« Math is Subjunctively Objective | Main | Does Your Morality Care What You Think? »

July 25, 2008


The problem with your idea is that the risk is so low that the price the tickets could command would not pay for building and stocking the refuge. If someone, moderately fearful and wealthy, built a refuge for himself, he might be able to defray the costs and get some company this way.

The trouble seems to be that it's hard to see how the entry contract is likely to be enforceable once disaster strikes. How will you prevent the refuge owners reneging? They can simply demand ever more resources, which you'll be willing to pay, in a classic 'holdup problem'. It's not even clear the refuge as a whole would be able to stop gangs of armed intruders forcing their way in (if they found out about it), unless the refuge itself was more heavily fortified.

I am going to have to contradict the first two commenters: I think it is a great idea and it sounds like a fun time too, the conditional idea is not even necessary.

As billswift suggests, merely building and maintaining the refuges has significant cost. That cost is will be undertaken, and a meaningfully-priced market in refuge tickets emerge, only if there is a long-sustained perception of lurking existential risk.

I might be wary of the motives of any refuge entrepreneurs. "RefugeCo, a synergistic subsidiary of Apocalycorp, the world leader in bioweapons, nanotech, weather control, asteroid relocation, and discount ocean-based toxic waste disposal."

I disagree with the commenting critics in this thread. The markets might not pay for the refuges, but that's not the point. I think if the refuges were subsidized, the market trends for the tickets could still tell us useful information.

Robin, I'm not as knowledgeable about prediction markets as you, but what do you think about near aversion markets? Because I think there's a bit of a paradox about prediction markets that are used to thwart the prediction -wouldn't the market figure that out quicker than we would? (Nevertheless, I don't mind a paradoxical refuge market if it saves my ass, I'll worry about the paradox later). I think a near aversion market could help address some of these problems without the pradoxical concerns. And the near aversion payoffs could be collected even decades after the fact, if our improved empiricism/analytical ability in the future allows us to see something the market saw and paid off twenty years prior: that a catastrophic risk almost materialized but was narrowly averted. It could allow groups of individuals autonomously both to bet that a catastrophic risk will be narrowly averted, and to engage in actions that facilitate the catastrophic risk being narrowly averted, without even realizing as individuals that they're doing the latter. It might be a form of algorithmic group behavior reward. This idea, if useful, needs to be made more coherent, but what do you think?

Why isn't anyone worrying about how to ensure that, conditional on a near-existential disaster happening with some people in shelters surviving, the resulting society will be of such a kind as to make long-term good outcomes probable?

Bill, I put an added on your point.

Behemoth, refuges should indeed be hidden and fortified, and run by people who depend on each other to keep their promises.

Hopefully, I don't yet see a concrete procedure to acheive what you suggest.

Steven, I'm open to suggestions.

To what degree could we get some of this information from better analysis of existing markets, such as housing markets? This question shouldn't be read as a vote against creating purer markets to measure catastrophic risk, such as refuge markets.

Also, how about the construction of refuges for their own sakes, not tied to markets? Something like the doomsday seedbank, but with people.

I could even envision a scenario in the future when we're wealthier or more efficient as a global society, where everyone has their designated cryonics unit they can evacuate to, while a subset of the population lives in mine communities to wait out the catastrophic threat and revive everybody.

Incidentally, this is similar to the "uncomfortable question" a blogger (de woolf of depressed metabolism?) asked recently, about whether many of us would be better off getting cryonically suspended now, in terms of minimizing information theoretic death odds until the living human condition is substantially safer. The idea is you're less likely to perish in a car wreck or die unnoticed and decompose in the iterim.

To finance much in the way of shelters you would need an argument that the potential benefit times the probability of a payoff exceed the construction costs. It seems like a difficult case to make to me. In the USA, the Federal Government cut off funding for the maintenance of fallout shelters decades ago - as a result of reduced perceived risk.

If AI and/or nanotechnology cause problems for most humans, they are relatively unlikely to be the kind of problems you can shelter from underground. More likely, if you take refuge in an underground bunker, such problems will follow you down there soon enough.

I'm reminded of the story about the group who foresaw WWII coming and went to a tropical island paradise to wait it out -- Guadacanal.

The shelters would need to be stocked for long enough, and there would need to be some method for terraforming the immediate surroundings afterward. Looks kind of bleak. To use your ticket you'd have to first win it and then think there's a strong chance the catastrophe will happen within the next 6 months. How would you feel if you spent 6 months cooped up somewhere, away from your job, and then the catastrophe happened in month 7?

But the idea does have possibilities, with some modification. How about this -- each 6 months hire 30 Hollywood starlets who're willing to take 6 months off from their search for the big time, and install them in the shelter. Then 30 men would have six months to try to seduce them, knowing there were only 29 competitors.

