Paper deep dive
BotVerse: Real-Time Event-Driven Simulation of Social Agents
Edoardo Allegrini, Edoardo Di Paolo, Angelo Spognardi, Marinella Petrocchi
Intelligence
Status: succeeded | Model: google/gemini-3.1-flash-lite-preview | Prompt: intel-v1 | Confidence: 96%
Last extracted: 4/1/2026, 1:37:41 AM
Summary
BotVerse is an event-driven, scalable multi-agent simulation framework designed for high-fidelity social simulation using LLM-based agents. It provides a controlled environment that grounds agent behavior in real-time content streams from the Bluesky ecosystem, enabling researchers to study social phenomena like disinformation spread without the ethical risks associated with live social network experimentation.
Entities (5)
Relation Signals (3)
BotVerse → usesdatafrom → Bluesky
confidence 100% · BotVerse seeds simulations with real-time content from Bluesky
BotVerse → integrates → Stable Diffusion
confidence 95% · BotVerse integrates a Stable Diffusion-based image generation pipeline
Synthetic Social Observatory → ispartof → BotVerse
confidence 95% · The system is organized into four tightly integrated layers: the Synthetic Social Observatory...
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:BotVerse is a scalable, event-driven framework for high-fidelity social simulation using LLM-based agents. It addresses the ethical risks of studying autonomous agents on live networks by isolating interactions within a controlled environment while grounding them in real-time content streams from the Bluesky ecosystem. The system features an asynchronous orchestration API and a simulation engine that emulates human-like temporal patterns and cognitive memory. Through the Synthetic Social Observatory, researchers can deploy customizable personas and observe multimodal interactions at scale. We demonstrate BotVersevia a coordinated disinformation scenario, providing a safe, experimental framework for red-teaming and computational social scientists. A video demonstration of the framework is available at this https URL.
Tags
Links
- Source: https://arxiv.org/abs/2603.29741v1
- Canonical: https://arxiv.org/abs/2603.29741v1
Trouble viewing inline? Open PDF directly →
Full Text
14,411 characters extracted from source content.
Expand or collapse full text
BotVerse: Real-Time Event-Driven Simulation of Social Agents Edoardo Allegrini ∗ Computer Science Dept., Sapienza University of Rome allegrini@di.uniroma1.it Edoardo Di Paolo ∗ Computer Science Dept., Sapienza University of Rome dipaolo@di.uniroma1.it Angelo Spognardi Computer Science Dept., Sapienza University of Rome spognardi@di.uniroma1.it Marinella Petrocchi IIT-CNR & Scuola IMT Alti Studi Lucca marinella.petrocchi@iit.cnr.it ABSTRACT BotVerse is a scalable, event-driven framework for high-fidelity so- cial simulation using LLM-based agents. It addresses the ethical risks of studying autonomous agents on live networks by isolating interactions within a controlled environment while grounding them in real-time content streams from the Bluesky ecosystem. The sys- tem features an asynchronous orchestration API and a simulation engine that emulates human-like temporal patterns and cognitive memory. Through the Synthetic Social Observatory, researchers can deploy customizable personas and observe multimodal interactions at scale. We demonstrate BotVerse via a coordinated disinformation scenario, providing a safe, experimental framework for red-teaming and computational social scientists. A video demonstration of the framework is available at https://youtu.be/eZSzO5Jarqk. KEYWORDS Social Networks, Social Bots, Agentic Systems, Disinformation ACM Reference Format: Edoardo Allegrini ∗ , Edoardo Di Paolo ∗ , Angelo Spognardi, and Marinella Petrocchi. 2026. BotVerse: Real-Time Event-Driven Simulation of Social Agents. In Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2026), Paphos, Cyprus, May 25 – 29, 2026, IFAAMAS, 3 pages. https://doi.org/10.65109/CKXV2098 1 INTRODUCTION Online Social Networks (OSNs) are central to modern communica- tion but also serve as conduits for disinformation, which is often amplified by automated bots [1,11]. The advent of Large Language Models (LLMs), like GPT and DeepSeek [5,14], has made these bots sophisticated social agents. They can produce content that is almost impossible to tell apart from human writing [3,13]. While it is important to understand the impact of these agents on public discourse, studying them on live OSNs carries significant ethical risks. The deployment of autonomous agents, particularly those simulating malevolent behaviour, among unsuspecting users gives rise to significant ethical concerns pertaining to deception and non- consensual human-bot interaction. This underscores the necessity for the establishment of controlled simulation environments [7], * These authors contributed equally to this work. This work is licensed under a Creative Commons Attribution Inter- national 4.0 License. Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2026), C. Amato, L. Dennis, V. Mascardi, J. Thangarajah (eds.), May 25 – 29, 2026, Paphos, Cyprus.