Intransitive preferences are a demonstrable characteristic of human behaviour.  So why am I having such trouble coming up with real-world examples of money-pumping?

"Because I'm not smart or imaginative enough" is a perfectly plausible answer, but I've been mulling this one over on-and-off for a few months now, and I haven't come up with a single example that really captures what I consider to be the salient features of the scenario: a tangled hierarchy of preferences, and exploitation of that tangled hierarchy by an agent who cyclically trades the objects in that hierarchy, generating trade surplus on each transaction.  

It's possible that I am in fact thinking about money-pumping all wrong.  All the nearly-but-not-quite examples I came up with (amongst which were bank overdraft fees, Weight Watchers, and exploitation of addiction) had the characteristics of looking like swindles or the result of personal failings, but from the inside, money-pumping must presumably feel like a series of gratifying transactions.  We would want any cases of money-pumping we were vulnerable to.

At the moment, I have the following hypotheses for the poverty of real-world money-pumping cases:

  1. Money-pumping is prohibitively difficult.  The conditions that need to be met are too specific for an exploitative agent to find and abuse.
  2. Money-pumping is possible, but the gains on each transaction are generally so small as to not be worth it.
  3. Humans have faculties for identifying certain classes of strategy that exploit the limits of their rationality, and we tell any would-be money-pumper to piss right off, much like Pascal's Mugger.  It may be possible to money-pump wasps or horses or something.
  4. Humans have some other rationality boundary that makes them too stupid to be money-pumped, to the same effect as #3.
  5. Money-pumping is prevalent in reality, but is not obvious because money-pumping agents generate their surplus in non-pecuniary abstract forms, such as labour, time, affection, attention, status, etc.
  6. Money-pumping is prevalent in reality, but obfuscated by cognitive dissonance.  We rationalise equivalent objects in a tangled preference hierarchy as being different.
  7. Money-pumping is prevalent in reality, but obscured by cognitive phenomena such as time-preference and discounting, or underlying human aesthetic/moral tastes, (parochial equivalents of pebble-sorting), which humans convince themselves are Real Things that are Really Real, to the same effect as #6. 

Does anyone have anything to add, or any good/arguable cases of real-world money-pumping?

Casinos. The habitual gambler would prefer to have money than not, and would prefer to take the negative-value bet than not.

So one description of this situation is that what a habitual gambler is paying for is the gambling itself rather than its outcome; that is, what the gambler derives utility from is the act of gambling. In that sense there's no money-pumping involved: the gambler is just paying for a service he strongly desires.

You can always redefine pumping as non-pumping by saying that the victim derives utility from being pumped.

You can always redefine pumping as non-pumping by saying that the victim derives utility from being pumped.

And this is why you can't (necessarily) distinguish a preference from a bias with behavioral studies alone, which should weaken our faith in the results of the heuristics and biases research program (at least, slightly).

Then you aren't modeling the pumping properly; the agent is getting something out of the cycle.

Any money-pumped agent is getting surplus from each transaction. That's why they're carrying out the transaction.

In the money pump, the agent isn't extracting the surplus from the cycle, because they trade everything they get back at some point and end up with strictly less than they started with, including intangibles.

If this is a strict requirement of a money pump, (and I can see the argument for it), it seems like the bases for human intransitive preferences don't qualify as intransitive if ephemeral / insubstantial gains are treated as concrete cases of surplus.

In fact, if it is a strict requirement, it seems like the money pump is a fairly useless model unless agents literally had "hard-coded" exogenous intransitive preferences, which doesn't seem to make it much of a practical worry in AI, either.

The presence or absence of an ephemeral/insubstantial gain in a given transaction is a fact independent of any other transaction. Rational agents with well-ordered values that consider insubstantial benefits not easily viewable to an outside observer could engage in behavior where they traded money for insubstantial benefits in a manner that looked exactly like a money pump to an outside observer. They would also seek ways to acquire those insubstantial benefits in ways that cost less.

Really, though, convincing someone that you have an intangible benefit that they want enough to pay for is simply good marketing.

I've heard that for badly addicted gamblers, what they want is the trance. Getting a jackpot is an unpleasant distraction.

One description of an archetypal money-pumping situation (where an agent prefers A to B, and B to A, and keeps swapping one for the other with another agent, to whom he pays commission) is that the first agent derives utility from the act of obtaining either A or B in exchange for its counterpart, and he's paying for that service.

