Patching Prediction Markets

argumate

@voximperatoris: Also, you keep saying you don’t like prediction markets, but I’ve never seen you make a decent argument why not. Clearly, on a small and illiquid market (or one with fake money), quality will generally be poor. But that is reason to expand their use.

All prediction markets will be small and lack liquidity, as the range of things people want to predict is very broad but the number of people able or willing to get involved is very limited.

Robin Hanson often suggests that corporations should use prediction markets to predict the outcome of various internal projects, which would limit the number of participants to dozens or hundreds of people at best, require the company to provide liquidity, be confounded by all kinds of internal politics, and is really just a clumsy way of providing random bonuses to lucky employees.

Large scale real money prediction markets have to compete with all the other places people might want to invest, many predictions are very long-term, and it’s very difficult to nail down exactly how to judge a prediction, even for something as structured as a US election (eg. consider electoral college shenanigans).

Hanson’s futarchy is particularly ridiculous. Say you have a poorly designed government program like cash for clunkers to get old cars off the road and reduce pollution and CO2 emissions. The limited reach and high overhead means it’s going to have no effect or even be counterproductive, and you bet accordingly. Then the global financial crisis hits, economic growth stalls, and CO2 emissions actually decline, unexpectedly. Nothing to do with the policy you thought was bad, but how is that handled? Do you have a panel of experts to carefully go through all these issues and pronounce a verdict on each one?

There are already ways to use predictions to profit. If you think that Trump will be elected and pivot towards Russia, that has implications for energy markets and Eastern Europe and the Middle East and currencies. If you think that China will enter an economic slump you can take advantage of that. If you think that a particular movie will fail at the box office you can short the studio. If strong AI will be developed earlier or later you can invest accordingly.

People who have good ideas about the future already have plenty of options!

Prediction markets may be useful in certain fairly limited situations, but it is necessary to actually demonstrate that on a case by case basis; mostly it just seems to be the equivalent of averaging a bunch of guesses, which typically outperforms a single guess but isn’t some radical new way to structure society.

thefutureoneandall

These are all good criticisms; I’ve torn into UNU elsewhere for this sort of stuff. (Their pitch is swarm intelligence, but check out their pictures. It’s just an overhyped prediction market with low stakes, plus they’ve added some totally novel bugs like letting people spread vote between certain pairs of outcomes, but not others.)

Futarchy has some way bigger problems, though. I think the common rebuttal to the things you cited is “all our existing systems suck too, so this might still beat them”. I’m not convinced by that, but I certainly think we can avoid the argument by citing more severe problems.

One is that futarchy can’t effectively deal in long-odds or long-timescale events. Black swan reasoning is always a bitch, but wisdom of the crowds is a particularly awkward way to approach it. And handling things like “global warming with lead to > 2°C of temperature rise by 2100″ has all kinds of secondary problems where running a bet that long is basically a nonstarter no matter how you assess results.

Another, which I think dooms the project, is that futarchy is catastrophically unable to reason about its own existence. If I propose a betting topic of “the prediction market will be dissolved and replaced with a dictatorship which cancels all bets”, there’s no way to coherently bet true and get profits.

If people think too big to fail is a serious issue with the stock market, they ought to be screaming in terror at the thought of futarchy. Because the equivalent faulty-downside predictions there are things like “let’s do nuclear brinkmanship with Russia to improve our economic standing when they back down”. The failure case doesn’t pay out, so it must be a good idea!

I’m being glib here, but I’ve seen this issues raised before and I’ve never seen even an attempt at a rebuttal. If someone has addressed things like “how to prediction market about 100 year x-risk issues”, I’d love to read it.

argumate

like an assassination market, but for the entire planet.

mitigatedchaos

Perhaps payouts along the way based on the estimated probability for long-term bets - I guess that would be Prediction Derivatives? That allows people to profit if the estimate of high global warming increases. However, I’m not sure on how it incentivizes holding the prediction in the first place.

