Bryan Caplan  

A Hobbesian Thought Experiment

Inconvenient Arithmetic and En... Do HSAs Work? Singapore, Sing...

Suppose a random person is living on a desert island without hope of rescue. Call him the Initial Inhabitant, or I.I. Another random person unexpectedly washes up on shore, coughing up water. Call him the New Arrival, or N.A. While N.A. is helplessly gasping for air, what does I.I. do? Just to make the story interesting, let's suppose that N.A. is much bigger than I.I.

Thomas Hobbes' prediction, on my reading, is that I.I. will immediately pick up a rock and murder N.A.:

And therefore if any two men desire the same thing, which nevertheless they cannot both enjoy, they become enemies; and in the way to their end (which is principally their own conservation, and sometimes their delectation only) endeavour to destroy or subdue one another. And from hence it comes to pass that where an invader hath no more to fear than another man's single power, if one plant, sow, build, or possess a convenient seat, others may probably be expected to come prepared with forces united to dispossess and deprive him, not only of the fruit of his labour, but also of his life or liberty. And the invader again is in the like danger of another.

And from this diffidence of one another, there is no way for any man to secure himself so reasonable as anticipation; that is, by force, or wiles, to master the persons of all men he can so long till he see no other power great enough to endanger him: and this is no more than his own conservation requireth, and is generally allowed. (emphasis mine)

Preemptive murder may seem paranoid. But here's the Hobbesian logic: If I.I. waits for N.A. to catch his breath, N.A. will be strong enough to overpower him if he so desires. It's therefore in I.I.'s interest to kill N.A. before N.A. becomes a threat.

In my view, the Hobbesian prediction is crazy. Virtually no one alone on a desert island would choose the route of preemptive murder. Yes, it's possible that N.A. will catch his breath and then attack. But it's far more likely that N.A. will catch his breath and say, "Boy, am I glad to see you. At least I'm not alone." And I.I. will say the same thing back. Two normal humans in a Hobbesian scenario become fast friends, not mortal enemies.

Dogmatic Hobbesians won't accept my prediction, but I think reasonable people will. But here's a bigger challenge: What is the minimum revision you would have to make to my thought experiment to get a Hobbesian outcome?

Comments and Sharing

COMMENTS (41 to date)
Brian Doherty writes:

Perhaps that there is an already functioning small group of cooperators--three or more--on the island already when the big N.A. comes ashore? I can at least begin to imagine the awful Hobbesian outcome in that case.

Daniel Corradi writes:

The Prisoner's Dilemna is the classic model for the Hobbesian state of nature, so in this case you need to change the payoffs to make it a PD.

Both men are on the island, but everyone assumes that there is a benefit from cooperating- this is simply the Assurance game.

Now, suppose that both men were just bitten by the same snake on the desert island and there is only one dose of antidote. There is only one thing that both men need. Then, you have a Hobbesian state of nature.

Gabriel writes:

We might expect the Hobbesian outcome whenever, due to the structure of the interaction, I.I. doesn't expect to be able to achieve the same utility level as before (in expectations, as it were, assuming risk-neutrality).

In other words, I don't expect a Hobbesian outcome whenever participation constraints are satisfied (everyone is at least as well off together as they could be on their own).

Adam writes:

I see a difference between your scenario and the Hobbesian:

And therefore if any two men desire the same thing, which nevertheless they cannot both enjoy...

I.I. may not know for certain that "they cannot both enjoy." Perhaps he thinks that the new arrival can actually make life better, or at least not so much worse as to override the benefit of someone to talk to.

In other words, neither I.I. nor N.A. would have perfect information nor be perfectly rational. I suppose if I.I. was certain that N.A.'s arrival would mean imminent death, then then Hobbes would be right.

N. writes:

What was the last meal I.I. ate?

"Even saints can act as sinners, when they haven't had their dinners..."

grant writes:

If there's only enough food on the island to support one person, killing the N.A. then becomes a very good decision.

Tom West writes:

While there being a water supply that only barely supplies the initial inhabitant might be a sufficient pretext, I suspect that very few I1's would pick up the stone.

We're pretty heavily socially conditioned not to murder in cold blood, so I suspect that most I1's would let NA wake up and hope that somehow both of them could find water that I1 has missed, etc.

I've always been amazed that in situations where the alternative is death (famines, etc.), there is still rarely total anarchic breakdown of society. People tend to follows positive social conventions, even if it will kill them (with lots of exceptions, of course).

