Slashdot Mirror
Left To Their Own Devices, Pricing Algorithms Resort To Collusion (popularmechanics.com)
Reader schwit1 shares a report: When you're browsing online, who sets the prices? An algorithm, most likely. A study from 2015 showed that a third of all items on Amazon [PDF] had prices set by an algorithm, and chances are that percentage has only risen. A new study shows how easy it would be for price-setting algorithms to learn to collude with each other and keep prices at a disadvantage for customers.
This sort of collusion would stem from a certain type of algorithm, the researchers say. Reinforcement algorithms learn through trial and error. In the simplest terms, a walking robot would take a step, fall, and try again. These algorithms have often been used to teach algorithms to win games like Go.
"From the antitrust standpoint," say professors Emilio Calvano, Giacomo Calzolari, and others from the University of Bologna in Italy, "the concern is that these autonomous pricing algorithms may independently discover that if they are to make the highest possible profit, they should avoid price wars. That is, they may learn to collude even if they have not been specifically instructed to do so, and even if they do not communicate with one another."
This sort of collusion would stem from a certain type of algorithm, the researchers say. Reinforcement algorithms learn through trial and error. In the simplest terms, a walking robot would take a step, fall, and try again. These algorithms have often been used to teach algorithms to win games like Go.
"From the antitrust standpoint," say professors Emilio Calvano, Giacomo Calzolari, and others from the University of Bologna in Italy, "the concern is that these autonomous pricing algorithms may independently discover that if they are to make the highest possible profit, they should avoid price wars. That is, they may learn to collude even if they have not been specifically instructed to do so, and even if they do not communicate with one another."
There's case studies many economics classes use of how the airlines in the 1970s would evade the collusion in restraint of trade laws. Basically, they developed a process for proposing fare changes. They would announce a scheduled fare increase a month in advance. If the other company followed suit they would enact it if they didn't they would retract it or post a fare decrease shortly after that.
It worked quite well till the regulators figured it out. Braniff (deceased high end airline) was the ringleader.
These pricing algorithms just accomplish this one time scales of hours rather than months.
but it's just old wine in new bottles. No one invented anything or changed anything. But the amplification and universality of it made the problem worse.
But what actually makes this interesting is that it may also be an emergent behavior as opposed to either intentional programming or an untended artifact of some algorithm. THat is, if an AI is simply asked to maximize profit in a multi-agent system it's entirely possible it will learn this tit for tat strategy. Humans do in cooperation-games in behavior theory.
Some drink at the fountain of knowledge. Others just gargle.
No, it's pretty obvious that Amazon DOES NOT adjust adjust pricing on a per-consumer basis, and they do not use your past purchasing history to set pricing specific to you. If they did, then sites like camelcamelcamel would not work. Amazon does adjust pricing over time, but at any given time, all buyers are offered the same price.