Cobra effect 226 points by shawndumas on July 8, 2011 | hide | past | web | favorite | 67 comments

 Nice link.TL;DR version - Cobras are a problem, pay a bounty for dead cobras, people start farming cobras to get the bounty. Kill of the bounty and at the end of the exercise you have more cobras than before.If you do systems analysis, you see this as optimizing the wrong variable. What they wanted were "fewer cobras" but they expressed by asking for "cobra corpses."The conundrum is identical to the logical fallacy of attempting to prove a negative statement. You can't articulate or measure a 'cobra void' so its hard to reward for it or incentivize it.In such systems you can either adapt the system to be unaffected by the undesirable trait (hard to do with cobras, there is only so much anti-venom to go around) Or to establish negative incentives around the presence of cobras (say fining people where a cobra is found on their property).The goal of course is understanding what is the desired state of the system and targeting an input that only amplifies that state and not some other state.
 What do you mean by "the logical fallacy of attempting to prove a negative statement?". I can prove the negative statement "I did not assassinate JFK" by citing the fact that I wasn't born yet.You're right in that in this particular case, the practical difficulties of quantifying "less cobra-ness" or "cobra void" as you call it is difficult, but there are plenty of cases where quantifying the absence of something is straightforward.
 Strictly speaking, you're proving the positive assertion that you were born after the assassination, which is incompatible with your having been the assassin. You can only prove or disprove a positive assertion. "Proving a negative" is done by proving a positive assertion that's contrary to, or incompatible with the thing you're trying to "disprove".
 If proving the positive assertion implies the negative assertion, what's the difference?
 I do not believe in the existence of fire-breathing dragons outside of fiction, but I can't prove they don't exist. See also: UFOs, angels, god(s), and so on. In logic, the onus is on the person making the positive claim (that such things do exist) to furnish the evidence to back it up. The skeptic's lack of knowledge is not proof of non-existence, of course.
 I'm talking about mathematical truths for which establishing non-existence follows directly from proving a positive assertion, not that white ravens don't exist.
 I understand that, but 'proving a negative' is philosophical shorthand for the latter rather than the former. You are just going to have to put up with the slight vagueness of the term.
 I imagine he was thinking of the problem of proving a negative general statement, e.g. "nobody has assassinated anybody". It's trivial to prove the statement is false (if indeed it is false), but it's nigh impossible to prove that it's true.The more controversial wording of this is that you can't prove there is no god.
 What people usually mean when they say this is that it is impossible to prove a global proposition about the real world. I don't know when the irrelevant word "negative" was added (after all, every "positive" statement can be trivially re-written in a "negative" form), but it's become very popular for some reason.For example: "There are no orange swans" is equivalent to "All swans are either colorless or colored other than orange." Similar techniques can be applied to all statements.
 People keep saying "prove a negative" when they mean "prove a universal". As if somehow existence statements were more positive than universal statements...
 What they wanted was "less damage and deaths by cobra."
 The idea is not to test for a single void but to test for two related voids..Offer a natural enemy to the cobra..a mongoose..
 Establishing a penalty for having a cobra appear on your property adds the property 'natural enemy of Cobras' to the existing population of humans which can be undone without having to hunt down and kill off yet another element of the system :-)
 This reminds me of that Simpson's episode "Bart the Mother". The town gets overrun with Bolivian tree lizards:Skinner: Well, I was wrong; the lizards are a godsend.Lisa: But isn't that a bit short-sighted? What happens when we're overrun by lizards?Skinner: No problem. We simply release wave after wave of Chinese needle snakes. They'll wipe out the lizards.Lisa: But aren't the snakes even worse?Skinner: Yes, but we're prepared for that. We've lined up a fabulous type of gorilla that thrives on snake meat.Lisa: But then we're stuck with gorillas!Skinner: No, that's the beautiful part. When wintertime rolls around, the gorillas simply freeze to death.
 Reminds me of:"The curious task of economics is to demonstrate to men how little they really know about what they imagine they can design." — Friedrich von Hayek
 But lizards are cold-blooded. Wouldn't they just freeze to death in the winter as well?
 Not as humorously.
 There are plenty of cold-blooded species that live in places that freeze in the winter, notably including all plants. So somehow your logic has a flaw.
 Yeah, that's a classic. Always reminds me of the Sugar Barons of Hawaii importing Asian mongooses to handle their rat problem; sadly, the rats are nocturnal, and the mongooses are diurnal. Also tasty are groundnesting birds. Oops.
