Simulation
We let electorates debate. Who reaches consensus and who does not?
Our simulation lets groups of voters debate freely, with real arguments. It draws on voice-based interviews with around 1,200 people in Berlin, as of August 2026.
That everyone in Berlin can live well and safely.
Adopted: matches statements from the other campThe data
Around 1,800 statements from voice-based interviews.
The statements come from several surveys among around 1,200 people from all major electorates, as of August 2026. They were collected in voice-based interviews, not in focus groups with social pressure.
Every statement sits coded in the system: which concern it expresses, how typical it is for the group, and in what tone it was spoken.
„I was homeless. And through social workers I was able to find a flat again.“
„except that petrol prices are so high and the state just wants to cash in“
„The biggest problem in Berlin is definitely crime, and also that the streets are simply very dirty.“
„The cut to state subsidies for heat pumps annoyed me a lot.“
The mechanism
The process.
You pick two groups and a topic you would like to see discussed. The groups debate their party preference and show what matters to them.
The outcome
After about thirty seconds the outcome of the discussion emerges.
It is not about who wins. We show:
- 01
Where the sides find consensus
- 02
Where they cannot agree
- 03
What matters to each of them
Because democracy means discourse, and consensus does not always have to be found.
Why it holds
A simulation built on real opinions.
The adoptions are traceable
The system does not judge whether an argument sounds convincing. It checks which statements from the interviews come closest to the point being made, and marks the hits. Every adoption traces back to a concrete recording.
The interjections come from the pool
While one side speaks, interjections from the listening group appear at the edge. Short sentences of the kind you know from everyday life. Every debate pulls them fresh from the interviews.
Failure is part of the design
In some constellations there is no convergence. The simulation shows exactly that and explains why. A result of zero is not an error here, it is the finding.
Data protection: participant statements are not shown verbatim. What appears in the simulation is a linguistically smoothed version that keeps the substance and the tone but allows no conclusions about individual people.
Who it is for
A tool against the easy explanation.
The module is for everyone who wants to understand why people vote the way they vote. And above all: under what conditions they would change their mind.