Does Positive Reinforcement Work?: A Quasi-Experimental Study of the Effects of Positive Feedback on Reddit

Social Platform Design & User BehaviorContent Moderation & Platform GovernanceContent Governance & Platform Compliance TeamsSociologists & Anthropologists

Research Background and Issues

  • What problems or challenges did the authors identify?
    This paper investigates the positive feedback mechanisms on social media platforms, particularly the impact of the widely used "Gold" and "Upvotes" features in Reddit communities on user behavior. Although positive feedback is theoretically believed to motivate users and encourage the production of high-quality content, there is still a lack of empirical research on its effects on individual user behavior and the overall community.

  • Why is this issue important?
    The success of online communities depends on users consistently contributing high-quality content, and effective incentive mechanisms can help users adhere to community rules and increase engagement. Understanding the effects of positive feedback can help optimize community management, making online communities healthier and more efficient.

  • Research Motivation and Related Work
    The authors draw on the theory of positive reinforcement in behavioral psychology and the framework of distributed community management to propose their research motivation. Using data from Reddit, they explore whether positive feedback effectively promotes higher-quality content, increases user engagement, and helps users adapt to community norms.


Solutions

  • What methods or solutions did the authors propose?
    The authors employed causal inference analysis methods to study how two primary positive feedback mechanisms ("Gold" and "Upvotes") influence users' future behavior, using a large dataset from Reddit (approximately 11 million posts).

  • What is innovative about this solution?
    The study combines the theory of positive reinforcement in behavioral psychology with data-driven analytical methods from the field of social computing. It examines the effects of "Gold" and "Upvotes" as variables on different user groups and communities. Additionally, a two-week observation window was designed to analyze the persistence of the effects of positive feedback.

  • What are the implementation steps and key techniques used?

    1. Data Collection and Processing: The Pushshift dataset was used to collect Reddit posts from May to September 2020, extracting posts that received "Gold" or high scores ("Upvotes").
    2. Matching Process: Stratified matching and propensity score matching were used to ensure that the characteristics of the treatment group (users receiving positive feedback) and the control group (users not receiving positive feedback) were similar.
    3. Difference Analysis: The causal effects of positive feedback on users' future behavior were analyzed, including content quality, community recognition, posting frequency, and content removal rate.
    4. Individual and Community-Level Analysis: Individual Treatment Effects (ITE) and Community Treatment Effects (CTE) were calculated separately, exploring which types of users and communities are more significantly influenced by positive feedback.

Research Findings

  • What specific findings were achieved?

    1. Users who received positive feedback showed significant improvements in the quality of their future posts and community recognition. Posts that received "Gold" awards saw their future scores increase by approximately 19.28%, while posts with high "Upvotes" scores saw a 57.11% increase.
    2. High scores ("Upvotes") not only encouraged users to post more frequently but also reduced behaviors that violated community norms (lower removal rates).
    3. Newer users (accounts created more recently) were more strongly influenced by high scores, with significant improvements in engagement and adherence to community norms.
  • What advantages does it have compared to existing solutions?
    Compared to traditional community management strategies based on punitive measures such as content removal or bans, positive feedback mechanisms more proactively encourage users to adhere to community norms while enhancing their creative engagement.

  • What were the experimental or evaluation results?
    The persistence of positive feedback effects varied depending on the specific mechanism: the effects of high scores remained significant within the observation window (14 days), while the effects of "Gold" diminished more quickly. Additionally, new users responded more positively to feedback, indicating that positive feedback has a unique advantage in helping newcomers familiarize themselves with community rules.

  • Limitations and Future Directions

    • Limitations:
      1. The analysis did not examine in detail whether the content receiving positive feedback truly "deserved" the rewards but was based on the community's overall scoring mechanism.
      2. The impact of positive feedback on bystanders (non-recipients) and its effects compared to specific administrator feedback (e.g., boosting post rankings) remain unexplored.
    • Future Directions:
      1. Develop more sophisticated content quality prediction models to uncover the underlying definitions of high-quality content as perceived by the community.
      2. Expand the research to explore the effects of positive feedback on comments or other content forms.
      3. Investigate whether bystanders can learn community norms by observing rewarded content.

Conclusion

This paper demonstrates the effectiveness of positive feedback in enhancing user engagement and content quality, as well as promoting the adoption of community norms. It not only provides theoretical support for improving user incentive mechanisms in online communities but also offers a series of design recommendations, such as prioritizing support for newcomers or highlighting positive content to enhance community management. These findings hold significant implications for the design of online platforms, content management teams, and scholars studying community behavior.

Quick Actions

Share

Share this page

ios_share

https://hci.top/en/papers/chi/188492/2025

AdRecommended

Learn AI Coding at CodeNow

open_in_newOpen DOI Link
DOI: https://dl.acm.org/doi/10.1145/3706598.3713830
At a Glance

Paper Snapshot

fact_check
dataset
Source
CHI
calendar_month
Year
2025
emoji_events
Award
No award tagged
group
Authors
3 authors
sell
Subtopics
Social Platform Design & User Behavior, Content Moderation & Platform Governance
work
Professions
Content Governance & Platform Compliance Teams, Sociologists & Anthropologists
article
Content Status
Full text indexed
hub
Related Papers
6 related papers