Nick L. writes:

Minimum revision?

If you were a psychiatrist, and you recognised NA as being an uncontrollable psychopathic patient of yours. What other choices would you have?

Or, you could be the man with six fingers, and NA could be Inigo Montoya - but that would be inconceivable!

Larry writes:

I'm not so sure a revision is required. In Guns, Germs, and Steel, the author describes real life in Papua. The people there live[d] in small villages, one in each isolated valley. If one man, walking alone on a forest path, meets another, there is one of two outcomes.

The men attempt to communicate (not guaranteed, since there are hundreds of languages there) and try to identify someone they know in common. If they succeed, all is well. Otherwise, they set upon each other and only one walks away. The presumption by each man is that the other is the scout for a raiding party that will do the usual rape and pillage number on their village if not stopped.

This doesn't perfectly fit our thought experiment, but it does have the benefit of having been tested in the field!

So I think the outcome of the experiment is that II would try to communicate with NA before rescuing him. If NA was able to establish his bona fides, he lives. Else, it's adios muchacho.

mthomas writes:

It seems that the minimum that you need to get the Hobbesian solution is a discontinuity in the sympathy gradient. For example if the person I.I. has a number tattooed on their arm from a concentration camp and the N.A. has a "Hakenkreuz" on his chest. This is an extreme example, but we can think of more subtle distinctions which would trigger the inside-outside prediction.
It seems that the island example maximizes rather than minimizes this sympathy connection between individuals, therefore we are seeing people not as who we would given the abundance of people Hobbes undoubtedly assumes, but the scarcity of people which might be the one case where my extreme example would suggest a possibility that the I.I. and N.A. eventually would get along.

ChrisA writes:

As Tom West says we are socially conditioned (or we are genetically programmed) to cooperate. This programming means that we don't have a rational paranoid approach to other humans. In normal cases this is fine, because most other humans are also conditioned in the same way.

BTW this is why I fear the future, humans minds uploaded into computers will be able to override their social programming and begin to act "rationally" that is, according to game theory. In this case the biggest and strongest mind will attempt to destroy the others, because if he doesn't, he will be destroyed.

So the small change I would make to Bryan's experiment (to make it Hobbesian) is to have it clear to the II that the NA will not be socially or genetically conditioned to cooperate, perhaps by having the NA look like a dangerous species (say a tiger).

Peter writes:

What is the minimum revision you would have to make to my thought experiment to get a Hobbesian outcome?

Kelly Brook?

Unit writes:

"if any two men desire the same thing, which nevertheless they cannot both enjoy"

I thought there are substitutes for everything.

The Sheep Nazi writes:

So there are. Caplan's hairy pelt looks like it would make a reasonable substitute for roofing material, and so I would bash him over the head and skin him out with a piece of sharp flint the first chance I got. Faith in humanity is for suckers.

Matt C writes:

One castway who finds another castaway on "his" beach probably wants to make friends.

One band of castaways who finds a different group of castaways on "their" shore, I think you have a very good chance of a war.

Matt writes:

Who do you know who is a self-styled "dogmatic Hobbesian"? This sounds like a straw-man argument to me.

Regardless, you did not correctly interpret the logical force of Hobbes' argument and then you employed this misinterpretation to show that Hobbes argument leads to untenable conclusions (which it most likely does, but this isn't proven by your hypothetical).

Here is what Hobbes said: If A *and* B, then C.

A=two men desire the same thing.
B=that "thing" cannot be shared (in economic terms, it has the quality of being a purely rival good).
C=they become enemies.

Point of contention: Both condition A and B are not met in your hypothetical. Maybe you meant that A and B were implied by the "desert island" stipulation of your hypothetical, but that is ambiguous at best.

Perhaps you have some interesting free-market point here that involves behavioral economics, but it gets lost in your confounding hypothetical.

And by the way, I love your book.

ryan writes:

I wonder if this scenario really is what Hobbes is talking about. When we say it seems unreasonable to suppose that II is so afraid of NA that II kills NA if he has the chance, I think we're imposing the sort of social conditioning that the state of nature by definition precludes. Wouldn't Hobbes say that it's only because we don't live in the Hobbesian state of nature, and we don't think many other people do either, that we believe the probability of II being killed is so low?

So the minimum revision is that II hasn't lived in a society, or that he thinks that relatively few other people have. (Note it can still be rational to kill NA if he thinks NA will probably not kill him -- presumably the utility gain of NA's guaranteed help is less than the utility of not being killed.)

