About

Eshwar Chandrasekharan is an Assistant Professor at The Grainger College of Engineering, University of Illinois Urbana-Champaign. His research intersects with Social Computing, Computational Social Science, Human-Computer Interaction (HCI), Natural Language Processing (NLP), and Human-Centered Artificial Intelligence. He studies how technology influences what people create, curate, and consume online, focusing on building technologies to make the Internet safer and more welcoming. His recent work involves using a combination of computational techniques and social computing theories to address issues related to online safety and inclusivity.

Research topics

• Computer Science
• Sociology
• Artificial Intelligence
• Political Science
• Social Science
• Internet privacy
• Computer Security
• World Wide Web
• Data science
• Psychology
• Social psychology
• Business
• Engineering
• Law
• Human–computer interaction
• Software engineering
• Public relations
• Advertising
• Economics

Selected publications

• Social Simulacra in the Wild: AI Agent Communities on Moltbook

  ArXiv.org · 2026-03-17

  articleOpen access

  As autonomous LLM-based agents increasingly populate social platforms, understanding the dynamics of AI-agent communities becomes essential for both communication research and platform governance. We present the first large-scale empirical comparison of AI-agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. Structurally, we find that Moltbook exhibits extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross-community author overlap (33.8\% vs. 0.5\%). In terms of linguistic attributes, content generated by AI-agents is emotionally flattened, cognitively shifted toward assertion over exploration, and socially detached. These differences give rise to apparent community-level homogenization, but we show this is primarily a structural artifact of shared authorship. At the author level, individual agents are more identifiable than human users, driven by outlier stylistic profiles amplified by their extreme posting volume. As AI-mediated communication reshapes online discourse, our work offers an empirical foundation for understanding how multi-agent interaction gives rise to collective communication dynamics distinct from those of human communities.

  PublisherOA PDF

• "Think about it like you're a firefighter": Understanding How Reddit Moderators Use the Modqueue

  2026-04-13 · 1 citations

  preprintOpen accessSenior author

  On Reddit, the moderation queue (modqueue) is the platform’s primary interface for reviewing user-reported and automatically flagged content. Despite its central role in Reddit’s community-reliant moderation model, little is known about how moderators actually use it in practice. To address this gap, we surveyed 110 moderators, who collectively oversee more than 400 subreddits, to understand how the modqueue fits into their workflows and what its design enables or constrains. We find substantial variation in modqueue use: some moderators treat it as a daily checklist, others use it to identify patterns or emerging issues, and many routinely leave the interface to gather additional context or coordinate with teammates. Respondents also described persistent challenges, including coordination issues such as collisions, incomplete or noisy information signals, and friction created by fragmented interface versions and reliance on third-party tools. Taken together, we show the modqueue is neither a one-size-fits-all solution nor sufficient on its own for supporting moderator review. We outline opportunities for more modular, better-integrated moderation infrastructures that support both item-level review and broader governance activities, and that better align with the collaborative and value-driven nature of volunteer moderation on Reddit.

  PublisherDOI

• The Language of Approval: Identifying the Drivers of Positive Feedback Online

  2026-04-13

  articleOpen accessSenior author

  Positive feedback via likes and awards is central to online governance, yet which attributes of users’ posts elicit rewards—and how these vary across authors and communities—remains unclear. To examine this, we combine quasi-experimental causal inference with predictive modeling on 11M posts from 100 subreddits. We identify linguistic patterns and stylistic attributes causally linked to rewards, controlling for author reputation, timing, and community context. For example, overtly complicated language, tentative style, and toxicity reduce rewards. We use our set of curated features to train models that can detect highly-upvoted posts with high AUC. Our audit of community guidelines highlights a “policy-practice gap”—most rules focus primarily on civility and formatting requirements, with little emphasis on the attributes identified to drive positive feedback. These results inform the design of community guidelines, support interfaces that teach users how to craft desirable contributions, and moderation workflows that emphasize positive reinforcement over purely punitive enforcement.

