There is a lot of talk right now about AI replacing jobs and causing layoffs. A new research paper tries to explain what this could mean for the entire economy. The paper is written by Brett Hemenway Falk from the University of Pennsylvania and Gerry Tsoukalas from Boston University. It uses game theory (a way of studying decision-making) to show how AI-driven layoffs could create a serious economic problem.
The “AI layoff trap” explained
The paper introduces a concept called the “AI Layoff Trap.” The idea is that companies use AI to cut costs, so they lay off workers. But when workers lose jobs, they stop spending money. When spending drops, companies start losing revenue. Eventually, even the companies that cut jobs to save money can end up struggling or going bankrupt.
This creates a dangerous cycle that if every company fires workers to cut costs, every fired worker stops buying product the revenue collapses across every sector and the companies that fired everyone go bankrupt.
Why companies can’t stop this
The study compares this situation to a “Prisoner’s Dilemma,” a well-known concept in game theory. Each company knows that too much automation could hurt the economy. But no single company can afford to stop.
If one company uses AI and others don’t, it wins. If one company refuses to automate while others do, it loses. So every company keeps automating, even if they know it could damage the bigger system.
The “red queen effect”
The authors call this ongoing race the “Red Queen Effect.” As AI becomes cheaper and better, the pressure to automate only increases. Companies must keep moving faster just to stay competitive. But this creates a problem that while companies become more efficient, people have less money to spend. So overall demand in the economy shrinks. In simple terms, companies are producing more but selling less.
Why common solutions don’t work
The paper looks at several popular ideas to fix this problem and finds that most of them don’t actually solve it. Universal Basic Income (UBI) can help people survive, but it doesn’t change the decision companies make about replacing workers. Taxes on profits don’t directly affect whether a company automates a specific job. Giving workers shares in companies helps a bit, but not enough. Agreements between companies to slow down automation don’t work because companies can cheat.Training and upskilling only works if people end up earning even more than before, which rarely happens.
What is the only solution?
According to the researchers, there is only one effective solution, a “Pigouvian automation tax,” also called a “robot tax.” This means companies would pay a tax every time they replace a worker with AI. The tax would match the amount of spending power lost when that worker loses their job. The goal is to make companies fully account for the broader impact of automation. This way, automation decisions would balance both business profits and the health of the economy.
The paper suggests that rising layoffs in tech and other sectors may not slow down on their own. Even if companies see warning signs, competition pushes them to keep automating.
Without the right kind of policy intervention, the study argues that this “AI Layoff Trap” could be unavoidable. What looks like a productivity boost from AI could actually turn into a slowdown in overall economic demand.