When Algorithms go shopping
In the world where algorithms make shopping decisions on our behalf, all bets are off.
On 18th April 2011, a book about the biology of a fly was offered for sale on Amazon for $23,698,655.93 (plus $3.99 shipping). This was not just a glitch—the price of the book was steadily going up in the days before. And then, the next day, the price dropped to $106.23 (plus $3.99 shipping).
What happened?
The high price was not a human decision. It was set in a strange bidding war between two algorithms that were trading on the online platform.
One of thee algorithms, likely in possession of the book, was regularly checking prices and offering its copy for sale at about 1% less than the top price. The other had a different strategy: it was offering the same book at about 27% more than the next best offer, hoping that a buyer wouldn’t notice the cheaper option. We can assume the second algorithm didn’t have the book and, upon receiving the order, it'd buy the cheaper book and get it delivered to the customer, earning the margin in the process.
The two algorithms got stuck in a cycle of “one percent back, twenty-seven percent forward” every day, reaching the multimillion-dollar price. Until, presumably, a human owner of one of the algorithms noticed the cycle and stopped it.
Photo by Craig Sybert on Unsplash
Impact on competition
For the past few years, we have lived in a world where algorithms, or software buying agents, “go shopping” on our behalf. LG ThinQ washing machines can automatically reorder detergent when they are running low. GE, Whirlpool, and LG offer dishwashers that automatically reorder washing pods. There are coffee machines, printers, and even smoke alarms that can reorder coffee pods, ink, and batteries.
While this sounds exciting, there is a concerning aspect about it.
Almost all of the “smart reordering” solutions are bound to one provider and always order the same product. While they fully automate the shopping process, they keep customers stuck to one retailer and one product.
This sounds very much like the quasi-monopoly position Microsoft enjoyed in the 1990s and 2000s, effectively forcing users to use Windows Media Player and Internet Explorer. Governments in the U.S. and in Europe stepped in to level the playing field.
Locking out humans
Bots are being used not just to shop for coffee pods, but also to book campsites and appointments. Campers share algorithms to book Yosemite Park camping spots and customers willing to make an appointment with Department of Motor Vehicles in New York can “hire” their own robocallers to make an appointment and keep checking if a better slot becomes available. Very recently, bots snatched practically all new graphics cards released by Nvidia.
Algorithmic shoppers can make split-second decisions.
They can make thousands of calls or website visits a day. Humans stand no chance against bots when trying to access products and services that might be in demand. However, blocking software buying agents is not addressing the problem. As technology advances, bots find new ways of accessing systems.
We need clear rules that govern the behaviour of software buying agents.
These can include developing robot exclusion standards, similar to those used to prevent search engines from accessing some websites. We also need a coordinated approach for software buying agents to disclose who they are, in situations where they can be confused for a human.
Sending waves through the markets
Majority of stock market transactions are fully automated and executed by algorithms.
What could go wrong?
In 2017, Dow Jones accidentally released a message about a merger of Apple and Google. For about one second, Apple’s stock went up from $2 to $158, exposing the vulnerability of algorithms in stock markets.
In response to this and similar cases, South Korea is now trying to increase control of algorithmic trading. Other countries will hopefully follow suit.
The emerging Economy of Algorithms, where software agents act on our behalf, has a potential of dramatically changing the way we live, work, and think. However, we need to ensure that clear rules are governing the behaviour of algorithms. With the right mechanisms for the protection of competition in the markets, regulating access to products and services, and enforcing minimum quality standards of algorithms that shop on our behalf, we stand a good chance of enjoying our future with algorithms that go shopping.