Using Scott Alexander as my guinea pig, I have created quite a few different types of scandel markets. They all have their own flaws.
The ones about crimes may be lower than you'd expect, since the legal system moves so slowly and the conviction might not go through before the end date. And there are lots of things Scott could do that will make his friends angry at him, but wouldn't be a felony.
The ones about accusations are easy for a bad actor to manipulate.
The ones about social outrage are highly subjective and will probably result in arguments over resolution, and they can be self-fulfilling if people make committments to ostracise anyone whose market goes above a certain percentage.
The one that's a poll may end up being more about the makeup of Manifold's user base than it is about anything Scott was or was not found to have done, and is also highly subjective.
I'm making this post as a central place to discuss the upsides and downsides of different market types, and hopfully arrive at a robust structure that we can use for markets on the reliability of public figures across the board.
And here's a version based on being "arrested". Any reason for the arrest can count, including minor misdemeanors. Arrests count even if the case is later dismissed, or charges are dropped.
https://manifold.markets/Jotto999/will-scott-alexander-psychiatrist-a
I'm fond of my "charges" wording, since convictions are unreliable and are often avoided. Admittedly, "charges" fails to capture social risks that aren't criminal charges. I like that it's so unambiguous though (you can tell I'm coming here from Metaculus lol)
Brief aside: I know Scott Alexander says he's completely certain charge markets would be corrupted, at sufficiently-high levels of adoption. That's plausible and I'm sure some sort of corruption would happen (though I suspect partial defenses would be thought of). Still, that universe where scandal markets are THAT accepted and widespread...the pendulum is currently sitting at a very opposite extreme, and has been as long as I've been alive. And there is big institutional force against prediction markets more generally. So I'm currently just not worried about those scenarios, though they're interesting to think about.
https://manifold.markets/Jotto999/will-scott-alexander-scott-siskind
I made a market on whether any market will cause real harm to someone, which would include someone spreading false rumors about them.
From Scott's post:
Let’s think about the possible failure modes.
Someone might make a false rumor (eg of sexual assault). Traders would trade on the false rumor and the scandal market would go up. This doesn’t exactly seem like a failure mode; this is the prediction market correctly reflecting the current consensus of smart people (which happens to be false). It’s not any worse than the normal thing where people make false rumors and other people believe them, except that there’s an easy way to access the fact that it’s happening. It might even be better - “always believe women” is more compelling as a slogan than as an investment thesis.
Someone might make a false rumor that they wouldn’t have made otherwise, specifically to move a market. That seems like a pretty cartoonish level of evil in order to get fake play money, but there are some real weirdos out there.
The resolution criteria could make things unfair in one direction or another. If it’s “convicted by a court”, very few sexual assault cases ever make it to court, probably that would be a low probability even if the person involved was a known assaulter. But if it’s “so-and-so thinks it happens”, it depends a lot who so-and-so is. There’s even risk of a hyperstitional cascade, where traders believe if the prediction market is high than so-and-so will believe it, so the prediction market ironically produces its own outcome.
What if someone tries to manipulate the market to destroy someone else’s reputation (or to protect their own)? In theory prediction markets are robust to this. In practice, low-volume prediction markets (like these) aren’t robust against someone willing to spend vastly more resources than everyone else combined and okay with losing all their money, which the friends or enemies of a scandal market target might be. I would be more comfortable with these markets if many people reviewed them frequently, which would make them stop being effectively-low-volume.
And from a response in a comment:
To expand on this: as long as weird nerds and prediction markets are ignored by whatever the prediction market is based on, the market can be accurate. (But useless, because it is ignored.)
As soon as the real-world decisions are based on the prediction market, the market becomes degenerate in exactly the same way as a "predict where this market will close at the end of 2022" market.
> The resolution criteria could make things unfair in one direction or another. If it’s “convicted by a court”, very few sexual assault cases ever make it to court, probably that would be a low probability even if the person involved was a known assaulter. But if it’s “so-and-so thinks it happens”, it depends a lot who so-and-so is. There’s even risk of a hyperstitional cascade, where traders believe if the prediction market is high than so-and-so will believe it, so the prediction market ironically produces its own outcome.
