Lesson 9 / 26

Core Ethical Principles

Name the common principles and turn each into a concrete question a team can ask.

A shared vocabulary

Many organisations and bodies (for example the OECD AI Principles and UNESCO's recommendation on AI ethics) converge on similar ideas: fairness and non-discrimination, transparency and explainability, accountability, privacy and data protection, safety and robustness, human oversight, and benefit to people and society. Principles are abstract, so each team should translate them into questions it asks before launch.

Principles into practice

Ethical principles only matter when they become questions, checks and decisions in real projects.

Four anchors: fairness, transparency, accountability, privacy.
Figure 3.1 — Fairness, transparency, accountability and privacy.

Principles as questions

Put these in the project template so they are asked every time, by default.

Fairness        Does it work equally well for all groups of users we serve?
Transparency    Will users know AI is involved, and can we explain the outcome?
Accountability  Who is answerable if it harms someone? Who can fix it?
Privacy         Do we use only the personal data we need, with a lawful basis?
Safety          What is the worst realistic failure, and how do we contain it?
Human oversight Which decisions can a person review and override?
Benefit         Who gains, and could anyone be worse off?

Ethics is not only compliance

Something can be legal and still unfair or unwise. Use the law as the floor, and the questions above to check whether the design is something you could explain openly to the people it affects.

Quick check: How should a team use ethical principles?

  • Ignore them if the law allows
  • Print them on a poster and move on
  • Turn them into concrete questions asked before launch
  • Leave them to the vendor
Answer

Turn them into concrete questions asked before launch — Questions in the process make principles actionable and checkable.