Governed autonomy
How can AI systems be given meaningful independence that remains bounded, revocable and auditable, and how should that autonomy expand as trust is earned?
Research in progress
Research
The questions we're working on.
Our research asks how intelligent systems can be given real responsibility while people keep meaningful control, and how technology can strengthen deliberate human choice.
Research areas
Where our attention goes.
Independent verification
How can work be evaluated by parties separate from those who produced it, so that results are judged on evidence rather than self-report?
Authority and accountability
How should permission be expressed, narrowed, delegated and withdrawn across people, organizations and machines, and how do we keep it traceable?
Provider-independent systems
How can governance and institutional knowledge outlast any individual model, vendor or execution environment?
Reaction and deliberate choice
How can technology help people notice their reactions and act on their values, without becoming another source of compulsion?
Human agency at scale
As capable AI becomes widely available, what structures ensure it expands what individuals can do rather than concentrating control?
How we publish
We share work when it's ready for scrutiny.
Our research is ongoing. We publish findings once they have been checked and are ready for outside review, and we would rather say less than overstate what we know.
Publications, technical notes and talks will be listed here as they become available. Until then, this page describes the questions we're pursuing rather than conclusions.
Researchers working on related problems are welcome to get in touch.
Publications
Forthcoming.
No publications are available yet. Check back, or ask to be notified when the first ones are released.