Physicists have a concept called the observer effect. In certain systems, the act of observing a phenomenon changes the phenomenon itself.

Prediction markets may be running into something very similar.

For years, these markets have been promoted as an elegant forecasting tool. The premise is simple. Let people bet on the outcome of real-world events and the market price becomes a probability. The wisdom of crowds reveals what the future is likely to hold.

In theory, it is a powerful idea. In practice, the reality may be more complicated.

Once real money, insider information and algorithmic trading enter the system, prediction markets do not just measure probabilities. They begin to create incentives. And incentives have a way of shaping behavior.

The result is a paradox. The very act of betting on the future can start to influence the future itself.

Prediction markets are not a new concept. Economists have been studying them for decades as a potential alternative to polling and expert forecasting. Platforms such as Polymarket and Kalshi have recently pushed the idea into the mainstream by allowing users to trade contracts tied to real-world outcomes.

A contract trading at seventy cents implies a 70% probability of the event occurring. The market aggregates information from participants who are willing to risk their own money on being right.

Supporters argue that this produces a more honest signal than surveys or pundit panels. People can say anything in a poll. They behave differently when their own capital is on the line.

That logic makes sense as far as it goes. Markets are powerful tools for aggregating information. But they also create incentives. Once large sums of money are attached to an outcome, participants stop being passive observers.

They become stakeholders.

At that point, the observer effect begins to emerge. When millions of dollars ride on an event, some participants may start thinking about more than prediction. They may begin thinking about influence. A trader who stands to gain significantly from a specific outcome has a financial interest in seeing that outcome occur.

In many situations, this dynamic is harmless. Prediction markets tied to sports or economic statistics do not carry obvious risks. The stakes change, however, when markets move into areas such as elections, geopolitical events or the deaths of public figures.

Critics have already raised concerns about contracts tied to conflict scenarios and the mortality of political leaders. These have been described by some observers as “death markets,” a label that captures the discomfort many people feel when financial incentives become attached to real-world tragedies.

Even when no one is actively trying to influence events, the perception alone can damage trust in the system.

The insider problem makes the situation even more complicated.

Prediction markets assume participants have roughly equal access to information. In reality, some people inevitably know more than others. When insiders enter the market, prices may reflect privileged knowledge rather than collective insight.

Reports have already surfaced of individuals trading on nonpublic information connected to corporate or geopolitical developments. In one widely discussed case, an employee at OpenAI was reportedly dismissed after allegedly using confidential company information to trade on Polymarket.

That kind of activity resembles insider trading in financial markets. When it occurs in prediction markets, the implications can extend beyond corporate earnings or product announcements. The information involved might relate to policy decisions, military actions or regulatory moves.

At that point, the market stops looking like a neutral forecasting tool. It begins to resemble a system where those closest to power can monetize information before the public even knows what happened.

Artificial intelligence adds another layer to this emerging ecosystem.

AI systems are becoming increasingly capable of forecasting complex trends by analyzing massive datasets. At the same time, generative AI has made it easier than ever to create synthetic information. Fabricated documents, deepfake videos and AI-generated narratives can circulate rapidly across social media.

Prediction markets could become highly sensitive to these signals. News, whether real or fabricated, moves probabilities. Those probabilities are then cited as evidence that the event is likely to occur. The feedback loop between information, markets and narrative can accelerate quickly.

In such an environment, prediction markets risk becoming less like forecasting tools and more like algorithmic casinos connected to real-world events.

Regulators are beginning to pay attention. The Commodity Futures Trading Commission is already evaluating how these platforms should be classified. Some lawmakers have also proposed restrictions on government officials participating in prediction markets.

Their concern is straightforward. If policymakers can place bets on events they influence, the incentive structure becomes problematic.

This debate is only beginning. Supporters of prediction markets argue that they provide valuable information signals that traditional institutions often miss. Critics worry that attaching financial rewards to sensitive events could create incentives that distort behavior or erode public trust.

Both arguments contain elements of truth.

Markets can be remarkably effective at discovering information. But they are also powerful engines for shaping incentives. Once enough money is tied to a particular outcome, the line between prediction and participation begins to blur.

Prediction markets were intended to reveal the future.

The question we may soon face is whether they also help create it.