The New Attention Economy

In 2021, playwright Mike Bartlett had a vision and, like Thespis before him, decided the best route was to show people “the future” on stage. He wrote a play, and, in early 2022, I found myself at The Old Vic in London to see it.

The 47th presented a mentally-unstable Joe Biden who begs Kamala Harris to run in his place. Harris becomes the nominee in a haphazard, opaque, primary-less process only three months before election day, so late that states had already printed—and in some cases mailed—ballots bearing Biden’s name.

Fast forward 5 years, and if Bartlett had similar foresight about much-watched political events today, he may not have written a play but instead made a bet.

Prediction markets have exploded online, with betting volumes on track to surpass $300 billion in 2026 on topics ranging from whether the Boston Celtics will best the New York Knicks to the dress Beyoncé will wear to the Met Gala. Polymarket and its peers (Kalshi, DexWin, Drift, etc.) are decentralized markets that facilitate the anonymous placing and settlement of bets on various events newsworthy and not.

Fueled by tech and marketing, the public imagination has been captured to make money instead of art. Kalshi and Polymarket are currently discussing valuations over $20 billion dollars.

The current boom in prediction markets neatly reflects the casino activity that predominates online today. Their mechanisms also might bring us closer to the worst problem online today—of the Internet eating real life, where crowd wisdom prospectively defines reality, rather than predicting it. There are other areas where we accept counterintuitive or troubling relationships between the future and the present (Parkinson’s Law being a classic example), but wagers are different.

It is important to separate novelty from innovation. History suggests that to bet is to be human. The role of wagers in our society is controversial but longstanding.

In the 1600s, wagers on the docks of Rotterdam and London over whether ships would reach their destinations were the predecessors of today’s multi-trillion-dollar reinsurance market. Prior to this, in the great markets of the ancient world in the Souk at Damascus or the Bazaar at Ray (today Tehran), people would bet against Silk Road couriers on whether goods would arrive before spoilage, essentially the first supply chain insurance system. Anonymous slip betting on proposition bets in London’s bookrooms has been around for at least two centuries.

So Polymarket is fundamentally neither a pioneer nor a unique repository for these propositions: its innovation is evolutionary, not revolutionary. To date, most critiques of prediction markets have been puritanical assaults on gambling, Luddite anxiety that there might finally be a mainstream crypto use case, or abstract attacks on insider windfalls. While there may be valid ingredients hidden in the first and last of these three, none by itself offers an edible, let alone delicious argument. The actual problem of newly digitized betting on the future is that it interferes with, or even writes, the future itself.

Market Mechanisms

All market activity can be thought of as some mixture of frequency and magnitude. High-frequency bets are those where spreads are small, reasonable minds can and do differ, and opinions are prevalent in the general population (e.g., who will win the World Cup?). Niche, high-magnitude bets with high information asymmetry drive Polymarket and Kalshi—and this is where the big windfalls happen.

While Polymarket famously started with a no-fee model to win market share, today it charges fees for bets on certain high-volume events (like NCAA basketball) and charges a higher fee (up to 1.5% or so) when the market is equally supportive of both outcomes (a 50/50 bet), while it charges a very low fee where the outcome is deemed more certain. Notably, it is a maker-taker system (only the taker, or person placing the order, pays a fee). Finally, Polymarket also makes money on its USDC treasury holdings, as holding the USDC token provides that benefit to both individual and institutional holders (importantly, the interest rate varies between markets and exchanges).

The “killer feature” of Polymarket is not the crypto tokens or much-discussed decentralized infrastructure it utilizes, but rather the ability to anonymously participate in high-magnitude wagering on niche topics and unusual events that invite insider trading. There is no question prediction markets allow trades either 1) prohibited or 2) impossible to monetize in traditional equities markets (there is no way to bet on the New York Stock Exchange which day a firm’s CFO will announce his retirement, but the CFO can place that bet and, at the same time, be sure he’ll win).

The problem with anonymity-as-a-feature is not in the privacy it offers but, rather, in the behavior it encourages. Social norms and collective distaste would normally discourage betting on celebrity death or illness, or whose name will appear in a pedophile’s email archive; however, in an anonymous universe, these frowned-upon wagers are elevated from backroom dealings to televised sports.

The moral concerns are more debatable than the legal ones. Nearly everyone agrees that if you hypothetically had knowledge two months earlier that the terrorist attacks of September 11, 2001, were about to occur, it would be inappropriate to use that information to short airline stocks. Yet many argue betting on which celebrities appear in the Epstein files is harmless fun. So, where is the line between the two?

In the context of the American-Israeli military action in Iran and Lebanon, some experts believe those with inside knowledge placed winning bets. If true, these people presumably discarded (or at least disregarded) their duty of loyalty and put their role, rank, and security clearance at risk to place rigged bets.

In the context of betting on whose name would appear in the Epstein files, the people best-positioned to make money were those who knew they’d appear in the files. Or people who were in frequent contact with Epstein and knew who’d be mentioned. This gives already-wealthy insiders another way to monetize their knowledge.

Do we want to empower these kinds of information traffickers as the key nodes in both our social and financial networks? Do we want to make someone powerful and wealthy because they were willing to break attorney-client or physician-patient privilege to bet on a hot piece of celebrity gossip?