If utilitarianism is to make sense as a model, you can't question someone's utility. If money-pumping is to be a meaningful description of a possible scenario, it has to be structural, not just a narrative that can be ascribed to an agent's preferences.

I don't think what I've written applies to arbitrary money-pumping. An agent with incoherent preferences doesn't have a utility function.

RichardKennaway models a habitual gambler as having incoherent preferences because the gambler prefers money to not having money but prefers taking negative-EV (in terms of money) bets to taking no bets. I model a habitual gambler as having a utility function which includes a term for money but which includes a second term for, I dunno, "excitement," and a negative-EV (in terms of money) bet has positive expected utility for such a gambler because of the gain in excitement. I think this is a more accurate model, and I also don't think habitual gambling is in isolation evidence of incoherent preferences. (I'm not claiming that humans have coherent preferences, though.)

RichardKennaway models a habitual gambler as having incoherent preferences because the gambler prefers money to not having money but prefers taking negative-EV (in terms of money) bets to taking no bets. I model a habitual gambler as having a utility function which includes a term for money but which includes a second term for, I dunno, "excitement," and a negative-EV (in terms of money) bet has positive expected utility for such a gambler because of the gain in excitement. I think this is a more accurate model, and I also don't think habitual gambling is in isolation evidence of incoherent preferences. (I'm not claiming that humans have coherent preferences, though.)

Yes. I'm bothered when people criticize gambling on the ground that it's negative EV in terms of money, given that (say) going to the cinema is also negative EV in terms of money but few people criticize that on that ground. (I guess that what's going on is that those people enjoy going to the cinema but don't enjoy gambling, and don't realize other people may be different.)

That's just the metaphorical reptilian brain winning the fight for control of the gambler's body over the metaphorical neocortex. There is a nice bit of fictional evidence about it in the movie Flight, when Nicole, a drug addict, dials her dealer while simultaneously praying for him not to answer the phone. The winner is not money-pumped, but is rendered the requested service.

That's just the metaphorical reptilian brain winning the fight for control of the gambler's body over the metaphorical neocortex.

Looking at the two halves of the circle separately does not make the circle go away.

Sure it does. The utility theory metaphor for the human agent is simply wrong in lots of cases. The human agent isn't a single agent but a plurality. The study of how pluralities allocate resources and interact is economics / game theory / voting theory, not utility theory per se.

Sure it does. The utility theory metaphor for the human agent is simply wrong in lots of cases. The human agent isn't a single agent but a plurality.

That doesn't help. Suppose we have two people, one with a preference for A over B, one with a preference for B over A. If they start out with various amounts of A and B, and trade, trade will eventually stop with a distribution of A and B that no trade can improve on. No money pumping is possible.

Money-pumping through non-transitive preferences can only happen if the preferences exist within the same person. You can talk about sub-agents of a person if you like, but they are not parts of the same person in the way that two people are parts of the same society. Even if we consider those two people fighting over A and B instead of trading, this still does not correspond to what happens within a conflicted individual. In extremis, one actual person can kill the other; the gambler sliding into ruin injures all of himself. Many who suffer from an addiction might wish to cut out and destroy the unwanted motivation, or lock it in jail, or any of the other things that can be done with undesirable people, but this is not currently possible.

The utility theory metaphor for the human agent is simply wrong in lots of cases.

I would say it is a description, not a metaphor, but I agree it is simply wrong in lots of cases. Utility functions define transitive preferences only. An agent with non-transitive preferences does not have a utility function and is not described by utility theory.

You can talk about sub-agents of a person if you like, but they are not parts of the same person in the way that two people are parts of the same society.

Even if we consider those two people fighting over A and B instead of trading, this still does not correspond to what happens within a conflicted individual.

Does it make sense to talk about the USA as if it were an agent that could get money pumped? An Obama administration is very different from a Bush Jr. administration.

I think one analogy here is that the human "agent" is ran by powerful executives that rapidly get voted out of office. To the extent that voting models what's going on inside human heads, negative results on voting also apply..

Does it make sense to talk about the USA as if it were an agent

In some contexts, yes, in others, no. And at best it's an approximation.

that could get money-pumped?

In everyday language, this is called "playing off one side against the other".

I think one analogy here is that the human "agent" is ran by powerful executives that rapidly get voted out of office. To the extent that voting models what's going on inside human heads, negative results on voting also apply..

