Offer Networks:

Infrastructure for a Post-Money Economy

Bitcoin has been getting lots of press lately as an alternative to traditional currencies. It’s an exciting innovation. But my feeling is that the future of exchange may lie in a very different direction.

I have always felt money is a crude instrument, a woefully inadequate way to represent preferences and guide exchanges and formalize “value.” Yet the only clear alternative, barter, is massively less convenient in modern society; money is obviously a large advance over barter in many ways. There has not been nearly enough exploration of alternate methods of systematizing the exchange of goods and services.

Money does work OK, obviously. Along with other mechanisms of modern capitalist economics, it has enabled the evolution of many amazing, positive aspects of modern society, including the Macbook I typed this blog post on, and the Internet via which you’re likely reading these words right now. But I’ve often felt that the simplification of preference and value it represents, is connected with various problematic aspects of modern society in various ways.

With all the advanced technology at our disposal today, isn’t there some better way to handle exchange of goods and services than crudely projecting the VALUES of everything into a single numerical scale (monetary value)?

After all, the values that we assign to various things, in our minds, are complexly multidimensional (or actually, probably better modeled as nondimensional). Human value is not one-dimensional like cash value. Yet for so many aspects of our lives, we’re study with a system that treats value as one-dimensional.

In this article I’ll outline an alternative I’ve been brewing in the back of my mind for a while. In the course of writing this article I gave this alternative a name: an Offer Network.

Offer Networks, like Bitcoin, would be enabled by Internet technology and advanced computer algorithms. Using Offer Networks, people would be able to make bids to provide one good or service in exchange for getting (not necessarily from the recipient of said good or service) some other good or service that they want. A computer algorithm would reconcile different peoples’ offers and try to find collectively satisfying arrangements.

Using Offer Networks, people would be able to get what they need and some of what they want, and offer what they are willing and able to, without having to pass all their preferences and values through the single dimension of cash value. A kind of cash value called OfferCoins would still exist, but these — as a kind of “post-money  money” — would be defined emergently in terms of a network of formalized non-monetary exchanges.

Expressed abstractly like this, the idea may sound pretty complicated – but actually using an Offer Network would be no more complex than using contemporary Internet transaction systems like eBay or Match.com.

There are echoes of SF stories like “Accelerando” and “Down and Out in the Magic Kingdom” in the Offer Network idea — but here I’m trying to propose a variety of post-money (and in some potential futures even post-economic) system in a straightforward and realistic way without too many fun, whimsical SF overlays.

Whether Offer Networks are really the future of economics, I don’t know. But as Abraham Lincoln and others have said, “The best way to predict your future is to create it.” Perhaps by promoting the idea of Offer Networks now, the odds of something like it coming to pass can be increased.

My feeling is that something like Offer Network is coming, and is going to play a key role in smoothing the socioeconomic changes we’re going to see during the next decades – as technologies like AGI and robotics advance and gradually conquer the job market. But I’ll get back to that aspect after elaborating on what an Offer Network actually is…

As a hint regarding my way of thinking, though … I should explain my reaction to the conventional economist’s argument that prices are a good way of assessing how much people value things. I think this argument holds in some abstraction of reality. But in the actual reality we live in, most people don’t really know what they want, in most complex situations…. And people are constantly changing: Bob today is not the same as Bob from last year. You can never step into the same Bob twice! … Since the notion of a persistent self is a crude model of the actual dynamical system of a human being, the notion of a consistent set of preferences associated with a particular self is also a very crude idea…. The interactions we partake of, including monetary or other exchange interactions, are part of an evolving, self-organizing network that defines who we are and what choices we make…

What is Money, Anyway?

I’ll work up to the Offer Network idea via incremental steps… starting in a hypothetical little village, and ending up eventually with an advanced AI-mediated post-economy.

We use money constantly, yet we don’t reflect often on what money really is. I’ll try to elucidate the nature of money by means of some very simple example cases.