© 2026 International Foundation for Autonomous Agents and Multiagent Systems (w.ifaamas.org). https://doi.org/10.65109/CKXV2098 wherein researchers can observe agent behaviour on a large scale without endangering actual users. Although recent studies have proposed agent-based platforms to investigate polarization and bot detection [4,6,8], they fre- quently rely on static datasets or unscalable iterative models. To address these limitations, we introduce BotVerse, a scalable, event- driven framework for controlled social simulation. Unlike static approaches, BotVerse seeds simulations with real-time content from Bluesky, but only allows interactions between agents in a separate environment. This design facilitates the secure analysis of synthetic content propagation, thereby mitigating potential ethical concerns. In this demonstration, we showcase three core capabilities: •Real-Time Fidelity: Agents react dynamically to live events that are ingested from real-world data streams. •Realistic Dynamics: Interactions (posts, likes, replies, re- posts) follow human-like temporal distributions. •Scalability: The event-driven architecture supports thou- sands of concurrent and highly customizable agents. 2 BOTVERSE ARCHITECTURE The BotVerse architecture, Figure 1, is a scalable, event-driven, multi-agent system (MAS) designed to support large-scale, socially grounded simulations. The system is organized into four tightly integrated layers: the Synthetic Social Observatory, the Orchestration API, the Factory, and the BotVerse Simulation Engine. The Synthetic Social Observatory ( 1 ) is a React/TypeScript frontend, to observe the simulation in real-time. It enables re- searchers to inspect the evolving “synthetic public square”, visualize interaction graphs, and analyze individual agent profiles, support- ing both qualitative exploration and quantitative analysis. The Orchestration API (2), implemented using FastAPI, acts as the system’s coordination backbone. It enables asynchronous communication between the simulation engine (4) and the fron- tend (1). This layer abstracts content generation and reasoning, supporting modular, plug-and-play integration of heterogeneous LLM backends (e.g., GPT-oss, DeepSeek). Additionally, it exposes real-time simulation state through RESTful endpoints, facilitating live monitoring and interaction. The Factory (3) is a PostgreSQL-backed persistence layer, re- sponsible for managing the state of thousands of agents and their interactions. The Factory’s approach enforces the isolation of data access and state management from execution logic, providing con- sistency and fault tolerance across concurrent simulation threads, and enabling scalable and reproducible experimentation. The BotVerse Simulation Engine (4) is the architecture’s core and governs autonomous agent behavior. In contrast to traditional arXiv:2603.29741v1 [cs.SI] 31 Mar 2026 13 Reply Comment Agent 1 LLM Core Post Agent 2 LLM Core Agent N LLM Core Agent 3 LLM Core Memory & PersonaMemory & Persona Memory & Persona Memory & Persona Like Repost Post Bluesky 4 Image Generator 2 Figure 1: Overview of the BotVerse architecture. Cognitive Modeling: The Extensible Agent Persona User Data & Details (e.g., username, age, gender, country, education) Psychographic Traits (e.g., political view, religion) Modular Add-ons Extensible (e.g., disinformer) Persona Builder Function (code logic) Agent Persona Prompt e.g., Figure 2: Persona prompt example. iterative simulations, BotVerse adopts an event-driven execution model, where agent actions are triggered by environmental stimuli or internally scheduled events. Environmental Grounding. To ensure high behavioral fidelity, agents do not operate in isolation. Instead, they perform contextual ingestion by sampling real-time content streams from the Bluesky (AtProto) ecosystem. This grounding mechanism enables the MAS to respond to real-world events, effectively blending external reality with internally generated synthetic discourse. Action Logic and Digital DNA. Agent behaviors are modeled with a custom version of the Digital DNA paradigm [2]. Each agent follows behavioral sequences (e.g., Post→Wait→Reply), governed by temporal distributions that emulate human OSN usage patterns, reproducing realistic activity burstiness and circadian rhythms. Cognitive Modeling: Memory and Attributes. Each agent has a Dynamic Memory Module and an Extensible Persona Pro- file to characterize its behavior via context-rich runtime prompts. Memory management follows a heuristic scoring mechanism in- spired by [9], rather than a simple FIFO buffer. Specifically, memory is computed as푆= 훼 · recency+ 훽 · importance,whererecencyis a time-based weighting factor that decreases exponentially as a post gets older (prioritizing newer content in memory selection) andimportanceis approximated via social resonance signals (e.g., count of likes and reposts received). This allows agents to remain selectively attentive to salient or viral content. Personas are repre- sented as high-dimensional JSON-based profiles encoding both de- mographic (e.g., age, education) and psychographic attributes (e.g., political and religious orientation). The persona model is inherently extensible, with the capacity to add new behavioural and personal- ity traits (e.g., the propensity to share disinformation) declaratively and dynamically injecting them into the LLM prompt at runtime. A pictorial example of prompt is in Figure 2. Multimodal Content Synthesis. To support multimodal social interactions, BotVerse integrates a Stable Diffusion-based image gen- eration pipeline [10]. When an agent’s internal decision logic trig- gers an image-based post, the LLM produces a semantically aligned textual prompt, which is subsequently rendered into a synthetic image, ensuring coherence between visual and textual modalities. 