2.1.5. Transparency
💡 First Principle: Transparency means people can understand how an AI system works and, crucially, when it is being used and what its limitations are. Because a model's reasoning is often opaque, transparency is the deliberate effort to make its behavior and boundaries understandable to the people it affects.
This shows up as disclosing that an AI (not a human) is making or assisting a decision, explaining in understandable terms what a system does and doesn't do, and being honest about its limitations and confidence. A loan applicant denied by a model deserves to know an AI was involved and broadly why. Transparency builds the user trust that lets AI be adopted responsibly, and it's a prerequisite for accountability — you can't hold anyone answerable for behavior nobody can see.
⚠️ Exam Trap: Transparency is about understandability and disclosure, not about open-sourcing code or revealing model weights. A scenario where users don't realize they're interacting with AI, or can't get any explanation for a decision, is a transparency failure regardless of whether the code is public.
Reflection Question: Why is transparency considered a prerequisite for accountability? What breaks if a system's behavior is completely opaque?
