Since Simon rejected the 15 wagers, I find language liker 'Ehrlich won/lost' or 'Simon won/lost' disturbing. If I propose to Caplan that the Sun will rise tomorrow - will I win ? Will Caplan lose anything in not taking the wager ?
I really appreciate how you've shown the true value of betting as a means of getting to the truth. This blog post is by a like-minded blockchain programmer named Jason Dreyzehner who is laser-focused on making massive decentralized prediction markets a reality. Worth a read to get his vision of a world with freely-accessible prediction market data (feel free to skip the technical details on implementation).
"By cutting through misinformation, prediction markets offer the potential to save countless lives, protect homes and businesses from preventable disasters, reduce the power of special interests, root out corruption, and help people better understand and plan for the future."
While I think bets have a place, and the need to set operationalizeable conditions for victory is enormously important, I think you are too bullish about how epistemically purifying they are.
The basic issue is that bets seduce us into thinking that the monetary stakes at issue represent the incentives of the parties, eg, if rational you only bet at odds of if you judge the probability of winning to exceed o/(1+o). However, this is virtually never the case.
In the vast vast majority of cases the amount of money bet is nowhere near enough to be the dominant factor. It's almost always prestige and those returns are very obscure, eg, since Ehrlich had already lost one bet losing again would be less costly to him than to Simon and winning would be more valuable to him.
And if you try to make the bets large enough to outweigh status concerns they may have to be bet your house sized big at which point the non-linear returns to wealth and differences in wealth between parties screw it all up.
I think the role of money in these types of bets is somewhat more complicated than either you or Caplan seem to believe. I agree that the expectation of winning or losing the amount at stake is not the primary explanation of people's betting behavior. However, attaching a monetary reward to a bet serves to properly formalize it and sends a credible public signal that both parties are taking it seriously and putting their reputation at stake.
I agree with that claim so I must not have been as clear as I thought.
What's important is the pressure to create some kind of clearly operationalized conditions for winning. I agree that it's important that there be concrete stakes be it money or some kind of specific performance.
However, I think what we can't do is read off confidence from the odds given in the bets. As such, bets are a pretty useful tool for settling disagreements between sides/theories where one side predicts something is highly likely to happen and the other side predicts it's highly unlikely to happen.
Where they aren't so useful is in disagreements where one side thinks some outcome will happen 20% of the time and the other thinks it happens 85% of the time. The bet forces the sides to operationalize winning but if I had the 20% side I wouldn't accept a public bet at 4.1:1 odds bc even if I'm right about my probability the expected payoff of the money is unlikely to compensate for the status effects of having to publicly concede they won the bet (which I think will happen most of the time).
In cases where theories predict many similar outcomes this can often be solved just by betting on aggregates (eg how many times out of 100 does the indicator happen) to push the probabilities to extremes.
In those cases it would be great to have a real, liquid, uncapped prediction market but damn fucking SEC.
To illustrate why I think bets aren't as purifying as you suggest imagine you're Ehrlich here. I think you'd happily accept an expected loss of 10k just for a 10% chance to win a bet with Simon so you won't be remembered in history as that guy who got trounced when it came to bet on his theory.
"Unless progress is amazingly uniform, you’d expect the richest 10% to pull away from the poorest 10%."
Ehrlich and Schneider probably had some half-baked theory about rich countries buying up all the world's resources and exacerbating inequality by pricing out poor countries out of the market entirely or something like that.
Since Simon rejected the 15 wagers, I find language liker 'Ehrlich won/lost' or 'Simon won/lost' disturbing. If I propose to Caplan that the Sun will rise tomorrow - will I win ? Will Caplan lose anything in not taking the wager ?
Great article & thoughts.
I really appreciate how you've shown the true value of betting as a means of getting to the truth. This blog post is by a like-minded blockchain programmer named Jason Dreyzehner who is laser-focused on making massive decentralized prediction markets a reality. Worth a read to get his vision of a world with freely-accessible prediction market data (feel free to skip the technical details on implementation).
"By cutting through misinformation, prediction markets offer the potential to save countless lives, protect homes and businesses from preventable disasters, reduce the power of special interests, root out corruption, and help people better understand and plan for the future."
https://blog.bitjson.com/prediction-markets-on-bitcoin-cash/
While I think bets have a place, and the need to set operationalizeable conditions for victory is enormously important, I think you are too bullish about how epistemically purifying they are.
The basic issue is that bets seduce us into thinking that the monetary stakes at issue represent the incentives of the parties, eg, if rational you only bet at odds of if you judge the probability of winning to exceed o/(1+o). However, this is virtually never the case.
In the vast vast majority of cases the amount of money bet is nowhere near enough to be the dominant factor. It's almost always prestige and those returns are very obscure, eg, since Ehrlich had already lost one bet losing again would be less costly to him than to Simon and winning would be more valuable to him.
And if you try to make the bets large enough to outweigh status concerns they may have to be bet your house sized big at which point the non-linear returns to wealth and differences in wealth between parties screw it all up.
I think the role of money in these types of bets is somewhat more complicated than either you or Caplan seem to believe. I agree that the expectation of winning or losing the amount at stake is not the primary explanation of people's betting behavior. However, attaching a monetary reward to a bet serves to properly formalize it and sends a credible public signal that both parties are taking it seriously and putting their reputation at stake.
I agree with that claim so I must not have been as clear as I thought.
What's important is the pressure to create some kind of clearly operationalized conditions for winning. I agree that it's important that there be concrete stakes be it money or some kind of specific performance.
However, I think what we can't do is read off confidence from the odds given in the bets. As such, bets are a pretty useful tool for settling disagreements between sides/theories where one side predicts something is highly likely to happen and the other side predicts it's highly unlikely to happen.
Where they aren't so useful is in disagreements where one side thinks some outcome will happen 20% of the time and the other thinks it happens 85% of the time. The bet forces the sides to operationalize winning but if I had the 20% side I wouldn't accept a public bet at 4.1:1 odds bc even if I'm right about my probability the expected payoff of the money is unlikely to compensate for the status effects of having to publicly concede they won the bet (which I think will happen most of the time).
In cases where theories predict many similar outcomes this can often be solved just by betting on aggregates (eg how many times out of 100 does the indicator happen) to push the probabilities to extremes.
In those cases it would be great to have a real, liquid, uncapped prediction market but damn fucking SEC.
To illustrate why I think bets aren't as purifying as you suggest imagine you're Ehrlich here. I think you'd happily accept an expected loss of 10k just for a 10% chance to win a bet with Simon so you won't be remembered in history as that guy who got trounced when it came to bet on his theory.
The new bets offered by Ehrlich et al were indeed much more milquetoast than his usual doomsayer rhetoric.
And as you pointed out, they were essentially admitting that Simon was at least half right.
Of course, approximately no one will take notice of any of this.
Simon didn't take the bets - how is he "right" ? Ehrlich was wrong-ish tho.
Excellent.
"Unless progress is amazingly uniform, you’d expect the richest 10% to pull away from the poorest 10%."
Ehrlich and Schneider probably had some half-baked theory about rich countries buying up all the world's resources and exacerbating inequality by pricing out poor countries out of the market entirely or something like that.
You should bet Richard Crim (https://daviddfriedman.substack.com/p/a-climate-falsehood-you-can-check/comments), especially now that he has got the attention of Scott Alexander (https://astralcodexten.substack.com/p/give-up-seventy-percent-of-the-way/comment/13435802)
I think Crim will remain beneath Caplan's notice if he doesn't comment under Caplan's posts.