Matt Lane · Notes← All notes
2026-11-19 · AI & future work

The Three Tests of Artificial Minds

As artificial intelligence becomes more capable, we need to distinguish what it can create, what it can decide, and what we might owe it.

These are three separate questions. An answer to one does not settle the others.

I. Creativity — Can it change what is conceivable?

Can it transform a conceptual framework to discover genuinely new, valuable possibilities?

II. Agency — Can it determine its own course?

Can it form, pursue, and revise purposes through reasoned judgment over time?

III. Moral Status — Can things matter for its own sake?

Can things go better or worse for it, independently of its usefulness to others?

These are independent questions, not steps toward personhood. Creativity does not establish agency, and neither establishes moral status. An entity need not be creative or autonomous to deserve moral consideration.

Moral status is not accountability. An infant can deserve moral consideration without being responsible for its actions. A corporation can bear legal responsibilities without having experiences or welfare of its own. Whether an AI can be held responsible, and who answers when it cannot, is a separate question.

Answering convincingly is not enough. Each question requires evidence; whether a system has morally significant interests is especially difficult to establish.

The unit of assessment matters: a model, a running instance, or a persistent system with memory and tools. A model can adapt within a conversation without changing its parameters. Lasting memory, evolving commitments, and continuity across sessions depend on the system's architecture. Each test must specify what is being assessed.

Rights do not follow automatically from passing a test. They require a further argument connecting moral standing to specific protections.