Four scenarios. Three minutes. Each one puts you in a situation where AI helped you do something. Sometimes it goes well. Sometimes it doesn't. We measure the gap between what you claim and what you accept.
No sign-up. No tracking. Your responses stay in your browser. When you're done, you can export your data as JSON if you want to contribute to the research.
16 arguments from court rulings, legal scholarship, and philosophy. 14 attack relations. Computed extensions. Here's the full map.
Seven arguments about who owns AI output — and why.
Six arguments about who's responsible when AI causes harm.
Three arguments that connect ownership to accountability. This is where it gets interesting.
When one argument logically defeats another. Each has a specific reason.
These are the arguments that no rational counterargument can defeat. The minimum consensus.
Consistent Attribution. You can't claim ownership of beneficial AI output and disclaim accountability for harmful AI output. They're logically bound.
Unified Control Threshold. The threshold for "I own this" and "I'm responsible for this" must be the same threshold.
Proportional Accountability. Responsibility scales with the degree of human direction and foreseeable risk.
Traceability. If you can't trace AI output to human decisions, you can neither own it nor escape accountability for it.
Non-Waivable Accountability. Terms of service can't shift foreseeable harm to people who never agreed to them.
A research lab studying what happens when humans invoke faith at strategic decision points found a perfect test case in AI governance.
A court in Beijing says prompting AI 150 times earns you copyright. A US examiner says 624 times doesn't. The EU regulates who's responsible when AI causes harm but defers to separate law for who owns the good stuff. Patent offices have reversed their guidance twice in two years.
Every one of these institutions treats ownership and accountability as separate problems. We used Dung's Abstract Argumentation Frameworks — the mathematics of rational disagreement — to prove they can't be separated. The threshold must be the same.
The formal analysis tells us what's logically true. This experiment tells us what humans actually do. Together, they measure the gap between what we should do and what we actually do when the stakes are real and the outcome is uncertain.
The Ownership-Accountability Asymmetry Index (OAAI) measures how much more ownership people claim versus accountability they accept for the same level of AI involvement. An OAAI of 0 means perfect consistency. Anything above 0 means humans are doing exactly what the formal framework says is logically incoherent.
We haven't fabricated results. We're collecting them — one participant at a time. Every response submitted through this site contributes to the dataset.
Devudaaa Research Lab studies the moment a person shifts from calculation to conviction — when the algorithm says one thing and your gut says another. The ownership-accountability paradox is one piece of that larger question.