  PublisherDOI

• Needling Through the Threads: A Visualization Tool for Navigating Threaded Online Discussions

  2026-04-13

  articleOpen accessSenior author

  Navigating large-scale online discussions is difficult due to their rapid pace and high volume of content. Platforms like Reddit employ “threads’’ to visually organize parallel discussions, but deep nesting obscures conversation flow. For moderators, this fragmentation compounds the difficulty of following evolving conversations and maintaining context across threads, which limits timely and effective moderation. In this paper, we present Needle, an interactive system that applies visual analytics to summarize key conversational metrics: activity, toxicity, and voting trends over time. Needle provides both high-level overviews and detailed breakdowns of threads, enabling moderators to identify priority areas without reading through entire nested conversations. Through a user study with ten Reddit moderators, we find that Needle provides a practical solution to maintain contextual understanding when navigating threaded discussions. Based on these findings, we propose design guidelines for future visualization-based tools that shape how people consume, interpret, and make sense of large-scale online discussions.

  PublisherDOI

• Social Simulacra in the Wild: AI Agent Communities on Moltbook

  arXiv (Cornell University) · 2026-03-17

  preprintOpen access

  As autonomous LLM-based agents increasingly populate social platforms, understanding the dynamics of AI-agent communities becomes essential for both communication research and platform governance. We present the first large-scale empirical comparison of AI-agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. Structurally, we find that Moltbook exhibits extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross-community author overlap (33.8\% vs. 0.5\%). In terms of linguistic attributes, content generated by AI-agents is emotionally flattened, cognitively shifted toward assertion over exploration, and socially detached. These differences give rise to apparent community-level homogenization, but we show this is primarily a structural artifact of shared authorship. At the author level, individual agents are more identifiable than human users, driven by outlier stylistic profiles amplified by their extreme posting volume. As AI-mediated communication reshapes online discourse, our work offers an empirical foundation for understanding how multi-agent interaction gives rise to collective communication dynamics distinct from those of human communities.

  PublisherDOI

• MoMoE: Mixture of Moderation Experts Framework for AI-Assisted Online Governance

  2025-01-01

  articleOpen accessSenior author

  Large language models (LLMs) have shown great potential in flagging harmful content in online communities.Yet, existing approaches for moderation require a separate model for every community and are opaque in their decision-making, limiting real-world adoption.We introduce Mixture of Moderation Experts (MoMoE), a modular, cross-community framework that adds post-hoc explanations to scalable content moderation.MoMoE orchestrates four operators-Allocate , Predict , Aggregate , Explain -and is instantiated as seven community-specialized experts (MoMoE Community ) and five norm-violation experts (MoMoE NormVio ).On 30 unseen subreddits, the best variants obtain Micro-F1 scores of 0.72 and 0.67, respectively, matching or surpassing strong fine-tuned baselines while consistently producing concise and reliable explanations.Although community-specialized experts deliver the highest peak accuracy, norm-violation experts provide steadier performance across domains.These findings show that MoMoE yields scalable, transparent moderation without needing per-community fine-tuning.More broadly, they suggest that lightweight, explainable expert ensembles can guide future NLP and HCI research on trustworthy human-AI governance of online communities.

  PublisherOA PDFDOI

• Venire: A Machine Learning-Guided Panel Review System for Community Content Moderation

  Proceedings of the ACM on Human-Computer Interaction · 2025-10-16

  articleOpen access

  Research into community content moderation often assumes that moderation teams govern with a single, unified voice. However, recent work has found that moderators disagree with one another at modest, but concerning rates. The problem is not the root disagreements themselves. Subjectivity in moderation is unavoidable, and there are clear benefits to including diverse perspectives within a moderation team. Instead, the crux of the issue is that, due to resource constraints, moderation decisions end up being made by individual decision-makers. The result is decision-making that is inconsistent, which is frustrating for community members. To address this, we develop Venire, an ML-backed system for panel review on Reddit. Venire uses a machine learning model trained on log data to identify the cases where moderators are most likely to disagree. Venire fast-tracks these cases for multi-person review. Ideally, Venire allows moderators to surface and resolve disagreements that would have otherwise gone unnoticed. We conduct three studies through which we design and evaluate Venire: a set of formative interviews with moderators, technical evaluations on two datasets, and a think-aloud study in which moderators used Venire to make decisions on real moderation cases. Quantitatively, we demonstrate that Venire is able to improve decision consistency and surface latent disagreements. Qualitatively, we find that Venire helps moderators resolve difficult moderation cases more confidently. Venire represents a novel paradigm for human-AI content moderation, and shifts the conversation from replacing human decision-making to supporting it.

