Independent · Nonprofit · Open Research
Human agency in an age of artificial intelligence.
As AI assumes more of the decisions once made by people, we study what becomes of the human capacities on which free societies depend.
KnowGence Research Foundation is an independent, nonprofit research organization. We examine how agency, responsibility, and individual freedom fare as decision-making is increasingly delegated to machines — and we develop rigorous, openly available measures to tell whether AI strengthens or quietly erodes these capacities.
Our Research
Bounded Agency, and the human measures of progress.
We explore the impact and necessity of Bounded Agency in AI systems — whether human judgment, accountability, and the ability to intervene are preserved as consequential decisions are delegated to AI. Alongside it, we develop human progress and resilience metrics that look beyond output alone to the agency, trust, and hope on which societies depend.
Our research, methods, and data are published openly, for anyone to use and build on.
Research Programs
Four lines of work, all published openly.
Bounded Agency
Bounded agency is the state in which AI acts autonomously, but only within limits people set and can enforce in time to matter, so that human agency is preserved. The framework defines the six conditions that state requires: authority boundaries, tempo matching, safe mode and rollback, intervention rights, reauthorization and accountability.
Board AI Governance Index
An annual, independent benchmark of how governable AI is across industries, and how boards oversee it. Directors and risk leaders are invited to contribute.
About the Index →An open evidence standard
What counts as evidence that AI governance actually works, as opposed to existing only on paper. Proposed levels, assigned to each claim and control: E0 self-reported, E1 documented, E2 configured, E3 tested, E4 operationally observed. A self-report can reveal a gap but never prove a control works. To be published for anyone to use.
Human progress and resilience metrics
Measures that look beyond output to the agency, trust and hope on which free societies depend, so progress in AI can be judged by what it does for people.
Publications
Where the work is published.
Bounded Agency: Who Is in Control in the Age of AI?
Ravi S. Bhagavatula, PhD, CFA
How authority migrates into automated systems, and what organizations, boards and institutions must do to keep AI in a state of bounded agency, acting within limits people set and can enforce, so that human agency is preserved.
The foundational Bounded Agency paper is available on SSRN. Read the paper →
Independence
Research that no one can buy.
- No vendor funding. The Foundation takes no money from technology vendors whose products its research could evaluate.
- Open methods. Methods, scoring approaches and sample sizes are published so others can check the work.
- Separate from KnowGence. KnowGence, an AI governance assessment firm, is a separate organization. The two share a founder and may fund or commission work from each other; such arrangements are disclosed, and no funder, KnowGence included, can influence findings or stop publication.
- Consent and privacy. Data reaches our research only anonymized, aggregated and with consent.
Get Involved
Ways to take part.
- Directors and risk leaders: contribute to the first Board AI Governance Index.
- Researchers: collaborate on Bounded Agency, governance measurement or human resilience metrics.
- Supporters: help fund independent research on human agency and AI.
Findings, open datasets and tools will be published here as the work develops. To take part, collaborate or support the research, write to us.