3 DEMONSTRATION AND USE CASES We realized a Disinformation Scenario to study disinformation spread with coordinating agents [12]. We initialized푁=500 agents in Factory: 350 benign agents (disinformation-skeptical) and 150 disinformative agents. The simulation unfolds in three phases: •Phase A - Seeding: Malicious (disinformative) agents use con- textual ingestion of Bluesky trends to craft deceptive narratives aligned with current real-world discourse. •Phase B - Amplification: Malicious agents use very careful plans to make their attacks bigger. Aside from broadcast posting and getting more users to engage with it (for example, by liking or reposting false information), they argue with people in reply threads. They use long, persuasive reasoning to challenge the scepticism of genuine users. •Phase C - Multi-level Analysis: The Observatory (1) enables inspection of both micro-level agent trajectories (cognitive states) and macro-level network dynamics (narrative diffusion). Broad Applicability. The BotVerse architecture may support di- verse research and industrial use cases, touching topics such as: •Computational Social Science: Study norm emergence via memory salience parameters훼(recency) and훽(importance) to model how persistence and resonance shape collective attention. •AI Safety & Red-Teaming: Researchers can use the Orches- tration API (2) to stress-test bots in charge of analysing and mitigating LLM-driven disinformative behaviors. •Policy & Crisis Response: Researchers can simulate grounded “What-If” scenarios (e.g., natural disasters) by injecting crafted narratives into real-time content streams. •Market Dynamics: Researchers can leverage Multimodal Syn- thesis (in 4 ) to evaluate brand resonance and sentiment across psychographic segments in a risk-free environment. 4 CONCLUSION AND FUTURE WORK BotVerse 1 provides a safe and scalable sandbox for studying au- tonomous social agents without the ethical risks of live-network experimentation. By integrating real-world Bluesky data with an event-driven simulation engine, BotVerse enables high-fidelity mod- eling of complex social phenomena, including disinformation spread. The modular architecture supports customizable personas, realistic temporal dynamics, and multimodal content generation. BotVerse bridges the gap between static datasets and live OSN research, offering a versatile platform for, e.g., AI safety, red-teaming, and computational social science. Future work will extend agent behav- ioral complexity and support cross-platform simulations. ACKNOWLEDGMENTS This work is supported by project SERICS (PE00000014) under the NRRP MUR program funded by the EU - NGEU. 1 Framework code available at https://github.com/netsecuritylab/BotVerse. REFERENCES [1]Alessandro Bessi and Emilio Ferrara. 2016. Social bots distort the 2016 US Presidential election online discussion. First Monday 21, 11-7 (2016). [2]Stefano Cresci, Roberto Di Pietro, Marinella Petrocchi, Angelo Spognardi, and Maurizio Tesconi. 2017. The paradigm-shift of social spambots: Evidence, theories, and tools for the arms race. In 26th World Wide Web Companion. 963–972. [3] Edoardo Di Paolo, Marinella Petrocchi, and Angelo Spognardi. 2025. Detection of LLM-powered bots using image classification. First Monday (2025). [4]Antonino Ferraro, Antonio Galli, Valerio La Gatta, Marco Postiglione, et al.2024. Agent-based modelling meets generative AI in social network simulations. In Advances in Social Networks Analysis and Mining. Springer, 155–170. [5] D. Guo, D. Yang, H. Zhang, et al.2025. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning. Nature 645 (2025), 633–638. [6]Maik Larooij and Petter Törnberg. 2025. Can We Fix Social Media? Testing Prosocial Interventions using Generative Social Simulation. arXiv preprint arXiv:2508.03385 (2025). [7] Yikang Lu, Alberto Aleta, Chunpeng Du, Lei Shi, and Yamir Moreno. 2024. LLMs and generative agent-based models for complex systems research. Physics of Life Reviews (2024). [8]Lynnette Hui Xian Ng and Kathleen M. Carley. 2025. Are LLM-Powered Social Me- dia Bots Realistic?. In Social, Cultural, and Behavioral Modeling: 18th International Conference, SBP-BRiMS 2025. Springer-Verlag, 14–23. https://doi.org/10.1007/978- 3-032-07715-8_2 [9]Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. Generative Agents: Interactive Simulacra of Human Behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. [10] Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 10684–10695. [11]Chengcheng Shao, Giovanni Luca Ciampaglia, Onur Varol, Kai-Cheng Yang, Alessandro Flammini, and Filippo Menczer. 2018. The spread of low-credibility content by social bots. Nature Communications 9, 1 (2018), 4787. [12]veraAI. 2024.Report on “Coordinated inauthentic behaviour detection tree”.https://w.veraai.eu/posts/report-revisit-coordinated-inauthentic- behaviour-detection-tree. [13]Kai-Cheng Yang and Filippo Menczer. 2024. Anatomy of an AI-powered malicious social botnet. Journal of Quantitative Description: Digital Media 4 (2024). [14]Junjie Ye, Xuanting Chen, Nuo Xu, Can Zu, Zekai Shao, Shichun Liu, Yuhan Cui, Zeyang Zhou, Chao Gong, Yang Shen, et al.2023. A comprehensive capability analysis of GPT-3 and GPT-3.5 series models. arXiv preprint arXiv:2303.10420 (2023).