may indicate interest in a rumour, conspiracy claim or inflammatory story, but that does not necessarily mean they believe it, endorse it or wish it occupied more space in their feed.
Why Curiosity and Anger Can Look Like Satisfaction
Engagement metrics measure actions. They do not directly measure whether those actions improved a person’s experience.
A misleading claim may attract attention precisely because it appears shocking or implausible. A user may click in order to verify it, mock it or argue against it. Similarly, angry content often generates comments and shares because people feel compelled to respond. From the perspective of a ranking algorithm, these reactions can resemble evidence that the content is valuable.
The Twitter audit provides a concrete example. The researchers found that engagement-based ranking amplified content expressing hostility towards political out-groups. Users reported that such content made them feel worse about those groups, yet the ranking system rewarded it because it generated strong behavioural signals. [arXiv]arxiv.orgSource details in endnotes.
Theoretical work by economists and computer scientists has argued that this mismatch may be unavoidable when systems infer welfare solely from behaviour. People frequently make choices that conflict with their longer-term preferences. A platform can therefore optimise engagement perfectly while still moving users away from experiences they would later judge as better for them. [arXiv]arxiv.orgSource details in endnotes.
This helps explain why myths can thrive in engagement-driven environments. False or exaggerated claims often trigger exactly the reactions that algorithms interpret as success:
- Surprise and curiosity.
- Outrage and moral condemnation.
- Identity-based defence of a group or cause.
- Arguments between supporters and critics.
- Repeated checking for updates.
Each of these behaviours increases measurable engagement, even if users later conclude that the content was misleading or not worth their attention. [ifo Institut]ifo.deifo InstitutRanking for Engagement: How Social Media Algorithms…by F Germano · Cited by 16 — This paper investigates the dynamic feedb…
What One Dataset Revealed About the Metric Gap
The audit of Twitter’s engagement-based ranking system offers one of the clearest datasets demonstrating the metric gap.
The researchers compared an engagement-ranked feed with a reverse-chronological baseline. They found that the engagement-based system increased the visibility of emotionally charged and out-group hostile political content. At the same time, survey responses indicated that users did not prefer many of the posts that the algorithm elevated. In other words, engagement and user-reported satisfaction diverged. [arXiv]arxiv.orgSource details in endnotes.
This finding matters because it challenges a common defence of engagement metrics: that they merely reflect what people choose. The study suggests that what attracts attention in the moment is not always what users endorse after reflection.
The result also provides a plausible mechanism for myth amplification. Content that provokes emotional reactions can accumulate engagement signals faster than careful explanations or corrections. If ranking systems treat those signals as evidence of value, myths may receive greater visibility even when users themselves would prefer a healthier information environment. [arXiv]arxiv.orgSource details in endnotes.
What Healthier Ranking Objectives Might Measure
Recognising the limitations of clicks and watch time does not automatically reveal a better metric. Researchers have therefore explored alternatives that attempt to capture user welfare more directly.
Several possibilities have emerged:
- User satisfaction surveys: asking users whether content was worth seeing rather than inferring value from behaviour alone.
- Quality and accuracy indicators: incorporating signals about credibility and informational value.
- Diversity measures: ensuring feeds do not become dominated by a narrow range of emotionally engaging material.
- Long-term outcomes: measuring whether users feel informed, connected or satisfied after using a platform rather than only during use.
- Preference-aware ranking: allowing users to express values such as accuracy, civility or relevance and incorporating those values into ranking decisions. [arXiv]arxiv.orgSource details in endnotes.
The Twitter audit tested one version of this idea by ranking content using users’ stated preferences rather than engagement alone. The alternative approach reduced angry, partisan and hostile content, though it introduced new trade-offs, including a tendency to reinforce some existing attitudes. The lesson was not that a perfect metric exists, but that engagement is only one possible objective among many. [arXiv]arxiv.orgSource details in endnotes.
Why the Metric Gap Matters for Myths and Misconceptions
A common misconception about social media algorithms is that they simply give people what they want. The evidence suggests a more complicated reality. Engagement metrics are powerful tools for predicting behaviour, but behaviour is not always a reliable guide to what users regard as valuable, accurate or beneficial.
When platforms rely heavily on clicks, likes, comments and watch time, they risk confusing attention with satisfaction. In environments where myths compete with corrections, that confusion can become consequential. Content that is emotionally irresistible may outperform content that is genuinely useful, not because users consciously prefer it, but because engagement metrics capture immediate reactions more easily than considered judgment. [OUP Academic]academic.oup.comuser engagement such as clicks, shares, and likes.Read moreOUP AcademicEngagement, user satisfaction, and the amplification of…by S Milli · 2025 · Cited by 244 — Social media ranking algorithms… 2arXiv
eBay marketplace picks
Marketplace Samples
Example marketplace items related to this page. Use the search link to explore similar finds on eBay.