First, let’s consider consider a very simple monetary-exchange scenario in a small community:

• Dean borrows $20 from Todd.
• Bob picks some strawberries from his farm, and sells them to Dean for $20.
• Bob then buys a hamburger from Karen for $20, which Karen produces from her own cow.
• Karen takes the $20 and uses it to buy home repair services from Jack; and Jack uses the $20 to buy some apples from Jane (who picked them on her farm).
• Jane takes the $20 to buy some oranges from Dean.
• Dean pays Todd back the $20 he borrowed.

What are the actual exchanges that have happened here?

• Bob supplied strawberries to Dean, and got a hamburger from Karen
• Karen supplied a hamburger to Bob, and got home repair services from Jack
• Jack supplied home repair services to Karen, and got apples from Jane
• Jane supplied apples to Jack, and got oranges from Dean
• Dean supplied oranges to Jane, and got strawberries from Bob
• Todd just served as the bank, giving a no-interest loan

The exchange of the $20 is just a way of mediating this network of exchanges of goods and services.

Now envision a different way of doing the same set of exchanges, with no $20 involved:

• Bob publishes desire to exchange Strawberries for Hamburger
• Karen publishes desire to exchange Hamburger for Home repair services
• Jack publishes desire to exchange Home repair services for Apples
• Jane publishes desire to exchange Apples for Oranges
• Dean publishes desire to exchange Oranges for Strawberries

Then, the whole little community looks at the various exchange-desires that have been published, and figures out the exchanges that can be done to make everyone satisfied. And Todd has nothing to do with it.

In this particular case, there’s no harm to using money; it’s just a simple way to mediate the exchanges that would happen anyway.

But in more complex cases, money may fail to capture the nuances of the desired exchanges; and can lead to various pathologies as we see in the modern world.

Offer Networks: A Simple Case

The simple example I’ve just run through leads up to a simple case of the Offer Network idea.

One elementary type of Offer Network would be defined as follows:

1. Each person publishes (into a secure clearinghouse system) their willingness to exchange X_i (some good or service) for Y_i (some other good or service)… Then
2. A “clearinghouse” algorithm figures out how to propagate exchanges so that each person gets some exchange that is acceptable to them.

A more sophisticated version would involve bidding, a la eBay and other auction systems. In a bid based Offer Network, each person would specify various possible exchanges (X_i, Y_i) and assign each one a degree of desirability.

Then, e.g.: Bob may be willing to offer an hour of medical consultation to get half an hour of skiing lessons. But first he may offer half an hour of medical consultation in return for half an hour of skiing lessons. If that fails he may up the offer to an hour.

(Or more generally: even if a person is willing to offer X_i to get Y_i, he may first offer some X’_i that is dominated by X_i (i.e. he would prefer to exchange X’_i for Y_i, than to exchange X_i for Y_i). If this initial offer fails to get him a good enough result via the clearinghouse, he may offer some other X’_i < X”_i < X_i, etc.)

A few more examples of specific exchange-pairs might be:

• Ben is happy to exchange knowledge about AI for knowledge about physics or biology
• Mary is happy to exchange home repair services for piano lessons. She’d rather exchange home repair services with a young guy for piano lessons, but is willing to accept someone between 40 and 50 if needed.
• Todd is happy to exchange 5 kilos of strawberries for 10 steaks or 3 whole chickens

An issue that arises immediately in the context of a system like this is the risk of dishonesty and deception. In the simple case of a small community (like the 5 people in the example given above), there is a strong reason for each participant to make good on whatever they’ve promised to provide in a given exchange. If they don’t, everyone else will see that they’re unreliable or dishonest, and will avoid entering into exchanges with them in future.

In a larger society this kind of direct social pressure can’t work, but alternatives exist, e.g. a reputation management system similar to that used in today’s online stores, in which each person rates each transaction they’ve had, so that each person gets an overall reliability rating. More complex variants are possible as well, e.g.

• reliability ratings that are context-specific (Jack sometimes reneges on offers to participate in musical jams; but always comes through with offers to provide programming services)
• an algorithm to determine the reliability of each person **as a rater**, based on whether their ratings tend to be outliers
• an algorithm to estimate the average rating given to a certain provider in a certain context, by raters with criteria similar to one’s own (e.g. you may not care if a house-cleaner shows up late; but John may… so if John down-rates Fred’s cleaning services because Fred shows up late sometimes, you may wish to ignore said down-rating)

Potentially, reputation management could end up being the most complex part of an Offer Network system – much as fraud detection is the most complex part of a system like PayPal.

