Source-linked AI summary
The anatomy of Reddit: An overview of academic research
Alexey N. Medvedev, Renaud Lambiotte, Jean-Charles Delvenne
TL;DR
Online forums offer rich data for studying collective thinking and practical platform behavior, but research needs reliable evidence. This survey maps Reddit research by post- and user-focused questions, synthesizes structural and dynamic methods, and identifies the platform’s research richness alongside data and modeling limits.
Problem
Researchers need to understand how behavior in online forums relates to collective thinking and practical outcomes using evidence that is broadly available for study.
Method
The survey categorizes Reddit research by focus on posts or users and reviews methods using system structure, dynamics, and textual features.
Results
The review finds that Reddit supports diverse research through rich data, while discussion-tree structure is relatively well understood and community prediction increasingly combines structural, dynamic, and textual features.
Takeaways & Limitations
Reddit’s diverse data and reconstructed networks support research on users, posts, communities, and future system dynamics.
Takeaways & Limitations
Heavy-tailed activity and popularity statistics may introduce sampling issues, while missing and corrupted records constrain some analyses.
Abstract
from arXiv · showhide
Online forums provide rich environments where users may post questions and comments about different topics. Understanding how people behave in online forums may shed light on the fundamental mechanisms by which collective thinking emerges in a group of individuals, but it has also important practical applications, for instance to improve user experience, increase engagement or automatically identify bullying. Importantly, the datasets generated by the activity of the users are often openly available for researchers, in contrast to other sources of data in computational social science. In this survey, we map the main research directions that arose in recent years and focus primarily on the most popular platform, Reddit. We distinguish and categorise research depending on their focus on the posts or on the users, and point to different types of methodologies to extract information from the structure and dynamics of the system. We emphasize the diversity and richness of the research in terms of questions and methods, and suggest future avenues of research.
1 Introduction
Online forums support large-scale study of communication, collective thinking, and practical platform outcomes. This survey reviews online discussion research, focusing on Reddit and organizing studies by their research focus.
- Online forums let researchers study opinion exchange, collective decisions, and collective thinking at scale.
- Forum research also has practical applications including improving user experience, increasing engagement, and facilitating democratic processes.
- The survey brings together research questions on online discussion platforms, focusing primarily on Reddit as the largest online discussion forum.
- The review organizes studies around datasets, post-focused research, and user-focused research before discussing future perspectives.
2 The Reddit dataset
Reddit combines ranked, topic-based content feeds with vote-based feedback and tree-structured discussions. Its large, open dataset supports longitudinal research, but missing, deleted, and corrupted records impose important limits on analysis.
- Users can upvote or downvote posts and comments, while registered users follow subreddits rather than individual users.
- Comments form rooted discussion trees in which the post is the root and reply-to relations connect comments.
- Reddit organizes content into topic-based subreddits, whose posts appear in feeds ranked by time and voting score.
- Reddit’s full historical dataset enables longitudinal studies and supports reproducibility because posts and comments are broadly available.
- Around 80% of posts before 2008 were reported missing, while 0.043% of comments and 0.65% of posts were missing across January 2006–February 2016.
- Deleted accounts, posts, and comments complicate user-participation, reply-network, and historical text analyses.
3 From the perspective of posts
Research on Reddit posts examines popularity, discussion structure, and predictive modeling using structural, temporal, textual, and author-related features. Studies also analyze how content quality, subreddit context, and platform design shape scores and discussion dynamics.
- Popularity prediction: Popularity research models post scores and discussion activity using structural, temporal, textual, and author-related features.Methods range from regression, Poisson, and Cox models to neural networks.
- Popularity prediction: Average post score is proportional to discussion-tree size, but individual score prediction requires more tailored estimation.The proportional relation is an average trend rather than a sufficient predictor for each submission.
- Popularity prediction: Comment scores reflect timing, relevance, novelty, discussion structure, and subreddit-specific preferences.High-scoring comments tend to appear early; low-scoring comments tend to arrive later, while stale memes and high user ranking do not improve scores.
- Popularity prediction: Adding linguistic features produced marginal gains in comment ranking, while an LSTM increased average F1 from 50 to 54.The LSTM combined structural, temporal, and textual features across three subreddits.
- Discussion trees: Generative and predictive models increasingly connect discussion-tree structure with temporal dynamics and can outperform contemporary cascade-dynamics models.One model jointly represents structure and dynamics to predict discussion flow.
- Discussion trees: Discussion-tree studies report root bias, variation in tree depth and width, and better reproduction by branching-tree models than preferential attachment models.Political discussions and political subreddits show larger depth and width, with subreddit differences partly associated with participation rates.
4 From the perspective of users
Research on Reddit users examines activity patterns, community loyalty, migration, moderation, and networks through observational and data-oriented methods. These studies show how users organize across communities and how their behavior relates to participation, retention, and interaction.
- Research approach: User-focused Reddit research primarily uses behavioral records and interaction networks to study activity and community organization.The reviewed approaches are mainly observational and data-oriented.
- Activity patterns: Only 16% of users produced more than half of recorded interactions, while users voted after viewing titles for 73% of posts.A further 17% followed links and browsed comments before voting.
- Community loyalty: Users generally explored new communities before distributing activity across a small set of communities with diverse interests.Posting patterns and first-post scores helped predict future settlement, whereas activity rate alone performed poorly.
- Community loyalty: Loyal communities had smaller but denser interaction networks, and linguistic features predicted loyalty with 63.6% average classification accuracy.These networks were less assortative and less clustered, indicating less fragmentation into groups.
- Communities and networks: Community studies examined migration, mobilization, hateful interactions, moderation, and network structure across Reddit and related discussion platforms.Examples include hate-speech subreddit bans, deleted-comment analysis, cross-community attacks, and reply-based social networks.
5 Discussion
The discussion highlights Reddit’s unusually rich, diverse data and surveys how its structure, dynamics, and communities support theoretical and applied research. It also identifies open directions involving user activity, community evolution, discussion dynamics, and computational methods, while cautioning that heavy-tailed statistics complicate sampling.
- Data richness: Reddit datasets combine potentially exhaustive access with user networks, text, social scores, discussion trees, and years of temporal activity.This diversity supports research on both social processes and platform-specific phenomena.
- Structure: Discussion-tree structure varies across platforms; on Reddit, trees are broadly distributed in size and are better modeled as near-uniform branching trees than preferential-attachment networks.The analysis also observes root bias and a characteristic depth/width profile.
- Research scope: Reddit data support investigations of communities, agreement, trolling, echo chambers, polarization, social manipulation, and platform activity.Applied problems include detecting and removing trolling or attacks and stimulating forum activity.
- Methodological caveat: Heavy-tailed activity, popularity, and post-success statistics can make random samples miss influential extreme cases, requiring caution with methods such as neural networks.The survey reports that combining structure, dynamics, and text has mainly been studied with black-box machine learning, including neural networks.
- Future directions: Future work includes more detailed user-activity studies, reconstructed community or user networks, evolving graph models, and analyses linking discussion-tree structure with dynamics.These directions target prediction, platform design, user experience, and community resilience.
- Future directions: Posting and commenting are coupled but not necessarily proportional, creating open questions about the growth and resilience of online communities.Figure 7 illustrates scenarios in which comment rates exceed posting rates, remain similar, or disappear while posts increase.