If we allow selling the shares of your position in the market, then people could hold long-term positions speculatively. For instance, I estimate a good chance that high global warming will happen, and that it will become more obvious in the intervening time, then I can buy low now and sell later.

argumate

the other issue with prediction markets is that we already have a wide range of ways to invest based on our beliefs about the future, eg. not becoming an apprentice buggy whip crafter after the invention of the automobile.

even betting on civilisation collapse can be done by stockpiling supplies or what have you.

mitigatedchaos

Or to put it another way, Argumate, Capitalism is ludicrously powerful, and I’m searching for a way to do that for government but with a different utility function.

argumate

I think the fundamental problem is this assumes that government’s role is to search the space of possible policies looking for optimal outcomes, when in fact it’s a formalised way of handling power struggles and dividing up the spoils with less bloodshed.

mitigatedchaos

I still think it can be improved to behave more like the former. Governments routinely spend over 30% of GDP in nations with GDPs in the trillions. Improving them even small amounts could yield huge dividends for human civilization, not to mention allocate the spoils in a way that makes them worth more.

Large gains are probably possible (based on the existence of states like Singapore). It wouldn’t even have to come with lower taxes. We could just buy another five hundred billion dollars in social goods. Seeing all sorts of raindance policies and wasteful policies is aggravating.

politics

collapsedsquid

argumate

the other issue with prediction markets is that we already have a wide range of ways to invest based on our beliefs about the future, eg. not becoming an apprentice buggy whip crafter after the invention of the automobile.

even betting on civilisation collapse can be done by stockpiling supplies or what have you.

mitigatedchaos

Or to put it another way, Argumate, Capitalism is ludicrously powerful, and I’m searching for a way to do that for government but with a different utility function.

argumate

I think the fundamental problem is this assumes that government’s role is to search the space of possible policies looking for optimal outcomes, when in fact it’s a formalised way of handling power struggles and dividing up the spoils with less bloodshed.

mitigatedchaos

I still think it can be improved to behave more like the former. Governments routinely spend over 30% of GDP in nations with GDPs in the trillions. Improving them even small amounts could yield huge dividends for human civilization, not to mention allocate the spoils in a way that makes them worth more.

Large gains are probably possible (based on the existence of states like Singapore). It wouldn’t even have to come with lower taxes. We could just buy another five hundred billion dollars in social goods. Seeing all sorts of raindance policies and wasteful policies is aggravating.

collapsedsquid

I think it runs afoul of so many measurement, externality, long tail, and other issues that what you’re going to get is a mess, especially if you try it on everything at once.

mitigatedchaos

It’s part of a general vein of policies that don’t have to be implemented all at once.

For instance, we could start by requiring legislators to make non-binding predictions about the legislation they pass when it’s passed, with specific, measurable outcomes, like “the annual murder rate will decline by 3-5% with 80% confidence”. No differences in pay, no firing, just a formal record to compare to. It might not sound like much, but it’s a step towards explaining the expected outcomes of massive bills in clearer language that can be verified.

Then if that works, we can expand it and start awarding compensation based on the outcomes. (Although I admittedly have been thinking about a far deeper version where percentile score in a legislative outcome prediction market is part of a formula that determines public funding of parties, with lots of delegated voting and other things.)

Other options to improve governance include automatically putting sunset provisions in bills and various regulations so that they don’t just keep accumulating and accumulating if they aren’t actually that politically important. Again, small step that can be reversed.

If people weren’t such idiots, we could get cities to volunteer to test various policies like basic income or wage subsidies before national rollouts instead of just kinda assuming they’d work because it’s morally virtuous without successively larger scales of testing.

We could probably also pay more to legislators, like they do in Singapore. It was only about $250 million or something to give all the congressmen $500k-$1M salaries, plus another $30M for the President. For a bit more, we could pay them significant pensions and forbid them from working for anyone but the government afterwards. If it saves the US economy well under 1% of GDP (0.0014%?) due to getting a better quality of legislator, it comes out to a net gain. The exact numbers here are less important than the orders of magnitude. Congress has an enormous amount of power over the economy, but they aren’t paid based on that. However, they can use that power to get wealth by converting it via corruption.