Jason Briggeman writes:

Bryan, in your story you do not have any "same thing" that both men desire -- and that is the key element in Hobbes's argument! In fact, by specifying "no hope of rescue", you go out of your way to diminish the one obvious desire that readers would impute to men stranded on a desert island. Yes, a hopeless situation is ideal -- but in Hobbes's argument, the role of hope is paramount.

Matt writes:

I should also point out that if you want to apply Hobbes' argument to your hypothetical, then you *must* assume A and B to be true. And indeed, if A and B are true, then the conclusion becomes tenable! For example, if there is only enough food for one man to survive, and if self-preservation is the highest virtues (which I am sure it is for some of your Randian readers), then of course it is reasonable for II to kill NA!

To be sure, if we negate either A *or* B, then we cannot reasonably surmise about the truthfulness of C (which it seems like you are doing when you inquire about "minimal revisions").

And also, we should never "assume away" hope! =)

Snark writes:

A Hobbesian outcome would prevail if N.A. were to enslave or subordinate I.I., resulting in a sort of "reverse" Robinson Crusoe scenario. Under these circumstances, the economic value of labor over capital represents the optimal state.

Billy writes:

If NA is wearing a prison uniform.

Humans are naturally social. We were social before we became humans, as humans evolved from social apes. Thus, Hobbes was utterly, completely wrong. The only person more wrong than Hobbes was Rousseau.

The minimum revision would be that one or the other would have to be a sociopath. A sociopath might take the Hobbesian approach -- if he can't think of a way to exploit the person.

Billy Jean writes:
Humans are naturally social. We were social before we became humans, as humans evolved from social apes. Thus, Hobbes was utterly, completely wrong

I don't know about that. It seems absurd now, but consider a place where there's no infrastructure and limited food, like, say, Africa.

Granted, it's not a war of all against all, but at the least it's of some against some.

Sorge L. Diaz writes:

Now, suppose Initial Inhabitant lives in the island with his girlfriend....

jurisnaturalist writes:

When I.I. washed up on the shore he murdered its only inhabitant.
Reminds me of the story of a boy with some marbles and a girls with some candy. They agreed to a total trade, all the marbles for all the candy. But before the deal went down the boy slid a few marbles in a pocket. He then laid awake all night, whilst she slept, wondering if she had given him all the candy.
If empathy is a strong motivator, we assume about others what is true of ourselves.

Carl Shulman writes:

Sorge Diaz beat me to it on variation in male sexual and reproductive of success as a source of conflict. "Kill the men, take the women and children as slaves" was standard military operating procedure for many cultures.

Eric writes:

Humans need food. Humans need company. II has a utility curve for food and a utility curve for company. In the absence of possible cannibalism, If the arrival of NA causes a net loss of total utility for II, II will kill NA. On the other hand, if the arrival of NA causes a net increase in total utility for II, II will welcome NA. Dr. Caplan, it is difficult for you to envision a Hobbesian outcome because you have reconciled that II's utility for company is greater than the loss of utility he gets from sharing the island's resources. Other respectable commentators on this thread have assumed the same, stating that "we" are socially conditioned not to kill, or that any inhabitant would want a friend. These presumptions reflect our own utility curves, derived from life experiences, temperment, conditioning or lack thereof, etc. It is a logical and empirical fallacy to think all individuals share the same utility curves. Clearly II's utility curves of company and resources are unique to II. The minimum change necessary is simply to assume, for whatever reason, that II experiences a net loss in utility upon knowing of NA's arrival.

Steve Sailer writes:

On a desert island, the most likely Malthusian limitation is if there were only enough fresh water for one person.

Patrik Åkerman writes:

In addition to Larry’s mentioning of Papua I believe that your thinking on this issue should include consideration of another example, North Sentinel Island in the Indian Ocean.

“All of these contacts were cut by short by the extreme hostility the islanders show towards anyone who sets foot on their tiny island. The shipwrecked sailors and fishermen were killed with arrows before they even got off the beach.”

Even in Africa, it is group against group. Humans are a social mammal, and we need others to survive psychologically. That isn't my or Caplan's personal opinion or private need, but a proven fact. People will choose somewhat less food for company. Hobbes was, simply, wrong. His anthropology was utterly wrong.

lecoeus writes:

You basically repeat what the British empiricist John Locke said: human beings will choose to become allies, or "friends" as you said, and try to overcome the hardships together, contrary to what Hobbes said.