There would probably be a much better market for that than for the chance to live 6 months underground while a catastrophe does or does not happen outside, with extreme uncertainty about what happens after the 6 months are up.

HA, what is your answer to the "uncomfortable question"? How come you haven't been frozen yet?

If we had prediction markets, this would be a good idea. Since we don't, this kind of brainstorming about what we could do with them doesn't seem to me more useful now than waiting for them. What is the second-order effect of an essay like this? RH probably does it because it's good for his career. That is probably good for the creation of prediction markets. But it may push him along the "colorful character" axis, which may be bad for his ability to promote them. Having extremely colorful characters may be useful for making other proponents look normal. Which force wins?

I reckon the neo-catastrophists will probably join forces with the global warming folks, and the GM naysayers and turn into a general technological foot-dragging movement - along neo-luddite lines.

It's a bit surreal to find that Hanson, Bostrom, Sandberg, and Yudkowsky are attending this conference, instead of the usual eco-disaster suspects - Lovelock, Joy and McKibben.

Douglas, most prediction markets are limited by legal barriers, so it seems worth pointing out an interesting version that seems not.

Tim, fear of catastrophe need not imply one wants slower tech growth.

I wasn't implying that fear of catastrophe necessarily means that one wants slower tech growth.

However, I do think there is an association. It's as though the luddites are looking around for excuses to stick pointy things into the wheels of industry - and eco-catastrophes are a good excuse to regulate resources, impose carbon taxes, regulate genetic tampering - and generally go for the brakes.

Such actions typically benefit them at the expense of more technical types - and there is a management issue surrounding whether these folks get much of a say in what happens - e.g. DARPA vs some luddite votes.

I'm reading a few misconceptions (I think) about Robin's idea. I suspect I get it, so I'll try to put it in other words in the hope of communicating it.

First, the idea that the tickets are to be exercised shortly before or after the apocalypse. That's what I thought I had read too the first time. But IIUC, the plan is that you as the ticket-holder go to the refuge at a pre-assigned time which might have nothing to do with the apocalypse. You stay there for 6 months. If during those 6 months, kaboom, you are likely to be safe or at least safer. Not because it's the written rule, but because there you are already sitting in the shelter.

When you go in, it's peacetime, barring unfortunate co-incidence. Maybe a bit tense in the world, but there are no armed men telling you your ticket means nothing, as there would be on The Day After.

The other thing was the idea that it was supposed to be self-funding, not much different than just buying timeshares in a commercial refuge (which nobody does AFAICT). I assumed that at first reading, but of course that wouldn't work. It has to be funded, at least to the point where there's real bidding on the subsidized tickets.

Of course the usual logic applies - if best knowledge thinks apocalypse is nigh, price goes up, cheap talk can't keep it down. Monitor the price, acquire the knowledge.

I think there are some subtler concerns: The price would vary according to other factors, making the measurement almost worthless. Some things that jump to mind: Changes in the quality of accomodations. Changes in the quality of defences. The rhthym of restocking the refuge. Timely outside events that people do not want to miss (Christmas, perhaps) or do want to miss (winter?). Well-regarded personnel entering or leaving the refuge's employ. This list could go on but I think the point is made.

There's another way in which it could predict the wrong thing: if other refuges open or close, that affects the supply which affects the price. You could say the price is reacting to
the real lesser / greater danger that people face with more / fewer refuges available, but that's not what we were trying to measure.

PS: I'm testing whether it was my link to my futarchy group that caused typepad to flag my posts as "Potential spam" - and I'm annoyed at typepad for that.

Unkown, good question. Here's my best answer:
I suppose I'm not convinced at a gut level of proof-in-concept. It's a no-brainer that cryonics seems to maximize my reanimation odds once I'm dead, but before then it becomes less clear.
So I don't think it's clear yet whether cryonic suspension is better than trying one's best to wait out improvements in regenerative medicine and safety-proofing out living environment. But I think more analysis on this is useful. I could see myself being persuaded and choosing cryonic suspension while still alive and healthy. But I'm not there (possibly because the critical thought hasn't been put into a careful, comparative analysis of the two options yet).

Robin Hanson,
I kind of understood that this was about avoiding regulation and meant to mention it, but it wasn't really clear to me and you should take that as a comment on clarity. The aspect of using prediction markets for large disasters swamped it in my mind.

Historical market data surely already exists for sales of personal nuke shelters and survivalist-themed periodicals.

Tom, yes refuge tickets are probably not self-funding, though who knows. Yes of course prices would vary with other factors, but noisy estimates are better than no estimates. On timing, there is a tradeoff between long periods to protect from long-lead pandemics and short periods to track events nearer to doomsday. Multiple refuges with staggered entry times could mitigate this.

Julian, shelter construction prices are very long delay estimates - refuge tickets let you get closer in time to doomsday.

