How AJLUN works
A research workflow that connects open-source evidence with human judgment.
- Open-source information
- Extraction and organisation
- Graph memory
- Evidence retrieval
- Analyst review
- Briefs and scenarios
Information and shared memory
Open-source information is collected and organised. AI assists with extracting information from text; routine collection and storage do not require a model at every step. Neo4j retains relationships between events, source records, extracted claims and actors so earlier research can be revisited.
Evidence retrieval and drafting
AI helps analysts find relevant sources and prepare drafts of explanations, briefs and alternative scenarios. Source links allow an analyst to inspect the evidence behind an assessment. A source link is not itself verification, and graph connections are not proof of causation.
Analyst review
People assess sources, compare competing explanations, identify unsupported claims and decide whether an output is suitable for publication. Analytical examples and human-review records inform the next round of development.
What we are developing next
Our development goals include analysing patterns over time, testing possible cause-and-effect explanations against evidence, and reusing analyst feedback in future research. Recording feedback does not automatically train a model. These capabilities require further development and evaluation.