Hamelberg AI

Evidence notes

Read the evidence. Keep the limits attached.

Practical analysis for readers who want to understand what an AI result or a public dashboard can actually establish.

These notes connect the questions behind the Observatory, The Brief and the European Security Monitor. Each follows a specific reasoning problem rather than treating a headline or chart as a conclusion. Worked examples are explicitly hypothetical; they are not findings from our projects.

AI at work: productivity and control

A customer-support study gives a concrete result, but a productivity measure answers only part of the question about human benefit. Follow the distinction between measured output, distribution and control over the work.

Read the evidence note →

When more sources do not mean more events

Use a small worked example to see how duplicate reporting and missing transcripts change the denominator. Learn why a news sample cannot support a country league table or a claim about public opinion.

Work through the counts →

Use the notes with the live projects

The Observatory provides retained reporting and its classifications. The Brief makes source-linked stories easier to read. The European Security Monitor presents sampled security narratives. These are distinct kinds of material; their counts and labels are not interchangeable.

For a reusable way to record a claim, source, comparison and uncertainty, start with the evidence record in the reading guide. For publishing responsibility, AI assistance and how to flag a mistake, see our editorial approach.

Published 14 November 2026. These explanatory notes use AI-assisted drafting and source checking; they do not report new fieldwork or experiments.