Beyond Traditional Economic Exchanges

The above kind of Offer Network is already pretty interesting. But it gets better – a moment’s reflection indicates that traditional exchange of goods and services is only a fragment of what this kind of system can handle.

An offer such as

Agent A is willing to exchange X_i for Y_i

is just a special case of a more general offer of the form

Agent A is willing to carry out action A_i,

which has post-condition Z_i,

IF precondition Y_i is met

Everyday examples of offers that aren’t classic economic transactions might be

• Matilda is willing to put 10 hours/week into a project aimed at emulating Oculus Rift in open source hardware (postcondition), if 10 other people with advanced degrees in science are also willing to put at least 10 hours/week into that project
• Maria is willing to go on a date on Friday 11/7 at 8PM with anyone living in the San Francisco area, if they are under 25, employed full-time, have a “Yahoo date rating” of at least .7, and have done at least 5 hours of verified social-welfare volunteer work in the last month
• Ben is willing to go play kazoo at Open Mic Night at Peel Fresco on 12/17, if there will be a drummer and a bass player there

More advanced and complex offers could be issued by various sorts of AI agents, as AI develops.

The clearinghouse can then have knowledge of which post-conditions imply which pre-conditions, and the whole process of clearing everyone’s offers becomes the execution of a giant “production system” style logic engine. Economy becomes cognitive in a very obvious and transparent sense. There is the possibility for the clearinghouse to optimize the degree of overall desire fulfillment via heuristically or probabilistically predicting which post-conditions are most likely to lead to which preconditions (or which combinations of post-conditions are most likely to lead to which combinations of post-conditions). And of course these calculations would be taken into account in the calculations defining the OfferCoin value of an offer.

Traditional economic actions involving exchange of goods and services become freely intermixed with other kinds of actions involving context-dependent willingness to commit to carry out various sorts of activities. But of course these more general actions are “economic” too in a broader sense, because they involve agents carrying out actions, hence expending energy…

As a simple example indicating the flexibility of this sort of system, various variations of “buy local” or “buy from a group I like” could be easily achieved via an Offer Network system . For instance I could offer a discount on some service I offer to anyone who participated in online transhumanist discussion … or to people who live in my village … etc.

There is a surface parallel between Offer Networks and old-style socialist ideas about computer-controlled centrally planned economies. In these old ideas, as in Offer Networks, a massive integrative system was often posited to balance everyone’s desires against each other. But a critical difference is that, in the Offer Networks approach the clearinghouse doesn’t try to determine what anybody’s needs or abilities are. Instead, people specify their offers based on their own value systems; and the clearinghouse just mediates the transactions, in a way that respects the various individuals’ desires and values insofar as they have articulated them within their exchange offers.

Emergent Money

Generalized offers as supported by Offer Networks have a richness and humanity to them that purely quantitative financial calculations do not. They also allow people to propose and execute exchanges that are faithful to the nontransitivity of their preferences.

However, money does have a certain simplicity to it. In some cases specifying exchange-pairs will feel overly complicated and the simplicity of some sort of money will be preferable. For this purpose, it is possible to define a sort of emergent money based on an Offer Network — an OfferCoin…

Further, it is difficult to see how a pure exchange network, with no abstract quantification of value into some money-like entity, would handle nonspecific deferred gratification.

If everyone were just living in the now and exchanging today-stuff for other today-stuff, then you could just have a clearinghouse of offers, with no need for symbolic, money-like tokens of value.

OR, if everyone were exchanging {today-stuff or specific promises of specific future stuff at specific times} for {today-stuff or specific promises of specific future stuff at specific times}, a pure exchange network would still work as well.

But the problem is that we don’t know what we’re going to want a week from now let alone 5 years from now… Yet we do want to be able to perform some services today, or exchange some perishable goods today, in exchange for being able to get some of “whatever we’re going to want” a year or 5 from now…

This kind of nonspecific deferred gratification is critical to the creation of large projects of any type, at least in the current era or any era before scarcity is wholly abolished.

