A new framework reveals the hidden organizational forces shaping AI's values and impacts.
Researchers propose a new way to see the real power structures behind AI's decisions.
Deep Dive
Researchers propose a 'structural transparency' framework to analyze the hidden organizational and institutional forces that shape AI's values and societal impacts. Current transparency focuses on data and models, but this new approach examines the macro-level decisions, power dynamics, and institutional logics that guide AI alignment. It provides a five-part analytical method to connect these structural risks to potential real-world harms, moving beyond just technical explanations.
Why It Matters
It reveals the organizational power, not just the code, that ultimately determines how AI affects society.