 First thing in my head too
 In Admiral Gallery's book "Clear the Decks!", he describes how as commander of the Fleet Air Base in Reykjavík, Iceland he put a 0.25 bounty per rat caught. He noticed an immediate decline in rats, but noticed that he was still paying out about the same amount of money in bounties. He later discovered that his personnel were paying people at the British and US Air Force bases a dime per rat and smuggling them into the Navy base. He decided not to crack down on it since he was cleaning up the other bases.
 When the British first started sending criminals to Australia, only 40% survived the journey. Measures such as 'increased standards' or paying more did nothing to increase this statistic.Eventually the decided to pay per prisoner delivered alive and well, instead of paying in advance. Survival rate increased to over 98%
 That anecdote doesn't pass the smell test. The article says "the 1860's". Quoth wikipedia: «The last convict ship to be sent from England, the St. Vincent, arrived in 1853». Another source[1] states that the worst death rate of the fleet was 33%, far below the 60% in your source.Makes for a great story/example though! Which I guess is how these anecdotes propagate.
 I read the same story in The Fatal Shore [1] which to my knowledge is a very well researched/factual book on early convict history (though I don't recall the success rate being initially as low as 40%, it was low though).
 Two thoughts:1. They should've reduced the reward over time. If the effort had been successful, there would've been fewer snakes, and less need to get the snake population under control. This makes economic sense event without considering this effect. If there's a million snakes, you really need someone to do something about it. If there's only 10k, who cares?2. This is why it's important to focus on the actual goal. The goal was not to kill snakes. The goal was to have fewer snakes. Killing snakes was just a means to an end. Actually, the goal wasn't even fewer snakes, it was to reduce the risk of dying from a snake bite. A successful anti-venom would've been just as much a solution as the eradication of the snakes. Of course it's a lot harder to measure, "Are you doing something which is likely to reduce the risk of me dying from a snake bite?" than "Did you kill a snake?"
 Breeding another, non-poisonous variety of snake to compete in the same environmental niche with the cobras would have likely been more effective. That's what my family does in Virginia to combat the rattlesnakes.
 If the other species is indigenous to the area that might work -- though in that case it clearly can't out-compete the undesirable species naturally.Otherwise our species track record with introducing foreign species to control a perceived pest has not been particular stellar and usually causes more problems than it solves.
 How would a non-poisonous snake compete? It will have fewer predator defenses, and hunting for its prey will be very different, possibly harder. Wouldn't the poisonous snake out-compete it, unless it ends up in a different niche?
 TANSTAAFL. Being poisonous takes effort. You need to have defenses against your own poison, spend energy producing often complex molecules, etc.Not spending energy on being poisonous means that a snake can spend it elsewhere; run a bit harder, lay more eggs, lay eggs earlier, whatever.
 Pythons and boas have been pretty successful. Realize that venom is not a silver bullet; have you ever noticed venomous animals very often pursue very small prey?Venom has a place and a use, but just because you've got venom doesn't mean you're a superior predator. In fact, it seems to me it is most often used simply to allow a delicate animal to hunt prey that can hurt it- not because the prey is fearsome, but because the animal is vulnerable.
 A better option is to simply set the reward lower than the marginal cost of raising the snakes or other pest to begin with. So that there is a benefit to killing and turning in one that is encountered but not for a primary enterprise - the real problem is lack of both economic understanding and long term thinking on the part of bureaucrats.
 > They should've reduced the reward over time.Wouldn't this result in locals just breeding more and more and more cobras to cope with the decreasing reward?
 It could if the reward wasn't reduced fast enough. At some point, the reward should go to zero. And everyone should know in advance that it's going to go to zero at some point.
 Cobras .. Coming soon to a bitcoin analogy near you.
 It would be difficult to find the right reduction function. If the reduction is too slow, the snake-breeders become a problem. If the reduction is too fast, there is no effect at all.
 Wouldn't it be optimal to make the reward inversely proportional to the number of snakes turned in?That way as the population goes down, snakes become harder to catch, thus more valuable dead than alive. It also makes "farming" a negative pressure, since it decreases the reward per unit.
 It seems to me the guy with the hidden snake farm and the "unfortunate fires" his fellow snake farmers had might lead to some problems with that plan.
 The moment it becomes cheaper to breed snakes than to hunt down one of the last wild ones, the logical thing to do is to breed them, rather than hunt them. That applies even in the case of one person 'competing' (but in that case, that person, of course, would need to turn in only one snake every time period). With multiple farmers, they will compete by producing snakes as cheaply as possible.
 The building I work in once had a contest: whichever floor recycled the most paper in a month got an ice cream party.
 This is a nice story, but I'm not sure it really adds up.If you're breeding poisonous snakes to turn in for a reward, and the reward goes away, are you really going to release live poisonous snakes, or are you just going to kill the ones you have and not breed more?