And it seems that to make nonspecific deferred gratification work, you need some abstract token of value, that can be cashed in for specific goods or services at the future time when you decide what you want…

To enable this within an Offer Network as described above, the most natural route is to estimate an “emergent money value” for a given offer, via defining a measure of how many “exchange units” or “OfferCoins” a given offer is worth.

One way to do this would be as follows:

• The raw OfferCoin value of X is: what percentage of all offers are dominated by X (in terms of what they could obtain in exchange, via the clearinghouse).
• The final OfferCoin value of X (the number of OfferCoins X is worth at a given point in time) is then a normalization of the raw value of X. Say: 1 trillion * raw value / (sum of all raw values)

The OfferCoin value can also be used in exchange-pairs. That is, barter can be mixed with $$ purchases. Someone can offer 500 OfferCoins for a side of beef. The going rate for a side of beef in terms of OfferCoins can then be determined via auctioning, similar to on eBay.

In a sousveillant society (where auctioning is less relevant as real preferences are more transparent) , everyone’s desires can be made open, meaning the whole clearing house can be open. Otherwise, the clearinghouse could be made auditable but secure, using strong encryption methods.

Note that qualitative offers could mix with quantitative OfferCoin-based ones in various contexts; e.g. the kazoo-playing example given above could mix with

• Jim is willing to pay 20 OfferCoins for a drink at Peel Fresco on 12/17, if there will be a kazoo player there
• Jean is willing to open her nightclub Peel Fresco for Open Mic Night on 12/17, if she has commitments for at least 500 OfferCoints worth of drinks to get bought and the open source hardware example given above could mix with
• Jack is willing to put 10/hours week into any available open-source hardware project in the gaming space, if he can find a Ruby programming contract paying him at least 50 OfferCoins per hour, for between 15 and 20 hours per week (or, if the contract involves developing open source code, he will accept a price of 40 OfferCoins per hour)
• Lisa is willing to pay up to 60 OfferCoins per hour for a Ruby or Python programmer who is also an active contributor to the open source community (at least 5 hours per week); she is also willing to donate 5 OfferCoins to African medical care for each hour that this programmer works

It might be desirable to enforce a fairly high inflation rate on OfferCoins. So, if someone’s precondition for an action involves getting OfferCoins, then they can receive OfferCoins and bank them. But if OfferCoins decay in value intrinsically, based on the time since they have been received, then hoarding intrinsically becomes infeasible…. Perhaps the inflation rate could be adapted dynamically via voting among all exchange members.

A big plus of an Offer Networks type system is that it allows one to flexibly embody various values into one’s exchanges and other offers. It’s straightforward to embody values besides simple monetary values in one’s exchange offers (e.g. the desire to hire a programmer who is an open source contributor; the desire to date someone who has done volunteer work; etc.)

To formalize things slightly more, the definition of “monetary value” implicit in the OfferCoins defined above is

OfferCoin_value(offer A) =

(normalization of the rank value of A, when all current offers made by
all people are rank-ordered in terms of what could be obtained from
the offer via a complex series of trades)
*
(fraction of total OfferCoin value that is assigned to current versus past offers)

The latter fraction is determined via the inflation rate: the higher the inflation rate, the greater the fraction of OfferCoin value assigned to the present time.

So as opposed to a Marxist “labor theory of value”, this approach is quite pragmatically assigning a OfferCoin value to something that is roughly proportional to what one could get by trading that thing. However, this calculation is made over the totality of specified non-monetary (i.e. non OfferCoin based) exchanges as well as over OfferCoin based exchanges.

If the inflation rate is voted to be reasonably high, then accumulation of wealth will be difficult and more of the wealth will go to people who are offering goods/services right now, that are judged valuable right now….