I put little value on these analogies and metaphors of people as collectives. They may be better than the idea of a mind as being a single thing without parts, but unlearning a falsehood does not mean that you now know the truth.

"All models are false."

I agree! My claim is game theory / economics / voting theory are complicated, but well-developed and informative models for human preferences. If they are wrong (as all models are..) they are wrong for more subtle reasons than utility theory. On LW people spend far too much time thinking/talking about utility theory and preferences vs these alternatives. Discourse will improve if more mindshare is moved to better models.

"All models are false."

I've never liked that slogan, not even with the completion "but some are useful". Of course, statistics is the science of how to make the best of bad data, and if bad data is all you can get, statistics is what you do. But "all models are wrong" strikes me as furnishing an easy excuse for intellectual laziness.

Rain causes mud; mud does not cause rain. Is that a "wrong model"? Matter is made of atoms; atoms are made of electrons, protons, and neutrons. Is that a "wrong model"? The structure of DNA -- a "wrong model"? Of course, if you search assiduously enough you can spy out faults in any assertion anyone makes, but to say that these are wrong in the way that a piece of statistical curve-fitting is "wrong" is, to borrow Isaac Asimov's quip, wronger than both of them put together.

If I visit my doctor with an illness, what I really want is for the actual disease process to be known, and an intervention that is known to actually fix the problem, in the same way that a car mechanic can find out what is wrong with a car and fix it by understanding what needs to be done and why. True models, that is. In the current state of the art in medicine, there aren't very many. "All models are false" is a counsel of despair.

Perhaps Box was only talking about statistical models. But we aren't.

But "all models are wrong" strikes me as furnishing an easy excuse for intellectual laziness.

Ok.

My point (I think we were talking about utilities), is that phenomena well known here on LW such as observed lack of transitive preferences, akrasia, precommitments, and so on, can be usefully viewed via a plurality model of human agency. It's fine if you don't like this model, but then you should like the utility function model even less (as its even less realistic).

but then you should like the utility function model even less (as its even less realistic).

I certainly do.

That's addiction. It's not addiction to losing money, it's addiction to playing the games.

That's addiction. It's not addiction to losing money, it's addiction to playing the games.

Arising from non-transitive preferences.

Source? Addictions don't always or even typically arise or result in non-transitive preferences.

People like my mother (who occasionally go to the casino with $40 in their pocket, betting it all in 5-cent slot machines a nickel at a time, then taking back whatever she gets back) go to the casino in order to have fun/relax, and playing casino games is an enjoyable past time to them. Thus while they lose money, they acknowledge that it is more likely than not that it will happen, and are not distressed when they leave with less money than they enter with because their goal was to enjoy themselves, not to end up with more money - getting more money is just a side benefit, something that happens sometimes (about one time in four that she goes, she winds up with more money than she entered with) but which is not really the primary purpose.

Ergo calling it a money pump in such cases is a bit silly.

On the other hand, people who genuinely believe they can win money at the lottery/gambling (against the house; it is not irrational to play poker or blackjack with the idea that you can win, IF you know what you're doing) are in fact engaging in money pumping activities.

But it really depends on the nature of the person involved as to whether or not it is a true money pump.

On the other hand, people who genuinely believe they can win money at the lottery/gambling (against the house; it is not irrational to play poker or blackjack with the idea that you can win, IF you know what you're doing) are in fact engaging in money pumping activities.

The only people who I know that do this, do it due to beliefs about their own capability of pre-cognition, not due to incoherent beliefs. There was a non-mistaken variation where someone reverse engineered the PRNG in the slot machines to win.

On the other hand, people who genuinely believe they can win money at the lottery/gambling (against the house; it is not irrational to play poker or blackjack with the idea that you can win, IF you know what you're doing) are in fact engaging in money pumping activities.

That doesn't strike me as money pumping so much as having false beliefs.

When I go into the pinball arcade, I would prefer to have more money than less (all other things being equal), but will still put $20 value-signifier of metal into machines in order to have fun.

I happen to have more fun playing pinball than playing casino games, but the principle is the same.

Have casinos ever tried to sell "gambler's insurance" that pays out some amount of money if you lose too much at the casino?

