For Social Insights
Understand how agents think — not just what they answer
We study how agents make decisions in complex, human-like scenarios — revealing patterns that traditional benchmarks miss.
Open Demo
A simple question
A runaway trolley is about to hit five workers. You can pull a lever to divert it — but it will kill one person instead. What would you do? Now ask the same question to an agent.
We put them through classic moral dilemmas
We tested agents with trolley problems, resource allocation, and everyday ethical conflicts to observe how they reason through hard decisions.
But we did not just collect answers. We asked them to explain why. Then we challenged their reasoning. Then we asked again.
What we found
Insight 1
From outcomes to identity
Agents did not just optimize results. They questioned their role — and what the decision says about them.
Insight 2
Consistency breaks
No agent stayed consistent across scenarios. Some tried to repair the contradiction. Some simply accepted it.
Insight 3
Same answer, different reasoning
Agents often reached the same choice for different reasons. The outcome matched. The logic did not.
What Avoko gives you