The inflation rate toggles between meritocracy and inheritocracy, in a sense (where I consider “inheritocracy” as the practice of assigning value to those whose past selves or ancestors provided perceived value at past times)

Of course the idea of grounding some kind of money in exchange is not new and is wholly obvious. But one thing that’s different here is the free intermixture of OfferCoin-based and “barter specification based” exchange … this keeps the OfferCoins quite directly grounded in peoples’ specified values. For instance, if

• X can be exchanged for 800 OfferCoins plus a few songs and paintings and happy wishes
• Y can be exchanged for 1000 OfferCoins

then X could be judged more valuable even if nobody wants to pay OfferCoins for the songs, paintings and happy wishes (but some people are willing to exchange other valuable things for these items)…

Futarchy

As a bit of an aside, there seems to be a natural connection between Offer Networks and Robin Hanson’s fascinating, hypothetical political scheme of Futarchy (vote on values, bet on beliefs; i.e. bet on methods for maximizing the values).

For instance in Futarchy, a vote could determine a goal such as: providing free health care meeting certain specific criteria to all people in a certain country, by a certain date. Then, various plans could be put forward by different people or groups, suggesting methods for achieving the stated goal. Then people would be invited to vote for or against the success of each plan. The plan that gets the highest odds of success is the one put into place, and then the bets for and against the plan are paid off once the success of the plan is observed. The underlying concept – which has a fair degree of empirical validation — is that groups of people betting their own money tend to make more accurate judgments than groups people simply expressing their opinions.

Futarchy would emerge from the Offer Networks system if

• people make offers of goods/services in ways that are based on their values
• people make bets on outcomes of attempts to realize their values
• institutions make choices (postconditions) based on observations of peoples’ betting behavior (preconditions)

Bets fit perfectly well into the Offer Networks framework; they are offers to give something in the future (the postcondition), conditional on some condition obtaining in the future (the precondition). Reputation ranking can be used to make it likely people pay up when they lose a bet…. A few renegers won’t matter much with political-type bets, as futarchy counts on a large number of small bets being made…

On the other hand, it is not clear to me that futarchy as specified by Hanson would really be needed in an Offer Networks context. It might be that betting turns out to be unnecessary and many aspects of appropriate governance can self-organize via the network of exchange.

Practicalities

To implement an Offer Network effectively in reality would require a lot of art as well as engineering and science.

It may at first seem very complicated for a person to have to specify offers involving pre and post conditions, rather than just specifying the price one is willing to pay for something. But I think this complication could be mostly eliminated via development of appropriate user interfaces.

Also note that, while there is a central clearinghouse in the above, it doesn’t actually have to be implemented in a centralized, monolithic way — it could be implemented as a distributed peer to peer system with no owner, using strong cryptography to maintain integrity, and spare computer cycles to do calculations.

I don’t think that the complexity of Offer Networks is greater than that of the current financial system, with all its derivatives and credit default swaps and central banks and what-not. It’s a different sort of complexity, though — more focused on individuals and their interactions and values than on the actions of specialists and large institutions.

The current, complex financial system emerged gradually, and I suppose that if a system like I’ve described here is going to happen, it’s also going to emerge gradually. I’m not sure what the first steps will be, but I suppose that they will occur in some online marketplace – or maybe even inside some multiplayer game. Perhaps they will happen in the developing world, where typical 21st century financial systems are not that well developed — there is a possibility for the developing world to leapfrog past the perversions of contemporary Wall Street into a more advanced sort of exchange network.

Possible Psychosocial Effects of Offer Networks

The impact of the money economy on psychology is subtler than commonly realized. It has often occurred to me that transcending or augmenting the money-based method of assigning value, in the way that Offer Networks do, might help resolve various problematic issues in modern society. For example, Offer Networks might end up wreaking exciting and positive havoc with our traditional notions of social status….

For instance, modern first world capitalism has a significant aspect of “Spending your life doing things you don’t want, so you can afford to buy things you don’t need.” (And as it happens, current Chinese socialist capitalism has this aspect as well….)

Given current technology, it seems it would likely be possible for most people in the developed world to work 15-20 hours per week at most, not 40 – and to spend the rest of their time simply enjoying themselves, or pursuing creative activities of some sort. Yet this is not the direction society has gone in. Instead, work hours are still fairly long, and work increasingly spills over into off-work hours. Salaries are largely spent in a consumerist way, doing things like upgrading one’s cellphone every year, or buying overpriced, toxically sugary Starbucks coffee concoctions.