This doesn't seem like money pumping to me. A money pump, as I understand it, means that you exchange money for A. Then you exchange A+money for B. Then you exchange money+B for A. Habitual gamblers just exchange money for the chance to win more money. The gamble that they take is never taken back by the casino.

It took me about five minutes to get there, but I'm actually kind of convinced by this example, and really surprised I didn't think of it since I have a massive dislike of institutionalised gambling.

Shops. The habitual buyer would prefer to have money than not, and would prefer to purchase goods than not.

Compulsive shoppers, yes. Ordinary people, no. Ordinary people simply prefer the goods they are buying over the money they are paying. They do not immediately sell the goods at a loss and then repeat the cycle. Preferring A to B does not create a money pump. Preferring A to B and B to A, that is a money pump.

Preferring A to B does not create a money pump. Preferring A to B and B to A, that is a money pump.

Precisely my point.

[joke]
Toll roads. People pay to drive along them in one direction, and then pay again to drive the other way, back to where they came from!
[/joke]

Not what you're looking for, but:

I didn't know him very well personally, but there was somebody in my circle of acquaintances who I heard was in the habit of doing things like buying a new gaming console and then realizing that he didn't have the money to pay his rent because of that. Then he'd sell it to someone for substantially less than the original price, so that he'd get at least some money quickly. Then he'd repeat essentially the same process a few months later.

Apparently his "friends" liked him because they'd get cheap stuff from him. I imagine that he would have been very easy to provoke into buying something (and thus selling it off later), but if anybody I knew was doing that intentionally, they never admitted it to me.

Some of these examples are not examples of "money pumping" but of something called "trade," or possibly even "arbitrage." Arbitrage is not quite the same as money pumping.

Money pumping isn't easy to find because people don't value consistency very highly and will change behavior in iterative games. It's difficult to precompute transitive preferences, but easy to change on the fly if burned.

Most of the examples I came up with, when I tried to formalise them, ended up being either rent-seeking or arbitrage.

My approach for finding money-pumping candidates seemed to be "in what ways are people scammed over and over again?" I've come to believe this approach is flawed, and my intuitions for what money-pumping would actually look like are off-base.

Given the dynamic nature of human preferences, it may be that the best one can do is n-fold money pumps, for low values of n. Here, one exploits some intransitive preferences n times before the intransitive loop is discovered and remedied, leaving another or a new vulnerability. Even if there may never be a single time that the agent you are exploiting is VNM-rational, its volatility by appropriate utility perturbations will suffice to keep money pumping in line. This mirrors the security that quantum encryption offers: even if you manage to exploit it, the receiving party will be aware of your receipt of the communication, and will promptly change their strategies. All of this assumes a meta-level economical injunction that states if you notice intransitivity in your preferences, you will eventually be forced to adjust (or be depleted of all relevant resources).

In light of this, it may be that exploiting money pumps is not viable for any agent without sufficient amounts of computational power. It takes computational (and usually physical) resources to discover intransitive preferences, and if the cost of expending these resources is greater than the expected gain of an n-fold money pump, the victim agent cannot be effectively money pumped.

As such, money pumping may be a dance of computational power: the exploiting agent to compute deviations from a linear ordering, and the victim agent to compute adherence thereto. It is an open question as to which side has the easier task in the case of humans. (Of course, a malevolent AI would probably have enough resources to find and exploit preference loops far quicker than you would have time to notice and correct them. On the other hand, with that many resources, there may be more effective ways to get the upper hand.)

Finally, there is also the issue of volume. A typical human may perform only a few thousand preference transactions in a day, whereas it may take many orders of magnitude more to exploit this kind of VNM-irrationality given dynamical adjustment. (I can see formalizations of this that allow simulation and finer analysis, and dare I say an economics master's thesis?)

And another one: If money-pumping is successful enough, the victim goes broke, and no more money pumping. Eventually all potential victims go broke or wise up.

This seems to happen. Penny auctions like Swoopo have been called "as close to pure, distilled evil in a business plan as I've ever seen", but where the data allows, people seem to gradually realize how they're being pumped and attrit away (see http://www.gwern.net/Sunk%20cost#fn35 )

It probably happens more often when the different stages of the pump are in different units. Fictional example: the archetypal Russian swindler used to sell the sandwiches his schoolmates freely shared with him back to them when they were hungry. Both legs of the transaction appear beneficial to the victim: feel-good sharing + hunger-quenching buy-back.