Endnotes
-
Source: academic.oup.com
Title: user engagement such as clicks, shares, and likes.Read more
Link: https://academic.oup.com/pnasnexus/article/4/3/pgaf062/8052060Source snippet
OUP AcademicEngagement, user satisfaction, and the amplification of...by S Milli · 2025 · Cited by 244 — Social media ranking algorithms...
-
Source: arxiv.org
Link: https://arxiv.org/abs/2202.11776 -
Source: arxiv.org
Link: https://arxiv.org/abs/2305.16941 -
Source: arxiv.org
Link: https://arxiv.org/abs/2604.11517Source snippet
arXivUnderstanding the Gap Between Stated and Revealed Preferences in News Curation: A Study of Young Adult Social Media UsersApril 13, 2026...
Published: April 13, 2026
-
Source: arxiv.org
Link: https://arxiv.org/html/2604.11517v1Source snippet
A Study of Young Adult Social Media Users13 Apr 2026 — The gap between stated and revealed preferences was measured by identifying cases...
-
Source: ifo.de
Link: https://www.ifo.de/en/cesifo/publications/2026/working-paper/ranking-engagement-how-social-media-algorithms-fuel-misinformationSource snippet
ifo InstitutRanking for Engagement: How Social Media Algorithms...by F Germano · Cited by 16 — This paper investigates the dynamic feedb...
-
Source: statista.com
Title: Social media
Link: https://www.statista.com/topics/1164/social-networks/?srsltid=AfmBOooSZhxkomCZn4fiPBsu0acLBI6RhMWf0DxMg2uGdqt1uBpD0mSiSource snippet
statistics & factsAs of early 2024, Facebook reported a year-on-year audience growth of 3.1 percent, demonstrating that the social media...
Additional References
-
Source: linkedin.com
Link: https://www.linkedin.com/pulse/crafting-meaningful-social-interaction-metric-facebook-growjunction-slndcSource snippet
LinkedInCrafting a Meaningful Social Interaction Metric for FacebookCrafting a meaningful social interaction metric for Facebook involves...
-
Source: papers.ssrn.com
Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5316506Source snippet
for Engagement: How Social Media Algorithms...by F Germano · Cited by 16 — This paper investigates the dynamic feedback loop between rec...
-
Source: facebook.com
Link: https://www.facebook.com/groups/NextUpAsia/posts/10160654553136657/Source snippet
Why Social Media Content Gets Low Views and LikesOne major cause of low views is lack of engagement. Facebook's algorithm rewards posts t...
-
Source: journalistsresource.org
Link: https://journalistsresource.org/media/facebook-engagement-patterns/Source snippet
Why most Facebook users get more than they giveThe average users in the sample “like” their friends' Facebook posts 14 times a month, but...
-
Source: facebook.com
Link: https://www.facebook.com/groups/1163865390677752/posts/2555700461494231/Source snippet
Why people prefer short facebook postsMost people who come on Facebook do not want to read long posts, and scroll past, once they realize...
-
Source: starsmedia.com
Link: https://www.starsmedia.com/wp-content/uploads/2022/09/Customer_Engagement_Social_Media_Framework_Meta_Analysis_AAM.pdfSource snippet
ent in social media (CESM) using a meta-analytic model of 814 effect sizes across 97 studies involving 161,059...Read...
-
Source: researchgate.net
Link: https://www.researchgate.net/publication/365032175_View_Like_Comment_Post_Analyzing_User_Engagement_by_Topic_at_4_Levels_across_5_Social_Media_Platforms_for_53_News_OrganizationsSource snippet
s five social media platforms (i.e., Facebook, Instagram, Twitter, YouTube...Read more...
-
Source: facebook.com
Title: Users must genuinely like or want to share and discuss content.Read more
Link: https://www.facebook.com/groups/3868728716717515/posts/4456305231293191/Source snippet
Facebook's interaction metrics change, focus on valuable...Facebook prioritizes meaningful social interactions over content consumption...
-
Source: facebook.com
Link: https://www.facebook.com/FacebookforCreators/posts/now-that-we-have-your-attention-lets-dig-into-the-real-meaning-behind-these-metr/1414027426754349/Source snippet
orithm loves you, boosting your reach organically. Likes: 3,857...
-
Source: knightcolumbia.org
Link: https://knightcolumbia.org/content/engagement-user-satisfaction-and-the-amplification-of-divisive-content-on-social-mediaSource snippet
Knight First Amendment InstituteEngagement, User Satisfaction, and the Amplification of...by S Milli · Cited by 3 — Social media ranking...
Topic Tree