AI Ethics and Responsible Use: Advanced

Twenty-two pages for after the beginner tutorial. You learn to scope an impact assessment for one bounded AI feature, identify stakeholders and harms, score risks, choose controls and record a decision that someone is accountable for. You then map principles, frameworks, standards and regulation to testable requirements, without assuming any one country's rules apply.

Further pages cover fairness measurement (selection rates, error rates, small samples and conflicting metrics), data governance, data rights and records, model cards, system cards and datasheets, and the design of human oversight that is more than a rubber stamp. Then come high-stakes use cases and stop-use rules, misuse and safety analysis, synthetic media, open versus closed models, labour and environment trade-offs, governance roles, vendor questions, audit and change approval, and how to measure whether controls work and run an incident review.

The Python in this tutorial uses only the standard library, runs offline and works on small invented toy datasets. Results illustrate the calculation only and cannot establish anything about the real world. Contested questions are presented as arguments from several sides. A practice lab and a capstone, a go/no-go assessment pack, finish the tutorial. It is educational and is not legal advice. Every page is available in English and Bangla.