Neglecting time inconsistency, the net effect of that is that the other party gives money in exchange for feeling good, i.e. what anyone giving money to a beggar does.

Nick Beckstead has this to say in his dissertation on decision theory and x-risk:

we never hear about people who get money pumped. Why? One possibility is that people never get offered these trades that would trigger money pumps. A more plausible answer is that people do not act on their preferences in the inexible way this argument assumes. When they get into a situation where they see that their preferences would lead them to get money pumped, they either change their preferences or refuse to continue to act on some of those preferences. Because of this, money pump arguments do not illustrate a practical danger for humans. It is plausible that having preferences which would be theoretically susceptible to a money pump displays a failure of perfect rationality, but, once again, that a certain approach is imperfect does not imply that an improved approach is meaningfully available.

It's possible that many people can detect many forms of money pump and wish to avoid them more than they wish to engage in the individual choices which form the pump.

Very interesting! I actually started having similar thoughts about money pumps and utility functions after learning Haskell. Specifically, that you can avoid the intransitivity -> money-pumpable implication if you just assume (quite reasonably) that humans' utility functions are lazily evaluated and have side effects (i.e. are impure functions).

In other words, humans don't instantly know the implication of their utility function for every possible decision (which would imply logical omniscience), but rather, evaluate it only as the need arises; and once they evaluate it for a given input, the function can change because it was so evaluated so that it has a different I/O mapping on future evaluations (the impure part).

EY has actually said as much about morality and human values, but used the term abstract idealized dynamic.

Anyone know how badly (or if at all) the standard implications of the VNM utility axioms break down if you take away the requirement that the utility function must be strictly evaluated and pure?

Edit: Do you have a cite for that quote? I google it and only get your post.

Beckstead's dissertation isn't online yet, and he asked me not to upload it.

Thanks for sharing the connections between human utility functions and programming functions.

Other works on that subject are Muehlhauser (2012) and Nielsen & Jensen (2004), both of which I cited in IEME, and also Srivastava & Schrater (2012), which was recently brought to my attention by Jacob Steinhardt.

Economists have pointed out that technical functions (i.e. the functions which yield the "outputs" for any given resource inputs and production techniques) are also explored lazily, as it were. It's quite likely that the existing literature on machine learning and search theory has extensively considered the implications of such exploration on the resulting behavior.

I have a hypothesis: being offered a bet that looks pumpable is generally sufficient evidence that the bet is losing, and the few people who still bet do so due to extraneous beliefs (e.g. a belief they are slightly psychic or can otherwise influence the chances; one can imagine a quantum and anthropic reasoning - confused person believe that by expecting very hard the future where he wins or committing to spending more time sleeping if he loses he can affect his subjective probability. One could also deduce the workings of PRNG in the slot machine, and then win, or lose if they're wrong about the PRNG. One could believe their subconscious can deduce PRNG, then be unlucky to confirm this at some confidence level).

I understand that many people with a regular monthly salary take very-high-interest paycheck loans, every month, over long periods of time.

I.e., they are not just desperate a few times and showing a strong time preference. Rather, every month for decades, they timeshift their salary back two weeks with paycheck loans. They should scrimp a little for a few months, and just shift consumption forward two weeks, but over decades, they never do that.

So, looking at this in the steady state, they can't even be said to have a high discount rate.

And what about people who buy a lottery ticket or gamble repeatedly; and when they win, re-gamble their winnings.

They could easily be said to have a high discount rate, and it's consistent. They give a finite value (sum of an infinite series of discounted amounts) to the infinite amount of money they lose to get more two weeks early, and it's always less than the added value of getting it two weeks early this month. I think many of them don't actually discount that much, they just judge their expenses more by urgency than by importance.

Then, when every last cent
Of their money was spent
The Fix-It-Up Chappie packed up.
And he went.
And he laughed as he drove
In his car up the beach,
“They never will learn. No.
You can’t Teach a Sneetch!”

- The Sneetches by Dr. Seuss.

a tangled hierarchy of preferences, and exploitation of that tangled hierarchy by an agent who cyclically trades the objects in that hierarchy, generating trade surplus on each transaction.

I used to be a used and rare bookseller. I knew what was desired and who desired it. Knowing one or both of those things I could go into one bookstore, buy a book at the price that store set, then sell it at another bookstore at a profit. I could even sell a book for credit, get a book and sell it at a third store for more profit. When I added the resources of the Internet I did even better. Then thousands of other booksellers added the Internet, then millions of casual book sellers, and I got out of the trade. There has never in human history been a better time to buy books. Not so much for the selling.

