Carissa Véliz

Predictions are often power plays in disguise.

They justify value-laden decisions under the pretense of facts.

Better understanding prediction matters more than ever, because we're relying on forecasting more than ever with AI.

And based on how we talk about prediction, we're being much too naive about it.

But AI is science, you might think.

It's cutting-edge technology, right?

Well, it depends on the kind of AI and how we use it.

AI can be a great technology to make predictions about molecules in the search for new antibiotics.

But predictions about human beings are fundamentally different than those about things.

Predictions about the weather don't influence the weather.

Predictions about people influence people.

Social predictions tend to act like magnets.

They bend reality towards themselves.

They affect the reality they purport to predict.

An algorithmic prediction about future disease can make someone's insurance premiums go up, leading to worse health outcomes from stress alone.

Predictions sound like descriptions of the world, like facts, but they're not.

Analyzed closely as assertions, they are what philosophers called speech acts.

That is, language that does something other than describe the world.

When you tell a child to clean up their room,

You're not describing the state of the room, you're issuing an order.