Research & Methods
This section is separate from public articles. It contains methodologies, working papers, DOI records, and formal research outputs.
A human-led method for moving from sources and observations to bounded findings in investigation and assurance. It defines how evidence is assessed, competing explanations are tested, and limitations are reported.
Read the publication View DOI recordExamines what must be established before an AI-system artifact can support a finding of harm, including situations in which the system operated as designed.
Read the working paperProvides structured tests for assessing how far the available record supports attributing a consequential AI-agent action to a natural person.
Read the methodology