I’ve heard libertarian-idealist colleagues discuss that “if everyone were on Polymarket, it would approach a perfect information-sharing market!” or assert that “when insiders push markets toward what’s actually going to happen, they can hypothetically turn prediction markets into a source of real-time unfiltered news.”

But these “market as news” folks (people who claim to want an information-rich Coasean fantasyland of “everyone in markets”) forget that if “everyone” were truly in these markets and betting on “everything,” it would erode any information advantages and look wildly different from the current market of erroneous odds and idiosyncratic windfalls.

In an “everyone in markets” world, you couldn’t make much money betting on when the next ChatGPT version is going to come out, because all your colleagues at OpenAI would also leave that meeting and immediately place bets; you couldn’t make much money betting on whether the star quarterback is too injured to play because everyone with any information, including the quarterback, would be participating and the market would integrate that information, and hence reflect the correct odds, almost instantly.

Such a world might be less “fun” and certainly would be a low-trust environment. To gain an information advantage, how far would someone go even if the odds weren’t extraordinary?

The New Attention Economy

Those betting on Polymarket are giving more than just their money to the platform—they’re giving their attention—which is something more concerning.

The most impactful thing I’ve done was long before my professorial days: I invented the dwell time framework in the late 2000s, which YouTube adopted as its primary content ranking algorithm in 2012 (and descendants of which guide nearly every feed-driven user experience today).

Back in the “clickbait” era, platforms measured success by the click—a cheap, low-friction action not unlike answering a casual political poll. Dwell time shifted this paradigm by measuring how long someone actually stays, effectively penalizing the quick exit (where the thumbnail was cool but the video was boring). This matters for the exact same reason asking people to bet their own money on an election outcome is superior to asking for their opinion on who will win that election. Whether you are risking your capital on a Polymarket proposition or your finite attention on a video, that tangible “skin in the game” forces a much more honest calculation of value. If you’re unhappy with the Surowieckian wisdom of crowds, put those same crowds in Thalerian casinos.

Dwell time illustrates a concept relevant here. We talk about “spending time” and “wasting time” and it’s not wholly metaphorical; attention is like investing. Wagering, voting (the term “wasted vote” has a pecuniary scent), and deciding where you’ll spend (or waste?) your time are all very similar.

Despite our culturally-endorsed multitasking fantasies, humans can really only invest their attention in one video, song, or piece of written material at a time. By watching one video over another, you are effectively demoting every other video in the library. This is just about the strongest revealed preference a person can express—equivalent to allocating 100% of one’s “attention” portfolio to a single asset.

Crucially, this investment of attention isn't a one-time transaction; it is a live, continuous wager. Every additional second a viewer lingers is letting a micro-bet “ride” and hoping the current content will deliver a higher payoff than the nearly-infinite alternatives waiting just one swipe or click away. The longer the dwell, the higher the conviction in that wager. The viewers aren't just passively consuming the next second of a video or song (or dating profile, the list goes on); they are actively re-staking bets on the belief that staying put is the winning wager.

A common critique suggests this convergence of betting and browsing is a joyless, clinical exercise—that reducing our interests to "investments" strips away the inherent fun of discovery. This view posits that most people don't actually want to "make things happen" or carry the burden of being right; they just want to be entertained. But my data from the dwell-time era suggests the exact opposite.

I found that the highest utility—the "deepest" dwell—occurs when the user feels a sense of agency, however small, over the outcome. The thrill isn't just in watching the wheel spin; it's in the infinitesimal, recurring decision to stay in the game. Betting on an outcome and then watching it manifest doesn't ruin the experience; it provides the ultimate feedback loop, transforming the viewer from a spectator into a stakeholder.

The person with a wager on the game watches the game not only more intently, but fundamentally differently. This may be tolerable for sport but dubious for politics or warfare.

This bettor-as-participant dynamic mirrors the addictive mechanics of prediction markets. And, as with all maturing markets, there are derivatives. So if you’re unsure which bet to place in a prediction market, you can instead log into your Noise.xyz account and bet on how popular prediction markets will be this month. This meta-betting (betting on betting) has many of the same defects as the underlying market, including that a person with enough resources could bet on market volume and then create that volume.

If there is any concrete (rather than theoretical) utilitarian justification for these prediction markets’ existence, it is as a hedge. In this use case, a person bets on something he or she fears and, in turn, is paid if it occurs. This allows the market to act as a cheap, narrow insurance policy against an outcome. For instance, a farmer might bet on low rainfall figures for a given county, reaping a reward that will partially make up for his crop failing.

However, we have a way to insure against key commercial outcomes. It’s called… insurance.

Adam Smith wrote that a gentleman aspires not only to be loved but to be “lovely,” and by that he meant respected, regarded as honorable, and seen as a good ally in our common human enterprise.

My central objection, then, is this Smithian one: Whether such wagers are allowed or disallowed, tasteful or abhorrent, gentlemanly or scoundrelish is not the question, as no question need be asked; markets like Polymarket make—and have always made—the world less human and lovely, and we should not be so thirsty for market efficiency that we lose all appetite for humanness and loveliness.

Illustrations by Pedro Nekoi.