VIRENA: Virtual Arena for Research, Education, and Democratic Innovation

Digital platforms shape how people communicate, deliberate, and form opinions. Studying these dynamics has become increasingly difficult due to restricted data access, ethical constraints on real-world experiments, and limitations of existing research tools. VIRENA (Virtual Arena) is a platform that enables controlled experimentation in realistic social media environments. Multiple participants interact simultaneously in realistic replicas of feed-based platforms (Instagram, Facebook, Reddit) and messaging apps (WhatsApp, Messenger). Large language model-powered AI agents participate alongside humans with configurable personas and realistic behavior. Researchers can manipulate content moderation approaches, pre-schedule stimulus content, and run experiments across conditions through a visual interface requiring no programming skills. VIRENA makes possible research designs that were previously impractical: studying human–AI interaction in realistic social contexts, experimentally comparing moderation interventions, and observing group deliberation as it unfolds. Built on open-source technologies that ensure data remain under institutional control and comply with data protection requirements, VIRENA is currently in use at the University of Zurich and available for pilot collaborations. Designed for researchers, educators, and public organizations alike, VIRENA’s no-code interface makes controlled social media simulation accessible across disciplines and sectors. This paper documents its design, architecture, and capabilities.

💡 Research Summary

The paper presents VIRENA (Virtual Arena), a novel research‑grade platform designed to overcome the methodological bottlenecks that have long hampered the systematic study of digital public discourse. Traditional approaches suffer from restricted access to real‑world data, ethical constraints on field experiments, and the need for substantial programming expertise to configure existing simulators. VIRENA integrates the strengths of several fragmented tool families—single‑user simulations, independent feed‑based platforms, chat‑based dialogue labs, large‑scale agent‑only models, and educational games—while eliminating their individual shortcomings.

Key contributions include:

1. Realistic Multi‑Format Replication – VIRENA faithfully reproduces the visual design and interaction patterns of five mainstream platforms (Instagram, Facebook, Reddit, WhatsApp, and Messenger). Users can scroll feeds, view images and videos, like, comment, up‑vote, and share, thereby preserving ecological validity.

2. Live Multi‑User Interaction – Multiple human participants can join the same simulated environment in real time, posting, commenting, and reacting to one another. This enables the study of group dynamics, collective deliberation, and social influence that single‑participant simulators cannot capture.

3. Configurable LLM‑Powered AI Agents – Researchers can add AI agents that behave either as “human‑like” or as overt “bots.” Agents are linked to external large language models via API, and their personas are defined through system prompts. Two dimensions of human‑likeness are controllable: response latency (including typing‑style delays) and linguistic style (prompt‑driven). This makes VIRENA uniquely suited for human‑AI interaction experiments.

4. Research‑Centric Moderation Engine – The platform treats content moderation as an experimental variable. Detection methods (keyword matching, regular expressions, AI‑based classification) can be combined with response actions (flagging, deletion, warning pop‑ups). Moderation rules can be scoped to specific content sources (human, AI, or both) and can include escalation thresholds. This allows direct comparison of policy interventions on user behavior and perception.

5. Precise Content Scheduling – Researchers can pre‑schedule posts, comments, or messages to appear at exact timestamps, ensuring tight control over exposure timing—a critical requirement for clean causal inference.

6. No‑Code Visual Configuration – All experiment parameters—platform type, participant limits, templates, moderation rules, AI agent settings, and visual customizations—are defined through a drag‑and‑drop interface. No programming is required, dramatically lowering the barrier for social scientists, educators, and public‑sector analysts.

7. Integration with External Survey and Recruitment Tools – VIRENA can be linked to Qualtrics, QuestionPro, Prolific, etc., facilitating participant onboarding, consent management, and post‑experiment data collection.

Technical architecture: VIRENA is a web‑based application built on PocketBase (a lightweight Go framework) with SQLite for storage, enabling a single‑binary deployment on a university server. All data—including user actions, AI‑generated content, moderation flags, and timestamps—are tagged with source metadata (human, bot, moderator, script) to simplify downstream analysis. The AI agent subsystem stores model endpoint configurations and supports multiple LLM providers, while human‑likeness controls are implemented via configurable delay generators and prompt templates.

The paper situates VIRENA within the existing literature, providing a comparative matrix that shows VIRENA uniquely combines realistic feed and chat interfaces, live multi‑user interaction, configurable moderation, and a no‑code workflow. It also outlines current limitations: the absence of proprietary recommendation algorithms (e.g., feed ranking), dependence on external LLM APIs (cost and latency), limited platform catalog (currently five), and the need for external ethical review and consent procedures.

Use cases described include: (a) measuring how different moderation strategies affect discourse quality and user trust; (b) investigating how AI agents influence opinion formation and misinformation spread in mixed human‑AI groups; (c) studying cross‑platform contagion by moving a discussion from a feed‑based environment to a private chat; and (d) pedagogical simulations for media‑literacy training.

VIRENA is already deployed in pilot studies at the University of Zurich and is offered for collaborative projects. The authors emphasize that the platform’s open‑source foundation ensures transparency, auditability, and institutional data sovereignty, aligning with Swiss data‑protection standards. Future work will expand platform coverage (e.g., adding Twitter, TikTok), integrate algorithmic feed personalization, and develop richer analytics dashboards.

In summary, VIRENA fills a critical methodological gap by providing a controllable, realistic, and accessible sandbox for social‑media research, education, and democratic innovation, opening new avenues for rigorous empirical investigation of online public spheres.