Creating a Less Wrong prediction market

I will bet 500 karma that a funny picture thread will appear on Less Wrong within one year. If anyone is interested in the bet, we can better define terms.

Right now the LW software doesn't support karma transfers. Until it does and we can develop a more robust prediction market, let's just record the karma transfers on the wiki page that already exists for this purpose.

I will also give 100 karma to anyone that donates $10 to the SIAI before the current fundraising campaign is over.

10,000 karma for the first person with a karma transfer source code patch?

52 comments, sorted by
magical algorithm
Highlighting new comments since Today at 6:29 PM
Select new highlight date

Karma provides an incentive structure for the production of LessWrong content; changing that structure without understanding the possible consequences, good and ill, might be unadvisable.

On the other hand, creating a new "play" currency, redeemable in donations and tied to various games that exercise the skills valued by this community - such as good calibration and discrimination - that could be a useful and fun project.

I agree. It seems that most people have a standby rationality mode which kicks in when their personal interests come into play.

Regarding permitting karma transfers for the purposes of betting:

Apart from Morendil's objection, which I also endorse, I don't think anyone's noted that it probably will spoil the usefulness of betting as I take it to be understood by people on Less Wrong, following Robin Hanson and others.

Betting supposedly gets people to leave aside the bullshit (I mean this in the technical sense, of course) by staking actual utilons. Less Wrong karma, sorry to say, does not consist of actual utilons, even though people here value it at least insofar as they would rather have it than not have it.

So using LW karma as betting currency seems like a way to bring betting back toward the realm of bullshit-filled barroom (or Usenet) argument rather than a way to force people to be clearer about stating exactly what they believe to be the case.

(NB I understand that dollars = utilons is not exact, but it seems to me the approximation has often been considered a useful one in other discussions. You can in fact use dollars to buy things you want and need, something not true of Less Wrong karma.)

I disagree that not-being-money is the problem. I do, however, think that mixing the betting currency up with karma could have undesirable impacts on both systems.

My impression is that even fake currency serves some, albeit not all, of the purpose of focusing cognitive attention on probabilities over impressive sound-bites. And the fake bets do help us to calibrate since they actually make us formulate specific predictions that are actually written down someplace.

All the benefits of betting "even fake currency" that you name: I absolutely agree -- particularly for people who are self-consciously interested in rationality.

But I think you overestimate the marginal utility of Less Wrong karma (in the amounts that frequent contributors to this site tend to possess it) versus money. Fundamentally, absolutely nothing changes for a person who loses most of their Less Wrong karma, except the ability to make more ill-considered wagers. We can't expect top-notch rationality skills from people when those are the stakes. So as I say: better than bullshitting on Usenet, less good than a true betting market.

Also, I have noticed that offers to wager are a subtle applause light on Less Wrong, garnering substantial karma at times for comments that consist of a prediction and an offer to back it up with a bet. If I am right about that, using karma as betting currency obviously somewhat changes the nature of the decision to propose a bet.

These considerations lead me to think that the result would depend too heavily on the good will of the commenters on the site. Again, though, I think you're mostly right.

Expanding somewhat on the problem of not using real money - if people lose all their money they can't bet any more. This combats obstinance. If people are simply too stubborn to update on evidence then they will no longer be able to bet. Their poor judgement is removed from the market. The only way this would come in to play on lesswrong is if the bets started to become seriously large. If the system was karma transfer (which I recommended against) then poorly calibrated predictors would lose their karma reserves. This would benefit the quality of large stake predictions but it would completely alter the meaning implied by karma.

Errr... now that I read that back to myself it tempts me towards perverse speculation that that the change in implicit meaning of karma could be a good thing. At least to the extent that karma can be used as a metric for quality of expected predictions (only a minor purpose).

I think a standard one-way virtual economy approach would be ideal for this: Track users' "betting scores" separately from their karma, allow them to purchase score with karma, but not vice versa. Then lousy betters can blow their general reputation if they so choose, but a skilled better still needs to earn karma the "honest" way.

It's pretty well accepted that people play far, far different poker games when real money rather than fake money is on the line e.g.. I predict a similar outcome if karma were used to bet regularly.

On a system of poker for karma, I would expect that people (and their pokerbots) would play much, much better than people play in the "play money" games of online casinos. In online casino play money games, most people give away extra money in every single hand. In more complex play money poker games, such as http://gpokr.com, most of the money is won by people that are cheating by colluding, but at least they know how to play poker.