In the middle, you can have "ostracism from the community of weird nerds with prediction markets" depend on a prediction market about a jury conviction. But notice that this strictly depends on normal people not caring about weird nerds with prediction markets. The very second that weird nerds and prediction markets become socially relevant, the prediction market directly causes its own outcome.
The way the system currently works is that some guy has bad vibes and people feel the feds ought to do something, so he gets convicted of a crime, and that crime can be completely untethered to the reasons for the bad vibes. This is, more or less by definition, a system that people are broadly happy with, in the same way that people are "happy" with Congress: nobody wants to change their piece of it.
If prediction markets become relevant, explicitly and specifically for the purpose of using the prediction market to ostracize people as Bad Monkeys, then the market creates its own outcome.
Try --- I know this is hard, but try to imagine a world where weird nerds and prediction markets weren't ignored by normal people. Try to imagine a world where "95% chance Austin Chen will be charged with a felony before 2030" was something that normal people cared about. In that world, if the prediction market says 95%, then Austin Chen WILL be charged with a felony before the next District Attorney election cycle. The specific felony could be anything; Austin Chen is a public figure, you can always come up with something.
To be more pointed, imagine a world where weird nerds and prediction markets weren't ignored by normal people, and imagine a prediction market on "Will Scott Alexander be convicted of sexual assault before 2030?" And imagine that that market sits at 73%. And imagine that Scott Alexander is not convicted of sexual assault before 2030. And imagine the headlines that will be printed after that, about Scott Alexander getting away with sexual assault after THE PREDICTION MARKET WAS AT 73%! (I kid, of course. The headlines wouldn't report what the prediction market had been immediately before the outcome was reported, the headlines would report whatever was the highest number the market had ever reached.) And imagine what would happen to the prosecutor who failed to convict, the judge who failed to grant timely justice --- God forbid, the defense witnesses who testified in favor of someone with a market at 73%. The defense attorney.
And then go back and try, again, to imagine that Scott Alexander is not convicted of sexual assault before 2030.
Put yourself in the shoes of the prosecutor, the judge, the defense attorney --- God forbid, the defense witnesses.
It won't happen. It is un-possible. In a world where the market sits at 73%, and that fact matters to people who matter...then 73% is in fact way too low.
...and now imagine being a trader, in that situation, looking at that market. You can run through this entire chain of logic and realize that 73% is way too low. So you buy. And the market goes up. Which makes nonconviction even more politically impossible. Which makes the market go up.
The problem with "predict where this market will close at the end of 2022" markets is that they pay out the same completely regardless of what happens in reality. They are, thus, simply and literally popularity contests. It doesn't cost anything, on net, to move the market, because moving the market pays out the same amount that moving the market costs.
As long as scandal markets are ignored, they're accurate but pointless. If scandal markets are not ignored, then they aren't predicting anything except the outcome of the scandal market itself.
I don't like the one where you are the arbiter - that requires any participants to model the ethics and peer-pressures you will have at the time of resolution.
How about instead:
Will {trustworthy well-known reliable person} publicly state that Scott Alexander did something highly unethical by ...?
I mean, that's the usual problem with trust in market creators. Having public statements by to base the resolution would help with that (in low-stakes situations)
@lu This seems like it's just /IsaacKing/will-scott-alexander-do-anything-be, but made by a market creator in which you place more trust.
@IsaacKing Yes, exactly. Also a market creator where people can somewhat estimate which things would be classified as highly unethical.
@NathanpmYoung reading that back...
I would like a market for "Will Scott be guilty of awful bad behaviour, either via legal proceedings or the view of the manifold community"
@NathanpmYoung Is that not already covered by /IsaacKing/will-scott-alexander-be-convicted-o-a372f8cf02bd plus /IsaacKing/at-the-beginning-of-2030-what-perce? Or did you want a single market that covers both?