 Perhaps the thinking was, "if we release them they will reinstate the bounty"? Alternatively, you could be correct and the story could apocryphal.
 I'm sure many people mean to kill them or they keep them for awhile as pets and then at some point they escape. Look at FL with all the boa constrictor/python problems they have.
 That might be the explanation.Are there really enough people who keep poisonous snakes as pets for this to happen? I don't think the Florida situation is analogous because of how much less dangerous constrictors are than cobras.
 It could also have been, that there was tons of prey available (and snake predators suffered too) and the wild living snakes just thrived after the bounty ended.
 This wikipedia article got quoted during discussions of how much to pay civilian casualties (caused by coalition/US forces and contractors) in Iraq and Afghanistan -- specifically, they didn't want to set compensation high enough to encourage people to intentionally get killed to provide for their families, because they definitely would. :(
 Reminds me of the story of archaeologists paying for bones. Which curiously became smaller and smaller with time.
 Sounds to be the same as "The law of unintended consequences" http://en.wikipedia.org/wiki/Unintended_consequences
 Or, even more precisely, perverse incentives.
 At its most sinister.
 Does anyone else see a possible corollary with Anti-Virus industry and virus writers?
 I dono. I have a family member (a database interface programmer) who insisted that virus writers are somehow being paid by the anti-virus companies, back in the boot sector virus days. I mean, his logic said, who would go through all that effort for free, and who benefits?
 True, and if your AV company also wrote all the viruses, your product would be the best at blocking them.
 And incentives for ethanol use lead to rainforest destruction:
 The plan back fired because the people optimized for the reward, instead of the long term goal of reducing the number of deaths, like the development of anti-venom mentioned previously.I read the article about Lean Software Development and one of the concept the stuck me is banning local optimization (http://drdobbs.com/184414744).
 See also, the iatrogenic effect, where the effects of the solution are worse than the problem.
 The term 'Cobra effect' stems from an anecdote set at the time of British rule of colonial India. The British government was concerned about the number of venomous cobra snakes. The Government therefore offered a reward for every dead snake. Initially this was a successful strategy as large numbers of snakes were killed for the reward. Eventually however the Indians began to breed cobras for the income.When this was realized the reward was canceled, but the cobra breeders set the snakes free and the wild cobras consequently multiplied. The apparent solution for the problem made the situation even worse.A similar incident occurred in Hanoi, under French colonial rule, where a program paying people a bounty for each rat pelt handed in was intended to exterminate rats. Instead, it led to the farming of rats.If the intention was to decrease the incidents of Cobras/rats causing problems to humans, then isn't Cobra effect a success?
 Short term success, medium term worsening of the problem, long term no change.
 Your company wants you to work more, so it has this "you must be in office for X hrs every day" and then they realize that people are spending 0.3x of their time in the canteen, so they decide to exclude canteen time, so the employees start playing games at their terminals. They decide to ban games - employees turn to facebook. Ban facebook, everyone becomes unproductive and grumpy. The company decides that this hrs. policy isn't working well, so decides to get rid of it. The joy stricken emloyees run wild and no work happens.
 My favorite version is during the decline of communist Russia. One example was quotas on furniture factories enforced by the pound, resulted in 2000lb sofas. Another was number of military boots made, resulting in all the leather and soles made but not assembled into an actual boot.
 They could have taxed "Cobra farmers" and introduce a nice reward for people pointing out to such farmers.
 That's the Vetinari Solution: http://lesswrong.com/lw/2ev/rationality_quotes_july_2010/29b...
 The power of human ingenuity. If there's money to be made, someone will try to make money off of it.
 "When this was realized the reward was canceled, but the cobra breeders set the snakes free and the wild cobras consequently multiplied."What if they had announced that rewards would stop being offered in a month? That would give cobra farmers a chance to cash out on their supply.
 From an economics perspective, there would have been a simple way to solve the problem. Create an imaginary market for cobra corpses. Basically the British government could have simply paid a price = 1 Rupee (or whatever max price they were willing to apy) /(total number of cobras purchased in the last Period of time) {where the period of time would obviously need to be adjusted for it to be sensible so that farmers dont just hold onto snakes and then come with many at once).The denominator could also be adjusted based on some biological factors - i.e. how long does it take to breed a cobra, what is the estimated number of cobras out there, so on and so forth.Further more there could be limits to the number of cobra that any one seller can sell in a the aformentioned period. As well as limit to the total number of cobras purchased in each period - (i.e. - we've bought out enough for now - thanks).I could go on...
 LOL but I think you're proving the point Hayek quoted above.In programming there's always another bug, in economics there's always another unintended consequence or negative externality.