There have got to be a hundred things we could do better.

Source: argumate politics

argumate

the other issue with prediction markets is that we already have a wide range of ways to invest based on our beliefs about the future, eg. not becoming an apprentice buggy whip crafter after the invention of the automobile.

even betting on civilisation collapse can be done by stockpiling supplies or what have you.

mitigatedchaos

Or to put it another way, Argumate, Capitalism is ludicrously powerful, and I’m searching for a way to do that for government but with a different utility function.

politics

argumate

the other issue with prediction markets is that we already have a wide range of ways to invest based on our beliefs about the future, eg. not becoming an apprentice buggy whip crafter after the invention of the automobile.

even betting on civilisation collapse can be done by stockpiling supplies or what have you.

mitigatedchaos

Right, but one might want organizations to make decisions based on more accurate information, not just individuals, and measuring the rate of underground bunker construction doesn’t tell you which apocalyptic scenario is most likely - and thus needs to be most addressed.

A continent-spanning superstate controlled by a network of computers implementing the thing that comes after the thing that comes after prediction markets over a vast and inscrutable state bureaucracy physically realizing a National Utility Function, attacking enemy nations in ways they don’t even understand, its terrifying efficiency only truly understandable as a creeping horror to the very few.

But unironically.

mitigated future mitigated fiction the invisible fist

Never give someone else direct write access to your mind.

flagpost

silver-and-ivory

mitigatedchaos

The Mitigated Chaos Plan for School

@silver-and-ivory

…that’s true.
I don’t know what a good solution would look like, but it doesn’t have to involve any more high-IQ individuals than we have now, just a better distribution of resources schools already have.
I want to test solutions to the current system, and to find many different possible set-ups that are different from the one we have now. (They might not scale well, of course.)
Even improvement in a limited geographical area or to some minor aspects, for relatively affluent middle-class individuals, would be really valuable to me.

Roight, let me suggest my plan, which would only help matters that you want tangentially most likely.

Are you familiar with Spaced Repetition? It’s used in programs like Anki. The basic summary is this: your brain flags things as important by whether or not you use them, and forgets them gradually over time. Spaced repetition brings the item up again at a certain point in the forgetting, so that your brain goes “oh hey this came up again, it must be important, I better remember it!”

Gamification is also a thing, and I have a theory that a big part of why people don’t like school stuff is that it doesn’t feel applicable, or that it will ever be applicable. But while I do not enjoy math for its own sake, I feel almost no resistance to doing math when I have to in order to accomplish some other task.

I’d like @argumate to read this post, too, and probably a few of the others as well.

So here’s my proposal:

1. This will be primarily implemented as a computer program. It will be implemented on a custom computer system that is not easily compromised.

2. All textbooks will be presented in both a fuller, contextualized format, and as semi-atomic facts of information, ready for use for spaced repetition memorization.

3. Exercises will be split between grinding and synthesis. Synthesis exercises will sometimes be in the form of game-like programs that have a complex problem which the students must integrate their knowledge of the subject to perform. (That is, students must be able to take the knowledge and use it and apply it, not just repeat it.) Other times, for other subjects like English, they will be items like essays that are manually graded by teachers. Students earn resource points to attempt synthesis exercises through grinding exercises, which are the rote learning component intended to reinforce the knowledge and speed up processing (e.g. of doing math). If you fail the synthesis exercise, you may have to do more grinding to attempt it again.

4. The computer program will conduct a review of all the subjects the student needs to know, based on spaced repetition algorithms and data about the student and their previous performance. This prevents the constant information loss that is pervasive in the American school system.

5. All of this is individualized. Students go at their own pace, and graduate when it has all been completed, or are pushed out of the school system at 21.

6. Homework is mostly rare or non-existent. Instead, students will stay another hour or two at school. Homework is for doing exercises, which we are having them do at school.

7. The school day will be broken up by various social activities to let students’ brains relax in between blocks of studying, which will still be somewhat unified by subject of study to make #8 easier.