That sounds more like arbitrage than money pumping; were you ever able to buy e.g. every book in a series from the same seller over a period of time, then sell the entire collection back at a premium to the same person?

Remember there was a time before computerized inventory. A book was bought, given a price and put on a shelf. That book might go up in value. The person who priced it might not work there, or remember to re-price it among the tens of thousands of books in a store. I could buy it because I recognize it went up in value then sell it back (though to be civilized about it, not that day). Some bookstores are so large they might sell a book from their warehouse then buy it back at a profit to me over their counter. Before computerized inventory this was far from unusual or irrational. Now a computer will remember what no person could remember.

An aside: The Old Brown Coat by Lord Dunsany is a short story on this topic.

That sounds like arbitrage combined with razor thin margins and imperfect information;

Imperfect information perhaps. Selling something to the former owner the next day for (if memory serves) fifteen times what I paid for it isn't a razor thin margin.

Even then, the reason this happens might be plausibly explained by the changing information of the bookstore rather than actual intransitivity.

Back when I was playing Everquest, at the point where my character gained the dual wielding ability, I got my hands on a so-so second weapon, but I decided I wanted a better one, so I set out on a trading expedition to get one.

I traded away my secondary weapon, along with some money, in a chain of deals, intending each one to bring me closer to my intended goal of a better weapon. Sometimes I realized after making a deal that I had lost ground on it, and would try to trade up the chain to gain more value.

Eventually, I realized that I had fallen far enough from my original goal that I traded back for the same weapon I'd started with, leaving me with considerably less money than I'd had to begin with.

In Australia, a mandatory 9% of your salary is paid into a super-annuation account. This money represents a loan from you to a financial institution, for which they pay you an interest rate X% per annum, once management fees etc are accounted for.

At the same time, almost all Australians take out a large loan from one of the same handful of financial institutions, with a term of much of their working life, for an interest rate of Y% per annum, and invariably Y > X. The only advantage of this arrangement to an individual is that superannuation savings are protected even in the case of bankruptcy, so the risk is attenuated. I don't know the percentage of Australians who go bankrupt, but I suspect it's quite small (we have no medical bankruptcies).

It's also fairly common for people to carry both a credit card balance at high interest, and a savings account at low interest.

I suggest that savings feel safer than debts feel risky, creating the circularity to be pumped.

It's also fairly common for people to carry both a credit card balance at high interest, and a savings account at low interest.

Savings provide liquidity.

On the other hand, people can be pretty stupid about when to pay off mortgages early vs saving more.

I think life insurance could be one of these examples. I heard that when people buy life insurance, they typically cancel it a few years later. And then they buy a new one, more expensive than the previous one, because they are older.

Then there are "financial advisors" who advise people to cancel all their existing insurance and buy new ones. I don't know how often people are willing to go through that cycle, though.

I suspect the cycles of the directed cyclic graph take a bit longer to run through than an obvious and quick cycle as this seems to imply.

I have a vague hypothesis that the cyclic nature of human preferences is something evolution hit upon to keep us working, working, working on enhancing our descendants' chances - that that's what the money pump actually is. Remember that a common failure mode in real-world AIs is them getting stuck. So if this is the case, then something won't capture human value until it captures the cyclic, pumpable nature of it. This has obvious and annoying implications for coherently extrapolating it, of course.

Do payday loans qualify as intransitive preference money pumps? I think some use cases might qualify under 7-time preference and discounting; there exists a time preference for each use case such that the use case is rational, but I don't think every embodied use case has an embodied time preference.

I have rejected the general case of miscalculated value of future repayments on loans. I am not completely comfortable with my rejection, but I can't fit it to my prototypical model of money-pumping without making questionable assumptions about time-sensitivity.

Assume that the borrower intends to repay the loan; the question then becomes "Is it better to have [thing I can buy] right now than to have [more things I could buy later] instead?" For major capital investments, like houses, it's reasonable to say yes. For things like consumers who keep constant payday loans but spend their money on things that they don't really want, it is less reasonable.

It's still reasonable to exclude 'money pumps' that require a minimum time to elapse, because the truest form of money pump is a series of exchanges that end up strictly worse that would all be accepted at the same time or in rapid series.