  PublisherDOI

• Creator Hearts: Investigating the Impact Positive Signals from YouTube Creators in Shaping Comment Section Behavior

  2025-04-24 · 3 citations

  articleOpen accessSenior author
  PublisherDOI

• The Chilling: Identifying Strategic Antisocial Behavior Online and Examining the Impact on Journalists

  Proceedings of the ACM on Human-Computer Interaction · 2025-10-16 · 1 citations

  articleOpen access

  On social platforms like Twitter, strategic targeted attacks are becoming increasingly common, especially against vulnerable groups such as female journalists. Two key challenges in identifying strategic online behavior are the complex structure of online conversations and the hidden nature of potential strategies that drive user behavior. To address these, we develop a new tree-structured Transformer model that categorizes replies based on their hierarchical conversation structures, offering insights into the latent strategies underlying these interactions. Extensive experiments demonstrate that our proposed classification model can effectively detect different user groups--namely attackers, supporters, and bystanders--and their latent strategies. To demonstrate the utility of our approach, we apply this classifier to real-time Twitter data and conduct a series of quantitative analyses on the interactions between journalistswith diverse cultural backgrounds and different groups of users--attackers, supporters, and bystanders. Our classification approach allows us to not only explore strategic behaviors of attackers but also those of supporters and bystanders who engage in online interactions. When examining the impact of online attacks, we find a strong correlation between the presence of attackers' interactions and chilling effects , where journalists tend to slow their subsequent posting behavior. Additionally, we find that attackers tend to negatively influence the posting behavior of other users within these conversations. As conversations deepen, replies often deviate from original posts and get more toxic. This paper provides a deeper understanding of how different user groups engage in online discussions and highlights the detrimental effects of attacker presence on journalists, other users, and conversational outcomes. Our findings underscore the need for social platforms to develop tools that address coordinated toxicity and foster healthier conversation dynamics. By detecting patterns of coordinated attacks early, platforms could limit the visibility of toxic content to prevent escalation. Additionally, providing journalists and users with tools for real-time reporting and de-escalation could empower them to manage hostile interactions more effectively. Enhanced moderation tools targeting coordinated behaviors, particularly among attackers, could ensure a safer environment for vulnerable groups like female journalists, ultimately supporting constructive discussions and resilient online communities.

  PublisherDOI

• ArgCMV: An Argument Summarization Benchmark for the LLM-era

  2025-01-01

  articleOpen accessSenior author

  Key point extraction is an important task in argument summarization, which involves extracting high-level short summaries from arguments.Existing approaches for KP extraction have been mostly evaluated on the popular ArgKP21 dataset.In this paper, we highlight some of the major limitations of the ArgKP21 dataset and demonstrate the need for new benchmarks that are more representative of actual human conversations.Using SoTA large language models (LLMs), we curate a new argument key point extraction dataset called ArgCMV comprising of 12K arguments from actual online human debates spread across 3K topics.Our dataset exhibits higher complexity such as longer, coreferencing arguments, higher presence of subjective discourse units, and a larger range of topics over ArgKP21.We show that existing methods do not adapt well to ArgCMV and provide extensive benchmark results by experimenting with existing baselines and latest open source models.This work introduces a novel KP extraction dataset for long-context online discussions, setting the stage for the next generation of LLM-driven summarization research. 1

  PublisherOA PDFDOI

Frequent coauthors

• Éric Gilbert

  14 shared

• Charlotte Lambert

  University of Illinois Urbana-Champaign

  7 shared

• Stevie Chancellor

  University of Minnesota

  6 shared

• Tanvi Bajpai

  6 shared

• Koustuv Saha

  Microsoft Research (United Kingdom)

  6 shared

• David Jurgens

  6 shared

• Amy Bruckman

  5 shared

• Frederick D. S. Choi

  University of Illinois Urbana-Champaign

  5 shared

Awards & honors

• NSF CAREER Award