Abstract
Human decision-making is subject to the biological limits of cognition. The fluidity of information propagation over online social media often leads users to experience information overload. This in turn affects which information received by users are processed and gain a response to, imposing constraints on volumes of, and participation in, information cascades. In this study, we investigate properties contributing to the visibility of online social media notifications by highly active users experiencing information overload via cross-platform social influence. We analyze simulations of a coupled agent-based model of information overload and the multi-action cascade model of conversation with evolutionary model discovery. Evolutionary model discovery automates mechanistic inference on agent-based models by enabling random forest importance analysis on genetically programmed agent-based model rules. The mechanisms of information overload have shown to contribute to a multitude of global properties of online information cascades. We investigate nine characteristics of online messages that may contribute to the prioritization of messages for response. Our results indicate that recency had the largest contribution to message visibility, with individuals prioritizing more recent notifications. Global popularity of the conversation originator had the second highest contribution, and reduced message visibility. Messages that presented opportunity for novel user interaction, yet high reciprocity showed to have relatively moderate contribution to message visibility. Finally, insights from the evolutionary model discovery results helped inform response prioritization rules, which improved the robustness and accuracy of the model of information overload.
Introduction
In the modern digital landscape, users are inundated with a constant stream of notifications, leading to significant "information overload." While human cognitive capacities remain relatively fixed, the volume and velocity of information propagation on social media have increased exponentially. To manage this, platforms utilize various "scaffolds," such as notification lists and news feeds, yet the underlying mechanisms by which users prioritize specific messages for response remain poorly understood.
Traditional diffusion models often simplify information engagement to a binary state. This research addresses this gap by identifying the specific message characteristics—such as recency, sender popularity, and content novelty—that drive response behavior among highly active users. By understanding these mechanics, we can better model how information cascades are constrained by human attention limits.
Methodology: Evolutionary Model Discovery
The researchers employed Evolutionary Model Discovery (EMD), a sophisticated framework that combines agent-based modeling with genetic programming and machine learning to infer behavioral rules from data.
- Coupled Modeling: The study integrated the Multi-Action Cascade Model (MACM), which treats conversations as branching trees of actions (initiating, contributing, sharing), with an Information Overload Model (IOM) that simulates a user's limited attention as a prioritized queue.
- Genetic Programming (GP): Instead of assuming a fixed prioritization formula, a GP algorithm evolved thousands of potential "utility functions" based on nine hypothesized factors.
- Random Forest Analysis: These evolved rules were then analyzed using a Random Forest regressor to determine the relative importance of each factor in predicting real-world user behavior.
The methodology was applied to a cross-platform dataset tracking discussions about cyber-vulnerabilities (CVEs) on GitHub, Reddit, and Twitter, focusing on the behavior of the most active community members.
Key Results and Prioritization Drivers
The analysis identified several key factors that govern how users prioritize responses under conditions of overload:
- Recency is Paramount: The strongest predictor of response was the recency of the message ($F_{Recn}$). This confirms that for overloaded users, notifications that are not immediately visible at the top of a feed are likely to be ignored.
- The Anti-Popularity Bias: Interestingly, the global popularity of a conversation's initiator ($F_{InitPop}$) was negatively correlated with response priority. Users in these specialized communities prioritized local relevance and engagement with "regular" peers over high-status "superstars."
- Novelty vs. Reciprocity: Users showed a preference for novel interactions—content they had not previously engaged with—while also favoring individuals with whom they had a history of mutual response (reciprocity).
Models incorporating these factors significantly outperformed baseline models in replicating real-world responsiveness, demonstrating the robustness of the EMD approach.
Discussion and Conclusion
This study highlights the role of social media interfaces as cognitive scaffolds that extend human working memory. The findings suggest that prioritization is not merely a matter of content quality but is deeply influenced by the structural properties of the information flow and the user's social network.
While the results are specific to a technical community, they provide a foundation for future research into how platform algorithms might be re-wiring human cognitive strategies. Understanding these evolved rules could eventually lead to the development of more effective AI-driven notification filters that align with human prioritization patterns to mitigate the negative impacts of information overload.
Summary updated