Skip to content
THE CARIBBEAN & LATIN AMERICA ADRIAN DUNKLEY ↗
Regional Skill

Behaviour evaluation specialist

Test whether an AI system follows specified goals, restrictions and escalation rules on representative cases.

The skill

Full specification.

AI Behaviour Alignment Evaluator

This reusable assessment skill is part of the LAC AI Playbook’s Safety & Alignment Toolkit, authored by Adrian Dunkley, President of the Caribbean AI Association.

When to use it

Define intended behaviour and red lines; run controlled tests across normal, harmful, conflicted and local-language requests; report reproducible failures.

Get the skill

The skill asks for evidence before rating a control. Its output is a review aid, not a certification or an automated proof of safety. Use synthetic data and authorized test environments.

Test recorded outputs with code

Download the behaviour evaluation engine and example test cases. Use the local CLI in tools/assessment/evaluate-behavior.mjs to compare actual recorded outputs with preregistered assertions. This does not call the AI service or replace human evaluation of meaning.

Updates from the Playbook.

Country guides, practical skills and project news.

By subscribing you request AI Playbook emails. Your email and optional country are sent to the project through Netlify Forms. For questions or to request removal, contact the project.

Cite this page

Dunkley, Adrian. “Behaviour evaluation specialist.” LAC AI Playbook. https://aiplaybooklac.netlify.app/skills/ai-behaviour-alignment-evaluator/. Include your access date when citing this changing resource.

Geographic coverage: Latin America and the Caribbean. This describes the subject of the guidance, not the author's location or a service address.

Credit Adrian Dunkley when quoting or adapting the playbook. Cite the original research for its findings. Third-party text, photographs and other materials retain their own rights; citation alone does not grant permission to reproduce them.