In early December, Time magazine named the “Architects of AI” its 2025 Person of the Year.
The lead-up to this announcement has always generated considerable speculation, but this year millions of dollars were riding on it: people placed $55 million in bets on the gambling website Polymarket and $19 million on Kalshi, a rival platform.
But while the platforms allowed users to bet on “Elon Musk,” “Jensen Huang,” and even “AI” as Time’s Person of the Year, no one anticipated the magazine’s exact phrasing.
The announcement ignited a furor. Which bets, if any, would pay out? The answer illustrates an important lesson about how prediction markets function.
In recent years, online prediction markets have flourished, bolstered by regulatory leniency in the United States.
On platforms like Polymarket and Kalshi, users bet on the outcomes of discrete political, social, and economic events.
The resulting “event contracts” vary considerably in the specificity of the conditions under which a bet is considered a win.
These markets have soared to prominence on a promise of superior information discovery and provision.
If motivated actors are willing to place bets on future outcomes within a sufficiently liquid market, the argument goes, the size of those bets should reflect the odds of a particular outcome. But what happens when the outcome itself is unclear?
Social facts
It should come as no surprise that the bets on these platforms are sometimes contested.
Many, if not most, of the events being wagered on are “social facts”: they come into being through social action and are subject to multiple interpretations.
Even contracts that specify payout conditions in great detail can be stymied by the actual outcome.
Time’s 2025 Person of the Year is a case in point. People who wagered on industry leaders like Anthropic CEO Dario Amodei claimed to have won, as did those who put money on AI itself.
Both Kalshi and Polymarket determined that the latter did not win, but disagreed about whether the former did.
Despite claims of automated information discovery, prediction platforms continue to depend on human judgment to structure their markets
As a result, if you placed money on Nvidia CEO Jensen Huang being named Person of the Year on Kalshi, you made money; but you lost money if you placed the same bet on Polymarket.
By reaching diametrically opposed conclusions, these platforms underscored the role that human interpretation plays in settling event contracts, even in increasingly automated digital markets.
It recalls “the paradox of automation’s last mile,” the phrase that anthropologist Mary L. Gray and computational social scientist Siddharth Suri coined in 2019 to refer to the new problems created by advances in machine learning, many of which only human labor can solve.
Despite claims of automated information discovery, prediction platforms continue to depend on human judgment to structure their markets.
Event contract
Polymarket and Kalshi have so far dealt with the need for contractual interpretation differently.
To settle controversial wagers, Polymarket uses a blockchain voting system, wherein holders of Universal Market Access tokens determine the outcome.
By contrast, Kalshi has no formal dispute-resolution system. It touts its “unbiased approach” to determining market outcomes in accordance with the event contract’s terms and conditions, which involves “meticulous adherence” to an unspecified set of “dedicated rules.”
The dispute over which bets on Time’s 2025 Person of the Year should pay out is representative of a more general phenomenon: event contracts are a form of derivative contract, analogous in some ways to credit derivatives, which provide compensation in the event of a default or other credit event.
When Greece retroactively introduced collective-action clauses into its sovereign-debt contracts in 2012, holders of credit-default swaps linked to Greek bonds argued that a credit event had occurred
Credit derivatives typically include extensive specifications, but still bump up against unexpected circumstances that are not covered by the contractual terms.
For example, when Greece retroactively introduced collective-action clauses into its sovereign-debt contracts in 2012, holders of credit-default swaps linked to Greek bonds argued that a credit event had occurred.
A “determinations committee” under the auspices of the International Swaps and Derivatives Association came to the same conclusion, triggering a stream of payments.
The existence of such committees, comprising representatives from buy-side and sell-side banks, is a tacit acknowledgement of the irreducible role played by human judgment in credit-derivatives markets.
Dependence on human interpretation
Online prediction markets have been pitched to regulators and the public as “quintessential truth machines,” in the words of Kalshi co-founder Tarek Mansour.
According to boosters, their value lies in collecting and distilling publicly available information through the price mechanism into credible predictions.
Online prediction markets remain shrouded in a thick mist of technical complexity and automation that serves to obscure their dependence on human interpretation
And like credit-derivatives markets, they remain shrouded in a thick mist of technical complexity and automation that serves to obscure their dependence on human interpretation.
But subjectivity is at the heart of resolving event contracts. A “truth machine” that dispenses with it is an empty promise. In this sense, prediction markets are no different from any other.
Erin Lockwood is Assistant Professor of Political Science at the University of California, Irvine.
