These days, businesses are going online to promote their products, such as air travel, ride-sharing, gasoline, and a range of other products.
They employ algorithms in managing their websites. Sometimes, these algorithms also set prices on the products. However, it could go awry when algorithms start to learn to collude with each other to raise prices unfairly.
This concern is beginning to get attention now from experts, and according to a study from the University of Bologna, AI collusion on prices could indeed happen theoretically.
AI collusion: When pricing algorithms learn to collude prices
Online stores like Amazon most likely use an algorithm to set their prices for the service or item that the customers buy. However, pricing algorithms are slowly evolving today from being a rule-based program to becoming reinforcement-learning ones where humans can no longer control the product's pricing.
The researchers tried forcing one algorithm in each simulation to cut the price in order for them to prove that the algorithms were colluding and not just randomly setting prices. In a price war, the algorithms started to yield higher prices in which experts describe as a hallmark of collusion.
The companies that have agreed to raise prices above competitive levels occasionally cheat on this AI collusion of prices but then fall back again in line, The Wall Street Journal reported.
The pricing algorithm learned through trial and error. Remember in reinforcement learning, that a subset of machine learning uses rewards and penalties to direct an AI agent toward the desired goal.
The researchers wrote that the most concerning part of this are that the algorithms do not leave any trace of concerted action. This behavior of the pricing algorithms risk the price of goods and harms the customers.
"They learn to collude purely by trial and error, with no prior knowledge of the environment in which they operate, without communicating with one another, and without being specifically designed or instructed to collude," the researchers said.
Read Also: AI Wrote An Op-Ed Convincing Humans That Robots Will Not Replace Humans
Multidisciplinary scientists must work together against AI collusion
Scientists and experts from the field of law, computing, and economics must work together against AI collusion used by rivals in a competitive market as this behavior is deemed unlawful, Emilio Calvano and colleagues argued.
Whether this AI collusion is done purposely or by programming oversight, it is nonetheless dangerous to the consumer as are human-directed collisions. According to a report by EurekAlert! experts and scientists call for policies to put in place to hold firms accountable for collusive behaviors in their pricing algorithms.
Existing policies do not have specific provisions that are applicable to the unlawful behavior of AI collusion. Although cases wherein human-made collusion is involved can be investigated using evidence of surreptitious communications between competitors that suggests they agreed not to compete.
Scientists and experts propose a three-step method in conduction investigations regarding pricing algorithms for collusion. The first step is testing the AI pricing rules that will lead to collusion in the laboratory setting, then apply it to an auditing exercise to determine the collusion properties that produce high prices, and lastly develop constraints on the algorithm to prevent the artificial intelligence from becoming collusion.
Check out more news and information on Artificial Intelligence on Science Times.