Karma poker should fall somewhere in between playing for real money and play money. People will play worse than they would with real money, but they won't just throw their money away every single hand like people do with play money. We also won't have the extent of collusion problems that gpokr has because the complete Bayesian history of every player will be available for analysis and people can call collusion whenever they see it.

We could eventually set up a karma poker market. 1v1 Texas Hold'em is a game that bots can currently win against humans.

I will bet 500 karma that a funny picture thread will appear on Less Wrong within one year. If anyone is interested in the bet, we can better define terms.

What odds do you give on the first funny picture thread being downvoted to less than -50?

This could get messy quickly though, I could generate infinite karma by giving everyone who upvotes this 5 karma. We probably need a lot of rules.

This could get messy quickly though, I could generate infinite karma by giving everyone who upvotes this 5 karma. We probably need a lot of rules.

As it stands, we should assume that nobody's trying to maliciously game karma. Sockpuppetry is already an option for that, so unless we start limiting upvotes somehow there's no reason to fix that problem.

What? No, you would lose 4 karma every time you did that (+1 from the upvote; -5 from the transfer). The karma transfers are zero-sum. Though, you could give half the people who upvote this 1 karma and come out ahead.

I was reading "this" as "this post" - one upvote is worth 10 karma, so that's +5 for each of them.

Yes, "this" meant the post, not the comment.

Oh. Yeah. That makes a lot more sense. Sorry.

He could give them karma by upvoting their comments. No-one's suggesting a finite amount of karma.

Oh. Yeah. You're right. I guess we should have a rule that karma transfers should be zero-sum, and you shouldn't upvote a post unless you actually like that particular post.

An interesting note here that is hinted in earlier comments but not suggested outright...

Some folks seem to use karma as some sort of proxy for Rationality Belt Level... a betting system using karma would more tightly bind to this idea, as we generally see making good bets as a sign of rationality (all else equal).

I'd take the bet, depending on the details. You mean a top level post for this purpose? What happens if it is voted down?

Do we all agree that this is an acceptable use for karma? I'm not sure if giving out karma for giving to SIAI is a great idea, it's an awful lot like buying karma.

I meant anywhere on the site. A top level post would count, or if were to be a subthread somewhere, such as this page or on "Babies and Bunnies". I think a 3 picture long thread should count. I'm not sure how I feel about including the karma of the picture containing comments in the bet, but I would consider it if you propose terms.

It is an awful lot like buying karma, but I'm not sure that's a bad thing. If people really want karma so much, why not let them buy it? They are the ones creating the monetary value here. Don't most people here think that karma is unimportant anyways?

I doubt that karma will regularly be given out for donations if karma transfers become popular here. I was just hoping to draw out two or three people to donate and make this post more interesting by linking it up with an existing meme.

For Less Wrong karma to eventually be worth something as part of a reputational economy, there is no way around allowing wealth transfers between karma and real world currency/services. Doing this in a way that scales to the globalist level while not being subject to anything like the USA PATRIOT Act is an unsolved problem.

So the Top Contributors section remains meaningful, we could separate that ranking list from a list for most karma.

For Less Wrong karma to eventually be worth something as part of a reputational economy, there is no way around allowing wealth transfers between karma and real world currency/services.

I'm fine with LW karma not being worth something as part of a reputational economy. In fact, I'd prefer it if we didn't total users' karma at all and just kept track of it per post, as a measure of how good an individual post is. The temptation is too strong to think of a user's total karma as actually meaning something ("reputational economy"), when there are all sorts of confounding factors like the number of posts made, how new someone is, how wittily they write, and how much people agree or disagree with their posts.

I second the idea that a LW prediction market should use real money because it's actually valuable.

I don't think many people use karma to judge a member; rather, frequently reading their posts causes me to respect them, and this correlates with their karma.

Having said that, I am generalising from one case.

Incidentally, there's already quite a karma-market in discussing karma. You could easily generate most of your karma discussing karma.

Incidentally, there's already quite a karma-market in discussing karma. You could easily generate most of your karma discussing karma.

Upvoted.

You could easily generate most of your karma being cleverly meta and self-referential.

Incidentally, there's already quite a karma-market in discussing karma. You could easily generate most of your karma discussing karma.

Upvoted.

Upvoted to give you your 50,000th karma :-)

Is there any bound to the number of levels this works at? And if so, is this post at level omega?