Of course, people are aware of the possibility of working less and living a less consumerist lifestyle; and there is a “simple living” movement tending in this direction, it’s just not very popular.

Why are we willing to work so hard at things we don’t like doing, in order to buy so many things we don’t need?

Part of the answer is addiction – we become psychologically addicted to certain routines and possessions, e.g. eating out in restaurants instead of cooking at home, or driving in a private car instead of taking public transport, or the feeling of wearing new & different clothes or jewelry.

And part of the answer is social status. Having a lot of money is itself an indicator of high social status; and money can be used to buy other status indicators. A tremendous amount of money is spent on status symbols of one sort or another. Of course, money is not the only status indicator. Einstein had a very high status in the public eye, yet was not terribly rich. The same for Linux Torvalds. But overall, money is surely the best single proxy for social status in modern societies.

Offer Networks could disrupt both of these factors, in indirect but powerful ways.

There is a great deal of data showing that addiction is most powerful among people (or animals) whose lives are not richly satisfying, including in social dimensions. A rat in an empty, boring cage will become addicted to morphine rapidly, after brief exposure to the joy of the drug. A rat in a cage full of fun, bustling social activity will ignore the morphine dispenser at the end of a tunnel attached to his cage, even after he’s felt the joy of morphine. Because the warmth of his social environment makes the morphine high unnecessary.

Offer Networks could offer more than just a replacement for many monetary exchanges – they could offer a flexible medium for obtaining all different sorts of interactions with other human beings. They could provide a richer social web, thus reducing the psychological emptiness that allows the addictive psychology of consumerism to take hold.

And Offer Networks would also provide alternative means of establishing and measuring social status. Reputation points would provide one means – and could be context-specific. Getting a high reputation score among people who tend to participate in a particular variety of exchange, would provide a form of status within a certain subgroup. Feeling the positive feedback of the people with whom one has done exchanges, may substitute somewhat for the status symbols that now provide people with ego boosts.

Exactly how the sociodynamics of Offer Networks would play out, nobody knows. But I think they have significant potential to disrupt the addiction and status dynamics that largely govern our society today. Doubtless they would bring new peculiarities and problems as well, only to be discovered once the actual usage patterns of Offer Networks have settled in.

Smoothing the Path to the Era of Abundance

It’s interesting to think of Offer Networks in the context of the Technological Singularity – the hypothesized point in the future when AIs become massively more intelligent than people, and the human mind is no longer the direct driver of most innovation on Earth. Ray Kurzweil has famously predicted 2045 as the approximate date for this kind of advance to arrive.

Peter Diamandis has phrased the Singularity in terms of “abundance.” Today we live in an era of relative scarcity. The resources we feel we want to fulfill our human desires, are often difficult for us to come by. But when technology has advanced enough, this may no longer be so. Infinite resources are unlikely to be available even post-Singularity, but one can only eat so many chocolate bars or surf so many waves each day. There may come a point where, relative to the scope of human desires and capabilities, scarcity is no longer a factor.

On the other hand, the same technologies that have potential to lead to radical abundance, also have the potential to obsolete human labor. If we really reach a point where robots and other automated devices are doing most of the work, then we will have to face questions like: A) what will everybody do with themselves instead of working?; B) how will people get money or other resources, when they don’t have the option to get paid for working?

My suggestion is that moving to a different, non-monetary method of exchange – like Offer Networks — might smooth out these difficulties considerably

As material scarcity decreases, more and more of our exchanges will involve our deeper engagement, rather than merely our “putting in time” carrying out an activity, or providing a material object to someone else. The conditions under which humans are desirous of providing deep engagement to some activity are more complex than a simple monetary exchange, making the specification of more complex offers more appropriate.

In the end, once automation has rendered “working for a living” unnecessary, human beings may end up spending most of their time on social interactions and creative pursuits, mediated by Offer Networks or similar systems. Beyond work and beyond money, but not beyond meaningful social interaction and exchange, not beyond creativity, not beyond the joy of life.