The Problem with Proxy Metrics in AI Governance

by | Jun 5, 2024 | AI Ethics

Every AI deployment makes implicit ethical choices. The training data encodes values. The objective function prioritizes outcomes. The deployment context determines who is affected. Pretending these choices are purely technical is itself an ethical position, and not a...

Transparency Reports for AI: What They Should Contain

by | May 31, 2024 | AI Ethics

Every AI deployment makes implicit ethical choices. The training data encodes values. The objective function prioritizes outcomes. The deployment context determines who is affected. Pretending these choices are purely technical is itself an ethical position, and not a...

The Precautionary Principle Applied to AI Deployment

by | May 22, 2024 | AI Ethics

AI ethics is not a constraint on innovation. It is a quality standard for innovation. An AI system that produces biased outcomes, violates privacy, or makes unexplainable decisions is not a good system that happens to be unethical. It is a bad system. The fundamental...

Epistemic Justice and AI: Whose Knowledge Counts?

by | May 8, 2024 | AI Ethics

AI ethics is not a constraint on innovation. It is a quality standard for innovation. An AI system that produces biased outcomes, violates privacy, or makes unexplainable decisions is not a good system that happens to be unethical. It is a bad system. When we talk...

Cultural Bias in AI: The Western Default Problem

by | Apr 17, 2024 | AI Ethics

Every AI deployment makes implicit ethical choices. The training data encodes values. The objective function prioritizes outcomes. The deployment context determines who is affected. Pretending these choices are purely technical is itself an ethical position, and not a...