Was that just ordinal-dropping, or is there actually a reason why that particular ordinal has something to do with this situation?

ciphergoth's post was at the level w (omega). Posts about ciphergoth's post and posts about those resulting posts are at level w + 1, w + 2, etc. Posts about that entire sequence of posts would be level 2w. Then you can talk about the sequence (w, 2w, 3w, 4w, ...) itself. The recursive tower seems endless, but there actually is a least upper bound, and that bound is the Church-Kleene ordinal.

Ok, you get 100 w.

If anyone wants some w+1, 2w, etc, please explain why.

I'm fine with LW karma not being worth something as part of a reputational economy. In fact, I'd prefer it if we didn't total users' karma at all and just kept track of it per post, as a measure of how good an individual post is. The temptation is too strong to think of a user's total karma as actually meaning something ("reputational economy"), when there are all sorts of confounding factors like the number of posts made, how new someone is, how wittily they write, and how much people agree or disagree with their posts.

I wonder if people's position on this correlates with how much karma they already have.

Real money would definitely be preferable but I'm pretty sure that idea is a non-starter. You run into gambling regulations with prediction markets and it appears that it is currently pretty much impossible to run any kind of gambling operation in a way that is reasonably accessible to anyone in the US. US credit card companies generally aren't allowed to process payments for online gambling sites. Even outside the US there is a minefield of regulation to navigate.

InTrade suffers from this - even though I live in Canada rather than the US I haven't been able to fund my account with a credit card.

We could just do it just like the informal bets here: trust each other to PayPal the money to the winner, or bet charitable donations.

Don't most people here think that karma is unimportant anyways?

Assert it? Yes. Believe it? Probably. Act like they believe it? No, and those most vocal in asserting their indifference seem to care about it more than others.

Mind you, people seem to care more about karma in the present moment on individual comments far more than they care about the number shown on the right hand of the screen, as can be expected. The same difference seems to apply to money, from what I can tell.

I don't care about karma, but I do care about the things that karma is supposed to be a proxy for. Therefore, I act mostly like I care about karma, except that I oppose anything which weakens the connection between karma and post quality.

I don't know if this would complicate things but it seems like karma sums can easily fail to tell us what we want to know about a poster. It would be nice to have karma stats like karma/comment, percentage of karma from top-level posts, subjective karma rankings (which is just the karma you've given out), karma weighted for posters you like, karma weighted for poster you dislike etc.

This is starting to sound like Whuffie. Which might be nifty, actually, but computationally difficult.

The social aspects of implementing Whuffie are much harder than the computational aspects.

:-) Er. If it is starting to sound like a Whuffie that is because I just stole the last three ideas from that wikipedia entry. I literally read that page five minutes ago, after I googled "reputation economy" and started reading the Doctorow book when Kevin brought it up above.

For Less Wrong karma to eventually be worth something as part of a reputational economy, there is no way around allowing wealth transfers between karma and real world currency/services. Doing this in a way that scales to the globalist level while not being subject to anything like the USA PATRIOT Act is an unsolved problem.

Can you say more about this?

It's kind of a vague statement acknowledging the limits of starting a new system that becomes more powerful than our existing global system while acknowledging that possibility. There's a long way to go between karma transfers on Less Wrong and the Whuffie system from Down and Out in the Magic Kingdom.

It seems that if there were karma transfers in place, betting against a funny picture post would be an almost guaranteed loss.

We could set up some ground rules that would exclude you specifically from starting the post, but I don't see how any rules set up in advance would prevent collusion to create the thread. Also, any karma lost for making an explicitly bad thread would be more than made up for with the 500 karma win.

Yes, I proposed a bet that was nearly a guaranteed win as long as I would be allowed to start the thread.

100 karma to Matt Stevenson for pointing that out.

You're assuming, of course, that you wouldn't be voted down to below -50, in which case no one wins.

I was planning on making sure the terms allowed for me to make the thread as a comment on an existing thread. I admit that giving myself favorable terms was bad for the virality of this post. The whole thing would have worked better if Jack was willing to continue discussing terms with me, and he would have been more likely to do so if the terms of the bet were actually good for him.

I might try to get people to transfer karma via a different post. Does anyone want to play flip a coin or rock paper scissors for karma?

The Donors' Comments section seems to be just a static page. Still, EY or someone else from SIAI should be able to confirm that you owe me 100 karma. The donation comment reads "CannibalSmith here. Kevin owes me 100 karma."

Should we record this on the existing wiki page for bets or make a new karma IOU page?

I also think that admins can manually adjust karma, an admin subtracting 100 karma from me and giving you 100 karma works even better.