Beware the power of prediction | Carissa Véliz

Episode Summary

Executive Summary: Carissa Vélez argues that many predictions, especially in AI, are not neutral forecasts but power-laden speech acts that shape behavior, justify decisions, and can create self-fulfilling prophecies. She warns that predictive systems can reinforce inequality, hide injustice, and enable manipulation unless society debates when prediction is appropriate and when transparent, contestable criteria are needed instead.

Main Topics: Prediction as power, not just knowledge (Priority: 5/5): Vélez reframes prediction as a tool often used to influence outcomes and legitimize choices, rather than merely describe the future. Historical framing of prophets and rulers (Priority: 4/5): She opens with the Louis XI astrologer story to show how predictions have long been tied to political power and survival. AI and social self-fulfilling prophecies (Priority: 5/5): The talk distinguishes between harmless predictions about physical systems and dangerous predictions about people, which can alter behavior and outcomes. Predictions as speech acts and veiled commands (Priority: 5/5): Social predictions function like instructions that push people toward the predicted outcome, especially when framed as inevitable. Manipulation through markets and tech hype (Priority: 4/5): Prediction markets and tech forecasts can be weaponized by those with resources to steer public perception and investment. Hidden injustice and lack of contestability (Priority: 5/5): Predictive decisions in lending, insurance, and other domains can be impossible to challenge, obscuring unfairness behind statistical authority. Defending uncertainty and democratic agency (Priority: 5/5): She concludes that uncertainty is good news because it leaves the future unwritten and invites public resistance to authoritarian predictive control.

Key Arguments: Most everyday predictions are closer to power plays than objective knowledge; they justify value-laden decisions under the pretense of facts. Predictions about humans differ from predictions about weather or molecules because people react to them, making them self-fulfilling or self-defeating. Social predictions are speech acts: they do something in the world by nudging, commanding, or pressuring people to comply. Claims that a predicted future is inevitable function as conversation stoppers and discourage scrutiny or democratic debate. Prediction markets can be manipulated by wealthy actors or political insiders who bet strategically to shape public perception and outcomes. Algorithmic prediction can create hidden injustice because future-oriented decisions are difficult or impossible to contest like ordinary factual claims. Using individual predictive scores for insurance undermines solidarity by shifting risk pooling toward individualized pricing. Society should distinguish between acceptable uses of prediction and unacceptable ones, especially in fairness-sensitive domains like lending, insurance, and governance. Uncertainty is not a problem to eliminate at all costs; it preserves freedom and the possibility of a better future.

Data Points: Anonymous accounts' earnings from betting: $1.2 million - Vélez cites six anonymous accounts that profited from betting on an attack on Iran. Number of anonymous accounts: 6 - She references six anonymous accounts involved in the betting example. Timing of funding: Hours before - Some of the wallets used in the betting example were funded only hours before. Recommended individual insurance use of predictions: Not for individuals - She argues predictions may be appropriate at a population level but not for individuals in insurance, to avoid unfair self-fulfilling prophecies.

Pivotal Quotes: "Predictions are often power plays in disguise." — Carissa Vélez: Core thesis of the talk, defining predictions as tools of influence rather than neutral knowledge. "Predictions are never facts. Facts belong to the past." — Carissa Vélez: Her argument that predictions are unverifiable and should not be treated as established truths. "Uncertainty is good news. It means that the future is unwritten, that it's ours to write." — Carissa Vélez: Closing call to resist fatalism and defend agency and democracy.

Implications: Listeners should treat predictive claims—especially in AI, insurance, and politics—as contestable acts of influence. The industry needs clearer limits, transparency, and public debate to prevent manipulation and unfairness.