8. In addition to grading work, teachers will also act as tutors to individual students. Students will be grouped in classes with students who are in a similar position of progress within the system. Teachers will go around the room answering various questions and helping students with items they are having trouble with. There may be some small lecturing sections, maybe.

The following is less necessary, but additional depending on your balance of Nationalism/Capitalism/Technocracy/etc.

9. Students will be awarded points based on a mix of (about 1/3 each) progress, attendance, and and percentile academic standing within their school. These points can be spent on a very larger variety (over 100) of uniform parts, snacks, media, and other items at participating retailers. This has the virtue of aligning the school’s social hierarchy more closely with the desired outcome of learning & academic performance, as well as giving students practical experience with small amounts of “money”.

10. Research shows that teaching math below a certain age doesn’t actually accelerate learning progress on it much at all, so for very young students, the system will focus on “moral/social” education and socialization and potentially language skills.

policy flagpost education

What GOP Must Do to Avoid an American NHS

Alright, US Republicans, listen up.

You think your goal right now is to “Repeal Obamacare”.

But that has issues. Given your values, your goal should not just be to repeal Obamacare, but to prevent the emergence of a single-payer healthcare system in the United States, in which the US Federal government monopolizes 1/5th of the US economy in a botched attempt at recreating the NHS.

This does not mean trying to pass laws to sabotage the attempt. They’ll just get overturned, and that will be celebrated as a Democratic party victory.

You have to deflate the demand for single-payer healthcare. That is what is necessary for market-based healthcare to continue to exist in the United States. As long as the demand is there - and it is growing, as middle-class families come under increasing pressure - the desire for single-payer will re-emerge.

Yelling at people about how they don’t deserve healthcare because they haven’t worked hard enough will not help. You just picked up a bunch of rural voters that got laid off at the factories and can’t meet their deductibles. You have to focus on preserving the market-based allocation.

There is a simple way to do this.

Each year, every American receives a healthcare voucher for $2,500. (Maybe more - this is way below our per-capita healthcare spending - but even you can do this much.) You can hide this as an earned income tax credit for the poor, a tax deduction for the middle class, etc, if you have to.

The voucher can only be used to purchase valid medical services and medical insurance. We have licensing for doctors, so this shouldn’t be too hard to figure out.

Any of the voucher not spent rolls over into a no-tax/low-tax Health Savings Account, where it builds up.

Private money can also go into the account.
Tax is only assessed above a certain amount / on death.
By default, it goes to a zero-interest government account, but this can be changed by opening a suitable interest-earning HSA account at any American bank or credit union.

Hospitals are still required provide emergency care for the uninsured/poor, however, up to ½ of future vouchers can be garnished, proportional to the size of outstanding claims.
Vouchers and money from the HSA can be given to spouses or children for their HSAs, though not other relations (normally). It can also be disbursed in a will, but if so then it will be taxed as normal income.

Now, first objection is probably that this will bid up the price of medical care to the new floor. I believe that is unlikely - because the voucher accumulates and people don’t actually like randomly buying healthcare, it makes much more sense to save it up and spend it later as you normally would, or else just buy insurance.

Second objection is that people might launder the money to get it for themselves. In this case, the person they are screwing over is mostly their future self, so they have incentive not to do this.

Third objection, which I’m likely to get from my left-wing readers, is that this isn’t enough money. That’s probably true, but this is likely of more benefit than whatever the Republicans are currently cooking up. In the current situation we have some poor people buying the mandated insurance, but unable to actually get medical care because they cannot afford the deductible. Under this system, they can go get medical care tomorrow. Likewise, for pre-existing conditions, this ensures that at least the value of the voucher is available each year to pay for it.

Later administrations could raise the amount, but the benefit for the Republicans here is the preservation of the market mechanism. This is likely to be a popular program.

This could be coupled with a variety of other reforms to reduce overall healthcare costs, such as requiring hospitals to post information about their prices, success rates, etc. Don’t just cross your fingers and hope the market works. Education alleviates information asymmetry and lubricates markets. Create the right framework for suitable informed competition to take place in.

policy politics healthcare republicans flagpost

See, that’s what the app is perfect for.