Academic skills course: Integrating Quantitative Data into Qualitative Research
A three-day CENTREPEACE academic skills course at PRIO in Oslo covered major datasets, spatial and survey data, and the craft of academic publishing.
Arguing with a Disinformer: ERIS at European Researchers’ Night 2026
Can you hold your ground against an AI whose only job is to talk you into something? At this year’s European Researchers’ Night at Masaryk University, visitors to our stand found out. Then the AI showed them exactly how it had worked on them.
This year’s European Researchers’ Night took place across the Czech Republic on September 25 under the theme of communication. At Masaryk University’s Bohunice campus, our stand, Disinformation in the age of AI and small language models, took that theme somewhere less comfortable: what happens when communication is built to mislead?
The starting point was simple. Disinformation rarely works because people are gullible. It works because it speaks to fear, uncertainty, a sense of injustice and distrust, feelings everyone has at some point.
To show how this plays out, CENTREPEACE presented ERIS, the Epistemic Resilience and Influence Simulator. Visitors scanned a QR code, opened a chat on Telegram, and were drawn into a story set in an invented world of fictional states, institutions and documents. For the next few minutes, ERIS played a seasoned disinformation operator. It framed, nudged and leaned on selective context and misleading interpretation, adjusting its tactics to how each person pushed back.
At the end of conversation, ERIS dropped the act and walked each participant through a debrief. It showed which techniques it had used and when, and what it had inferred about them along the way. These were rough guesses, for example about how they handled sources or whether they seemed at home with legal and security terminology. It was explicitly barred from guessing anything sensitive, such as political views, religion, health or sexuality.
Two lessons sit at the core of the exercise. First, individual pieces of information can each be true while their selection and arrangement produce a conclusion none of them supports. Second, an AI can sound strikingly sure of who you are while doing nothing more than guessing.
The conversations at our stand picked up from there. Did anything surprise you? Was there a moment when the argument started to feel convincing? What did you do when you weren’t sure? Did its read on you fit? These questions were less about the machine than about how people reason under pressure.
Checking whether a claim is true is one skill. Noticing how a conversation is steering you is another, and it is much harder to practice. ERIS is an attempt to make that second skill visible.