Robin, the sort of fallout shelters you dig into your garden aren't especially long-term (cf. the Cuban missile crisis). Also, you could probably do a trend analysis of canned food sales (and canning equipment), gun sales, survivalist magazine circulation, first-aid kit and antibiotic sales, the gold spot price, etc, and get a good short term estimate. In fact, there will be existing market data covering a range of time horizons (imprecisely, but them's the breaks) given the particular product/service and whether the contract is immediate or for a time in the future.

My point: you probably have plenty of data already to chew on!

I have never heard a plausible argument explaining why trading in conditional sports tickets is not in violation of the wire act or UIGEA or state laws. Has anyone? The tickets are things of value and the participants are at risk. Best that I can tell, the FBI just has better things to do.. like preventing disasters.

In some scenarios, those refuge ticket prices are going to track with handgun prices.

Also, the "uncomfortable question" isn't much of a challenge unless you assign a very high probability to successful re-animation (and don't care much about your contemporaries).

J Thomas:
But the idea does have possibilities, with some modification. How about this -- each 6 months hire 30 Hollywood starlets who're willing to take 6 months off from their search for the big time, and install them in the shelter. Then 30 men would have six months to try to seduce them, knowing there were only 29 competitors.

I'm not entirely sure if I understand what's meant here, but it gave me the idea of a series of shelters that were funded by a reality TV series. After all, we already have reality shows where the participants are confined into an isolated environment for a long time, so why not do it somewhere that's isolated in reality? You could then keep the shelter constantly occupied, with a new show starting each time the old one ended.

Of course, you couldn't tie this into prediction markets, as the prices would reflect people's willingness to be on TV, not their estimate of an existential risk being near. It's also questionable whether a competitive environment such as the one in most reality shows would be a very good thing in a community that might end up needing to rebuild humanity. Most "isolated environment" reality shows also, AFAIK, rely on a format where contestants are gradually eliminated one by one, bringing the late-season population to a non-sustainable level - but that could be fixed by coming up with a better format. As an additional plus, the tasks given to the contestants could be one that trained the skills they'd need in case of a collapse.

Kaj, this just seemed like an implausible sort of lottery. You invest your money, and if your ticket wins you get to sit in a shelter and get the dubious advantages of living through a giant catastrophe assuming the catastrophe happens in one particular six month period. You put your life on hold for this.

If the catastrophe does come then you find out how well the shelter is prepared to survive it. Not something that can be properly tested ahead of time though if you know precisely what kind of catastrophe to expect you can make some reasonable guesses.

If you have the funds you can outbid everybody else two or three or four times in a row. Then if you're pretty sure the catastrophe will happen within the next two years, you win. You will be in the shelter when it happens. If you miscalculate you might not have the funds available to win the fifth time, and then after two years of sitting in the shelter waiting for the end, you get thrown out -- maybe a week before the catastrophe actually comes.

Something about this behavior just doesn't fit human nature as I observe it. I could more easily see sending your *children* to a shelter for 6 months. One that's stocked with great teachers and the accumulated knowledge of the ages. I could see sending a son there when times look a bit perilous. I'd hesitate to send a daughter if too many other girls have come back pregnant, though.

J, it should be obvious that in equilibrium only a very tiny fraction of humanity would hold tickets at any one time. So of course tickets would not seem a good deal to most people. But we just need a few actual ticket holders to create a market and get market prices to inform us about how to avoid disaster.

Robin Hanson, we could just as easily figure out just how to avoid the vampire threat with a few ticket holders that way. We could have a retreat that's stocked with lots of garlic and stakes and silver bullets, and you have to show your reflection in a mirror before the security guards let you in....

Why would one market predict better than the other?

I don't get what the resort is for.

Quite a lot of Robin Hanson's posts have this kind of feel to them:

1) Find complex problem
2) Wave hands and cast "markets" at problem (see also: "emergence", "complexity")
3) Information appears from nowhere, problem is solved.

Where would the accurate information about the probability of catastrophe upon which the ticket buyers are acting come from? If people watch the news and are more afraid of a pandemic of brain tumours caused by cellphone use, that doesn't seem to tell us anything about the real likelihood of such a pandemic, or ways to avoid it...

Si, why should I summarize the concept of and evidence for prediction markets on each and every post that mentions them? I've already written lots on that subject elsewhere.

Maybe have a handy link to a FAQ about it? People who aren't completely clear on the concept might not understand why prediction markets inevitably summarise all the best information, discard the common disinformation, and weed out the misinformed-but-confident gamblers.

A quick link to the FAQ each time would be even easier than a 2-line comment saying why you won't explain.

Robin, point made, I apologise for criticising without first learning enough to know if my criticism has any merit. (While trying to be aware of the bias of learning overly specific lessons).

The comments to this entry are closed.

Less Wrong (sister site)

May 2009

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30