IMHO, it is important to see here that Hobbes talks about the natural state of the man whereas Locke considers people with a modern habitus, education and conditionings of the modern society, just like you. So you would think preemptive murder is crazy now after living in this society for years, but maybe if you were a caveman without that kind of experience, education, or values, you might go with the rule of the jungle.

So, to make this now-Lockean experiment Hobbesian, you would have to reduce the consciousnesses of II and NA to those of cavemen at the beginning of the evolutionary cycle, when human bodies were just machines to keep the genes alive.

Dr. T writes:

The minimum revision is simple:

The food and fresh water supplies are so scarce that I.I. has barely survived. To keep living, he must kill N.A. The two alternates (N.A. recovers and kills I.I. or N.A. and I.I. agree to coexist peacefully) result in I.I.'s demise and N.A.'s demise (because if the bigger N.A. kills I.I., he won't have enough food and water and will slowly waste away).

In this scenario, killing N.A. is both logical and compassionate. Logical, because I.I. lives and compassionate because the immediate killing saves N.A. from a slow death from dehydration and/or starvation.

I make no such assumption. Chimpanzees and bonobos, our two closest relatives, are social. Chimpanzees do conduct wars and raids, but they also have rituals which allow new members to join a troop. Bonobos are almost completely nonviolent toward each other. Humans are between the two, being more accepting than chimpanzees, but also willing to war, raid, etc., unlike bonobos. Humans in their natural state are social and live in small social groups. Pre-humans also were social and lived in small social groups. There never was any such thing as the Hobbesian individual man in his natural state. Humans are a social species, and any theory of man that does not take that into consideration is utterly wrong at its base, and will come to wrong conclusions as well. There was a primatologist who once said that one chimpanzee is none. The same is true of humans. Humans are in fact even more social than the other social apes -- even in our most primitive situations. This is basic primatology, anthropology, sociobiology and evolutionary psychology. It is extremely well established that Hobbes' basic assumption about man in his natural state could not have been more wrong.

Will Luther writes:

Add a female to the mix and NA gets the rock.

Pareto writes:

Dr. Camplin, unfortunately Hobbes did not have the benefit of evolutionary psychology, though we wish he could have taken that into account. Still, the state of nature may be more applicable before a veil of ignorance (which would be much more amenable to us if Rawls didn't use maximin).

MouseJunior writes:

Make II female, and NA gets the rock. Men without fear of punishment aren't known for treating women well.

Aristotle managed to get it right without the benefit of evolutionary psychology. And when a person's ideas prove to be wrong, due to a discovery of a contradicting set of facts, you don't defend the person by saying they didn't have the benefit of that knowledge -- you abandon their ideas.

Mammon's Gladiator writes:

Evolutionary psychology? Is there irrefutable evidence demonstrating that the human need for company is not maleable? If so, then produce a citation. I for one think NA killers exist. All that is required is the right set of environmental conditions to produce an individual who would kill another, if not for a rational "Hobbesian" motive, then for a psychological one. There are individuals in Federal prisons who fit what I have described. They number more then you may first imagine.

I have already mentioned sociopaths, which make up 4% of the U.S. population (and a much, much larger percentage of the prison population). One could perhaps mention solitary monks, but they are no alone, but in communion with God. If you believe in God, then they are not alone; if you don't believe in God, then the fact that they think they are in communion with God speaks volume about our psychological need for company that we would imagine that we have some when we are alone. Humans are social. The exception doesn't negate the rule.

Mammon's Gladiator writes:

I concede that humans are naturally social. However we are both in agreement that sociability may be altered through environmental, if not inheritable, factors as evident in the instance of a sociopath. Therefore it cannot be assumed prima facie that II is sociable because II is human and all humans are social. This argument was forwarded by previous posts. I thought it erroneous and wanted to refute it by pointing out the instance of sociopathy. I apologize for my strong use of language. I did not see your previous mention of the sociopathic instance.

My wife is reading a book on sociopathy which is very interesting. As with any behavior, there is likely a number of gene involved, and there will be some combinations which will give you a spciopath no matter what, while others are more sensitive to the environment. The U.S. has a higher percentage of sociopaths than other countries -- but this may be somewhat self-selecting, since one of the features of sociopaths is risk-taking, and there is an allele dominant in Americans that makes us more likely to be risk-takers (which makes sense, if you consider why people have immigrated to America). If that allele is also one of the alleles associated with sociopathy, we would expect there to be more sociopaths in the U.S. than in other places.

Comments for this entry have been closed
Return to top