Figure 4: An actor on an open market is facing the problem of
deciding with whom she shall make business.
We have already observed some, and hope to observe more of fraudulent behaviours as we expand the model to make the actors more sophisticated. Some of these behaviours are described below.
In an electronic market economic transactions take place at a rate much higher than what is usual in the ordinary market. Therefore it is possible to make a lot of money even on very small crimes. By falsely claim to perform a certain small and insignificant task, a person can rapidly con a large number of actors and make a lot of money.
It is not probable to think that an actor always is able to
decide whether he/she have been ripped off or not. For example,
if I want to know the size of the population in Canada anyone can
propose to answer and give me a random number between ten and
fifty million.
By the time the information is
taken into use it might not have been possible to verify it.
Figure 5: One way to choose business partner is to pick one at random.
This is very vulnerable to con strategies since there is no way
to retaliate on bad behaviour.
If we run a simulation in a one-room world with mostly good sellers where the buyers choose between all the sellers at random, we find that bad sellers who are charging a high price and giving a low value in return will make the most money. Good sellers will loose market shares to bad sellers since more and more buyers go bankrupt and more new bad sellers enter the market. Soon there are so many bad sellers that the buyers disappear completely from the market.
Figure 6: With many buyers there is always someone new to con.
On a global market there will be so many customers that even larger, detectable frauds can be profitable. Since there are so many people to con, there will certainly be people who haven't heard of the fraud and who will fall for it.
Figure 7: A buyer who is sticking to the same seller if the seller is
nice will manage ok, but he will not find the optimal value.
A simulation where the buyers use their favourite seller, that is, a seller that didn't fool him the last time, illustrates buyers who can distinguish that they have been fooled. We find that the malicious sellers disappear from the market. But, if their view length is limited (35) and they move about a lot (10) they will occasionally choose other sellers since they loose track of their favourite (this illustrates new buyers) . There will be a low but constant number of bad sellers.
If the buyers choose the cheapest seller then all the buyers will quickly disappear since all the money will go to cheap malicious sellers. There is no way for the nice sellers to make enough money to stay in business. All the buyers will run out of capital and be removed from the market. But, if there is very little migration on the board then there will sometimes be an some buyers who only find a nice seller within their range of sight. The system will finally stabilise at a population of nice sellers, but at a terrible cost. Almost all of the original buyers will have gone bankrupt. This leads us to conclude that even though several mechanisms can arrive at ``nice'' populations, some can be more ``drastic'' than others in that they demand the elimination of a lot of innocent actors.
In an electronic market new electronic actors might be introduced in such a high rate that it is impossible for a single customer to know whether the new actors are actually doing anything, or whether they are only ripping customers off. If it is possible to create a new electronic identity, old actors can reappear in new clothes, repeating a fraud again and again.
If a large part of our day-to-day transactions are made through an open system with largely unknown components, it might be possible for someone to collect information about the persons using the system. This information can be used to create a computer shadow of a person, which can be used for purposes ranging from directed advertising to blackmail. How can we prevent this from happening?
If we are using electronic agents, who are migrating programs,
then programs of possibly unknown origin can execute on a
person's local computer. It could be a virus or a Trojan horse
trying to steal information. Remember that since the system is
open, anyone can let anything in into the system. A malicious
program doesn't even have to be written in spite. It's not far
fetched to imagine an agent turning into a computer virus because
of unintended bug.
If everybody use the same sources for information about which other actors to use then a monopoly will quickly be established. This leads to extremely sensitive systems both because the buyers are in the hands of the whims of the seller, and because a change in the demanded services is less likely to be satisfied, since there are no competing, slightly different services being offered.
A simulation with one room where the buyers use the room reputation (a ``top ten list'' for that room), there will finally be only one seller left since the others have gone bankrupt. If the seller is permitted to mutate to a bad seller the entire population will finally go bankrupt.
Figure 8: Monopoly can be established when everybody are choosing
using the same algorithm of choice in a global market.
To avoid monopoly we can either create a system with many different different valuation measures and different price levels. We can also introduce the notion of location in the agent system. If it is costly to move from one place to another then an agent will promote the alternatives close to him. This allows for the proliferation of several different sellers in different regions.
If we rerun the same simulation in a world with, say, 9 rooms, there will be more sellers, about one in each room. They don't have monopoly since the buyers can move from one room to another. If the seller changes to bad behaviour, the buyers who stay will go bankrupt and the ones who move will survive. Finally the bad seller goes bankrupt and a new good seller can be introduced.
By introducing location we have to sacrifice the notion of optimal choice (or redefine it to include proximity in its measure). This can be difficult to accept, but perhaps this is better than the alternative, monopoly.
Still this mechanism yields a very low number of sellers on the market. This makes the buyers very sensitive to when the seller leaves their range of sight. When this happens, they often have to choose at random again, and they are possibly choosing a bad seller. Since the bad seller makes a lot of money on this he can afford only being used very seldom.
___________________________________________________
Lars Rasmusson
