Sure, "incentives matter" if you're seeking healthcare. That's why you're nibbled to death by co-pays and deductibles - if you could get healthcare whenever you felt like it, you might get *too much* healthcare. "Incentives matter," so we have to make sure that you only seek care when you *really* need it:

https://pluralistic.net/2025/04/14/timmy-share/#a-superior-moral-justification-for-selfishness

2/

But *rich people* don't need to be disciplined by incentives. They can get no-bid contracts with Uncle Sucker without being tempted to rip off the USA. They can force their workers into nondisparagement clauses without being tempted to act like a colossal asshole, secure in the knowledge that they can sue workers who tattle on them.

3/

They can force their workers into noncompete clauses without being tempted to underpay and abuse their workers, secure in the knowledge that they can sue workers who take their labor elsewhere. They can force their workers into binding arbitration clauses without being tempted into maiming or killing them, secure in the knowledge that the workers can't sue *them*.

4/

So incentives matter...when you're fucking over working people. But incentives don't matter, when you're gilding the Epstein class's lilies.

But incentives really *do* matter. That's the premise of Goodhart's law: "When a measure becomes a target, it ceases to be a good measure." This comes up all the time. Google got its start by observing that people who made websites linked to other websites that they found important or worthy or informative.

5/

With this insight, Google repurposed the academic practice of "citation analysis" to predict which pages on the internet were most authoritative, calling it Pagerank.

Google Search, powered by Pagerank, was vastly superior to any search engine in history. But as soon as Google became the most popular search engine, people started making links to *bad* websites - sites filled with spam and malware and junk - in order to game the results.

6/

The metric - inbound links - became a target - *get* inbound links - and stopped being a useful metric.

There is something quite wonderful and life affirming about the idea of Pagerank: the idea that people are, on average, pretty good at figuring out what's good. Rather than taking Yahoo's approach of having experts rank and categorize every website on earth, Google trusted "the wisdom of crowds" and it worked (until they created an incentive to subvert it).

7/

"The wisdom of crowds" was in the air in those days. James Surowiecki had a massive bestseller with that title in 2004, expounding on the idea that people were, in aggregate, good at figuring stuff out:

https://en.wikipedia.org/wiki/The_Wisdom_of_Crowds

8/

Surowiecki's book revolved around a famous anecdote from 1906, when 800 people at the Plymouth county fair were invited to guess at the weight of a slaughtered and dressed ox. Statistician (and eugenicist creep) Francis Galton noted that the average guess of 1207 lbs was within 1% of the actual weight, 1198 lbs.

9/

This turns out to be a repeatable phenomenon: if you get a lot of people - non-experts, experts, people paying close attention, people who barely think about it - to guess about something, the average is surprisingly accurate. Importantly, it's often more accurate than the best guess of experts.

This idea of the wisdom of crowds inspired a lot of 2000s-era internet projects. Some of them (Yahoo Answers) were pretty bad. Others (Wikipedia) were astounding.

10/

Of course, economists observed that "the wisdom of crowds" sounds a lot like the idea of "price discovery" - the idea that markets are a way of processing widely diffused information about desires and capacity in order to derive and emit signals about what should be produced.

11/

Economists have long spoken of future events being "priced in" to markets - for example, the price of oil today reflects more than the diminished supply resulting from Trump's military blunders, it also reflects "the market's" belief that oil production capacity will be disrupted for a long time to come.

12/

Add up all the different buyers' and sellers' guesses about the future of oil (incorporating diffuse knowledge about damage to infrastructure, capacity to rebuild, and intentions of the actors) and (we're told) we'll get a number that accurately reflects the real situation.

And, unlike Pagerank, this number can't be manipulated by flooding the system with spurious, self-serving inputs. If you want to move this price, you have to buy or sell something, which costs money.

13/

And because the market is "deep" (with a lot of participants), the sums you'd have to inject into the system to alter its consensus is incredibly large - more than you could possibly stand to make by manipulating the price itself. Incentives matter.

Put "markets," "the wisdom of crowds" and "incentives matter" together and you get "prediction markets."

14/