Fashion. Every time someone updates their wardrobe to stay in style, they are happy they're doing it. And it's not uncommon for some styles to come back into fashion years later.

That would be a money pump if they preferred style A over style B, and style B over style A. But that isn't what they really prefer. What they prefer is to wear whatever style is in fashion. Style A is preferable to style B if and only if style A is in fashion, and vice versa. Preferring specific styles is an instrumental value, staying in fashion is the terminal value.

Similarly, imagine one day I prefer pizza to tacos, the next day I prefer tacos to pizza, and the day after that I prefer pizza to tacos. My preferences weren't intransitive, because I didn't really prefer those foods at all, what I preferred was eating tasty things, preferring tacos or pizza were instrumental values. During the day I preferred tacos to pizza, I had gotten sick of pizza and no longer found it tasty.

You still lose money, though.

There seems to be some confusion. When we say "money pump" we are talking specifically about the dutch book. We aren't listing (arguably) inefficient ways to spend money.

Oh, indeed! I was just noting that preference change over time is a problem, not the same problem as money pumps.

You do, but saying fashion is a money pump because you need to keep spending money to stay in fashion is like saying that grocery stores are a money pump because you need to keep spending money to not starve.

Oh, sorry, I wasn't saying fashion is a money pump, just noting that it's a problem for agents who's preferences change with the passage of time; just like needing to purchase food is something most people would avoid, if possible.

a combination of 2 and 3 seems most plausible to me. If the loop gets at all obvious people will become indignant and this will change their preferences.

Those who know, won't tell. They just do it. Trevor_Blake told us now, when it is all over.

Here's a version of hypothesis #3 that sounds plausible to me: Humans have a preference for being consistent, and for other people to be so as well. If someone points out their inconsistency they often get upset, or try to rationalize it away. Many times a person caught being inconsistent will change their behavior to avoid future embarrassment. A common way to stir up anger against politicians is to point out times when they have behaved inconsistently.

Now, the obvious armchair ev-psych explanation for this consistency preference evolving is that consistent people are more predictable and stable, and hence make better allies and trade partners. But it occurs to me that this preference for being consistent in one's behavior, especially in front of others, might also serve to protect us from money pumps to some extent.

On one occasion, gamblers in Las Vegas played these kinds of bets for real money, using a roulette wheel. And afterward, one of the researchers tried to explain the problem with the incoherence between their pricing and their choices. From the transcript:

Experimenter: Well, how about the bid for Bet A? Do you have any further feelings about it now that you know you are choosing one but bidding more for the other one?

Subject: It's kind of strange, but no, I don't have any feelings at all whatsoever really about it. It's just one of those things. It shows my reasoning process isn't so good, but, other than that, I... no qualms.

...

E: Can I persuade you that it is an irrational pattern?

S: No, I don't think you probably could, but you could try.

...

E: Well, now let me suggest what has been called a money-pump game and try this out on you and see how you like it. If you think Bet A is worth 550 points [points were converted to dollars after the game, though not on a one-to-one basis] then you ought to be willing to give me 550 points if I give you the bet...

...

E: So you have Bet A, and I say, "Oh, you'd rather have Bet B wouldn't you?"

...

S: I'm losing money.

E: I'll buy Bet B from you. I'll be generous; I'll pay you more than 400 points. I'll pay you 401 points. Are you willing to sell me Bet B for 401 points?

S: Well, certainly.

...

E: I'm now ahead 149 points.

S: That's good reasoning on my part. (laughs) How many times are we going to go through this?

...

E: Well, I think I've pushed you as far as I know how to push you short of actually insulting you.

S: That's right.

-Zut Allais

You are correct. I was thinking the kind of arbitrage where people are offering a futures contract at a different price than the current price of the good (modulo the risk-free rate, storage costs and consumption of the good). Then they will be arbitraged and lose money for no good reason.

Financial arbitrage is nothing but this, essentially...

Exploitation of money pumps can be considered arbitrage but not all arbitrage is the exploitation of money pumps of this kind.

I was thinking the kind of arbitrage where people are offering a futures contract at a different price than the current price of the good (modulo the risk-free rate, storage costs and consumption of the good). Then they will be arbitraged and lose money for no good reason.

Investors know not to do this, but it's rarely immediately obvious to people who start learning how to invest.