Drawing on epidemiology, computer science and moral psychology, Tope Christopher Falade’s studies examine how direct interactions and coordinated groups shape harmful online conversations.
Online toxicity is often treated as a content problem: an inflammatory post appears, users react, and hostility spreads across the discussion. Research by Tope Christopher Falade suggests that this explanation overlooks a more immediate influence the relationship between one comment and the reply that follows it.
Falade, a Ph.D. candidate in Computer and Information Science at the University of Arkansas at Little Rock, studies how toxic behavior and misinformation move through online communities. His research draws on methods from network science, artificial intelligence, moral psychology and public-health epidemiology.
In one study, Falade examined more than one million comments from a health-related Reddit community that was later banned for spreading vaccine misinformation. He traced conversations comment by comment to determine which factors most strongly predicted whether a response would become toxic.
The analysis found that the comment immediately preceding a reply was a stronger predictor of toxicity than either the original post or the broader tone of the online community. The finding indicates that harmful exchanges are shaped substantially by direct interactions between users rather than solely by the subject under discussion.
Falade subsequently tested the pattern using 5.2 million Telegram posts from political channels discussing the Russia–Ukraine war. Despite the differences in platform, subject matter and user behavior, the analysis produced a similar result: the tone of the message to which a person was directly responding had a greater influence on the reply than the general topic of the conversation.
His initial study, presented at the Americas Conference on Information Systems in 2024, had received more than 20 scholarly citations by November 2025. The citations indicate that researchers working in related areas have used or discussed the study in subsequent academic work.
Falade’s later research shifted from individual exchanges to the behavior of entire networks. He adapted an epidemiological approach normally used to study how disease spreads through populations and applied it to the transmission of toxic behavior among connected social-media accounts.
That analysis identified a distinction between individual influence and coordinated influence. According to the findings, a small group of accounts acting together could spread toxicity farther and faster than the most prominent individual account in the same network.
The result has implications for content-moderation systems that concentrate primarily on highly visible users or isolated posts. A network may consist of accounts that appear unremarkable when examined separately but exert greater influence when they repeatedly interact, reinforce one another’s messages or distribute similar material across connected communities.
This form of coordination has become a concern in research on election interference, conspiracy narratives and cross-platform influence campaigns. Studies surrounding the 2024 U.S. election documented networks that circulated partisan and conspiratorial content across multiple platforms. Falade’s work provides a framework for examining the interaction patterns through which such campaigns may gain reach before any single account attracts significant attention.
The same network dynamics also apply to public-health communication. During emergencies, misleading health information can circulate through closely connected communities more rapidly than corrections or official guidance. Falade’s combination of epidemiology and computational network analysis offers a way to study not only what information is being shared, but also how repeated interactions help it travel.
A separate part of his research addresses the limited transparency of automated moderation systems. Many artificial-intelligence tools classify a message as toxic without providing a clear explanation for the decision. That lack of interpretability can make it difficult for researchers, moderators and users to evaluate whether a classification is justified.
Falade developed a model grounded in moral-foundations theory, a framework from moral psychology that examines values such as care, fairness, loyalty and respect for authority. Rather than relying only on particular words, the model considers the moral values invoked or violated within a conversation.
Tested on approximately 1.27 million conversations, the model achieved a reported accuracy rate of 88.2 percent. It was also designed to provide reasons for its classifications, allowing users to examine the basis on which a message was flagged.
Falade’s research has been presented through peer-reviewed venues including the IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, the IEEE International Conference on Tools with Artificial Intelligence, and the Complex Networks and Their Applications conference. His broader publication record includes work presented at the Hawaii International Conference on System Sciences, where one paper was nominated for a best-paper award, as well as the CySoc workshop at the International AAAI Conference on Web and Social Media, the Pacific Asia Conference on Information Systems, Social Network Analysis and Mining, and the IEEE Journal of Social Computing.
His academic background spans public-health epidemiology, information systems management with a concentration in enterprise security, and information science. He holds three master’s degrees in those fields and is a member of the Institute of Electrical and Electronics Engineers.
Across the studies, a consistent argument emerges: online toxicity cannot be fully understood by examining vocabulary or individual posts in isolation. Direct exchanges, recurring relationships and coordinated groups provide signals that content-based systems may miss.
As researchers and technology platforms prepare for the increased online activity surrounding the 2026 U.S. midterm elections, Falade’s findings place attention on the structure of digital conversations who responds to whom, which accounts repeatedly act together and how those interactions allow harmful behavior to move through a network.
