Evidence-centred AI
Source grounding, measurable uncertainty, citations, and abstention when evidence is insufficient.
Planned output forms: protocols, benchmarks, reports, replications, and negative results.
Research programme
OSForYou research on evidence-centred AI, memory, safe agentic systems, and the societal and environmental impact of AI.
Programme thesis
Source grounding, measurable uncertainty, citations, and abstention when evidence is insufficient.
Planned output forms: protocols, benchmarks, reports, replications, and negative results.
Memory with provenance, controlled forgetting, and a clear boundary between fact and hypothesis.
Planned output forms: protocols, benchmarks, reports, replications, and negative results.
Autonomy constrained by policy, human approval, auditability, and fail-closed controls.
Planned output forms: protocols, benchmarks, reports, replications, and negative results.
AI’s impact on law, work, and institutions, with European control over data and infrastructure.
Planned output forms: protocols, benchmarks, reports, replications, and negative results.
Responsible AI for energy transition, emissions reduction, and sustainable development.
Planned output forms: protocols, benchmarks, reports, replications, and negative results.
Method
Every future study will describe its question, sources, measurement date, procedure, limitations, funding, and conflicts of interest. Lack of conclusive evidence remains a result — not a reason to hide it.
The publication section will launch once the first materials meeting these standards are ready.
Research cycle
Not every piece of work must complete the full cycle. Each should clearly state where it stands.
A claim that can be challenged.
Conditions, data, and evaluation criteria.
Results with errors and limitations.
Review and attempted falsification.
A path to independent repetition.