Controlled experiments in realistic social media environments
VIRENA (Virtual Arena) is a research platform developed at the University of Zurich. People and AI agents interact in replicas of social media platforms and messaging apps. A study can include only real participants, only AI agents, or any combination of the two, and dozens of participants can take part at the same time. Researchers set up the study in a visual interface, without programming.
It is designed for researchers, and also for educators and public organisations that want to study, teach or practise digital communication in a safe, closed environment.
How it works
- Platforms
- Each experiment copies one platform: Instagram, Facebook, Reddit, X, WhatsApp or Messenger. Participants see the familiar interface and use its usual options, such as likes, reactions, votes, comments and messages.
- Experiments and scenarios
- A researcher creates an experiment and adds scenarios, which are the conditions of the study. A scenario holds scripted posts and replies that appear at set times, the AI agents that take part, and the moderation rules for that condition.
- Participants
- Participants enter through a link, and there are two kinds. With the experiment link, participants wait in a room until a group has gathered, and the group is assigned at random to a scenario. With a scenario link, participants go straight to one scenario. This suits designs in which randomization at the individual level is handled by other software, such as Qualtrics. In both cases a run can hold one participant or several. Participants interact with each other in real time, and with AI agents if the scenario includes them.
- AI agents
- Agents are language-model personas described by a prompt. Researchers decide when and how quickly they respond, and connect their own API key for the model provider they choose. Agents are optional: a scenario can include real participants only, agents only, or both.
- Moderation
- Moderation rules detect content with keywords, patterns or an AI prompt, and then flag it, delete it or show a warning. Because rules belong to a scenario, moderation can differ between conditions.
- Attention
- VIRENA records how long each post or message is visible on the participant's screen (attention time, in milliseconds), and how long a comment section is open on platforms that have one. On Facebook and Instagram it also records whether a video was played. Attention time is based on the viewport, the part of the page that is on screen. It therefore depends on screen size and on scrolling: content that is never scrolled into view receives no attention time, and on a large screen several posts can be visible at once. It records time on screen. It does not record where a person looks or whether they read the content.
- Data
- Every interaction is recorded with a timestamp, together with the attention measures above. Exports state the scenario each participant was in.
- Linking to other tools
- The participation link can carry an external participant ID, for example from Prolific. VIRENA stores the ID and includes it in the export. When a run ends, VIRENA redirects the participant to a web address set by the researcher, such as the next page of a Qualtrics survey. In this way one participant can be followed across tools, for example from Prolific to Qualtrics to VIRENA, back to Qualtrics and back to Prolific, and characteristics collected in a survey can be linked to behaviour in VIRENA.
The preprint describes all of this in detail.
Team
VIRENA is developed at the Department of Political Science, University of Zurich.
Stay updated
VIRENA is currently available through research collaborations, and general release is expected at the end of 2026. Sign up below and we will write to you when it opens. To use it now, email hoes@ipz.uzh.ch about a collaboration.
We use your address only to tell you about VIRENA and will not share it. We use Brevo to store addresses and send emails. You can unsubscribe at any time with the link in every email, or ask us to delete your address. If the form does not load, email hoes@ipz.uzh.ch.
Preprint
The preprint explains the design of VIRENA, what it can do, and how experiments are built and run.
Hoes, E., Klüser, K. J., & Gilardi, F. (2026). VIRENA: Virtual Arena for Research, Education, and Democratic Innovation. arXiv:2602.12207.