Tautalus
Regular Member
- Messages
- 588
- Reaction score
- 1,503
- Points
- 93
- Ethnic group
- Portuguese
- Y-DNA haplogroup
- I2-M223 / I-FTB15368
- mtDNA haplogroup
- H6a1b2y
AI is taking up an increasingly larger place in our economy and in our lives, and its impacts are becoming more and more visible. One problem that is becoming increasingly worrying is the AI-related layoffs. An increasing number of businesses are replacing employees with AI and automation, leading to significant workforce reductions.
The authors of this paper argue that AI-driven layoffs can create a market failure that pushes firms to automate more jobs than is economically optimal, even when managers fully understand the long term consequences. The central claim is that competitive markets alone cannot prevent an AI-driven cycle of excessive automation because each firm's incentives are misaligned with the collective interest.
The central argument is that when a company replaces workers with AI, it enjoys the full reduction in labour costs. However, those laid-off workers also lose income and therefore spend less as consumers. Since consumer spending supports all firms, not just the one making the layoffs, each company bears only a small fraction of the demand it destroys while receiving all of the cost savings. This creates what the authors call a demand externality: the benefits of automation are private and the costs are spread across the entire economy. As a result, every firm automates more than is collectively desirable.
The AI Layoff Trap resembles a Prisoner's Dilemma, if all firms restrained automation, overall consumer demand would remain stronger, benefiting everyone. But any individual firm that refuses to automate while competitors do will suffer higher costs and lose market share. Therefore every firm's rational strategy is to automate and collectively everyone becomes worse off because demand collapses. The paper argues this remains true even if every CEO perfectly foresees the outcome. The problem is incentives, not ignorance.
One of the paper's more surprising conclusions is that greater market competition increases excessive automation. With more competing firms each firm internalizes an even smaller share of the lost consumer demand, therefore each has even stronger incentives to automate. By contrast, a monopolist would fully internalize the lost demand and would automate much more cautiously.
Contrary to the common belief that more capable AI naturally benefits society, the paper argues that lower AI costs, higher AI productivity, easier deployment, all strengthen the incentive to automate. Rather than solving the problem, better AI intensifies the automation race because every firm seeks the same competitive advantage.
Who loses? The authors argue the outcome is not merely a redistribution from workers to capital owners. Instead, workers lose wages, firms eventually lose profits because weakened consumer demand reduces sales, society experiences a deadweight loss. Thus, over automation ultimately harms both labour and capital.
The paper evaluates several commonly proposed solutions.
Upskilling and retraining : Helps by enabling displaced workers to regain income, reduces the problem but usually does not eliminate it.
Universal Basic Income : Raises living standards, does not change firms' incentives to automate, therefore does not solve the externality.
Capital income taxes : Reduce profits, leave automation incentives unchanged, ineffective.
Worker ownership or profit-sharing : Helps recycle some income back into consumption, narrows the distortion but cannot eliminate it in realistic settings.
Voluntary agreements between firms : Fail because every firm has an incentive to defect, the game remains a Prisoner's Dilemma.
Pigouvian automation tax : The only policy the model identifies as fully correcting the externality, a tax on each automated task equal to the external demand loss aligns private and social incentives, the resulting revenue could fund retraining or income replacement.
The authors acknowledge their model deliberately simplifies reality. It assumes that AI directly replaces labour, that workers spend a larger fraction of income than capital owners, that re-employment after layoffs is incomplete, that firms compete in the same market , that consumer demand depends heavily on labour income.
They also discuss situations where these assumptions may be weakened, such as faster worker re-employment, stronger capital-income recycling, or broader general-equilibrium effects.
The paper's central message is not that AI is inherently harmful. Rather, it argues that competitive markets can systematically over adopt labour replacing AI because each firm ignores part of the demand it destroys. Left unchecked, this creates an "automation arms race" that may reduce both employment and long run profits. The authors therefore conclude that policy should address not only the social consequences of AI displacement, but also the competitive incentives that drive firms to automate beyond the socially efficient level.
The paper is available here
arxiv.org
The authors of this paper argue that AI-driven layoffs can create a market failure that pushes firms to automate more jobs than is economically optimal, even when managers fully understand the long term consequences. The central claim is that competitive markets alone cannot prevent an AI-driven cycle of excessive automation because each firm's incentives are misaligned with the collective interest.
The central argument is that when a company replaces workers with AI, it enjoys the full reduction in labour costs. However, those laid-off workers also lose income and therefore spend less as consumers. Since consumer spending supports all firms, not just the one making the layoffs, each company bears only a small fraction of the demand it destroys while receiving all of the cost savings. This creates what the authors call a demand externality: the benefits of automation are private and the costs are spread across the entire economy. As a result, every firm automates more than is collectively desirable.
The AI Layoff Trap resembles a Prisoner's Dilemma, if all firms restrained automation, overall consumer demand would remain stronger, benefiting everyone. But any individual firm that refuses to automate while competitors do will suffer higher costs and lose market share. Therefore every firm's rational strategy is to automate and collectively everyone becomes worse off because demand collapses. The paper argues this remains true even if every CEO perfectly foresees the outcome. The problem is incentives, not ignorance.
One of the paper's more surprising conclusions is that greater market competition increases excessive automation. With more competing firms each firm internalizes an even smaller share of the lost consumer demand, therefore each has even stronger incentives to automate. By contrast, a monopolist would fully internalize the lost demand and would automate much more cautiously.
Contrary to the common belief that more capable AI naturally benefits society, the paper argues that lower AI costs, higher AI productivity, easier deployment, all strengthen the incentive to automate. Rather than solving the problem, better AI intensifies the automation race because every firm seeks the same competitive advantage.
Who loses? The authors argue the outcome is not merely a redistribution from workers to capital owners. Instead, workers lose wages, firms eventually lose profits because weakened consumer demand reduces sales, society experiences a deadweight loss. Thus, over automation ultimately harms both labour and capital.
The paper evaluates several commonly proposed solutions.
Upskilling and retraining : Helps by enabling displaced workers to regain income, reduces the problem but usually does not eliminate it.
Universal Basic Income : Raises living standards, does not change firms' incentives to automate, therefore does not solve the externality.
Capital income taxes : Reduce profits, leave automation incentives unchanged, ineffective.
Worker ownership or profit-sharing : Helps recycle some income back into consumption, narrows the distortion but cannot eliminate it in realistic settings.
Voluntary agreements between firms : Fail because every firm has an incentive to defect, the game remains a Prisoner's Dilemma.
Pigouvian automation tax : The only policy the model identifies as fully correcting the externality, a tax on each automated task equal to the external demand loss aligns private and social incentives, the resulting revenue could fund retraining or income replacement.
The authors acknowledge their model deliberately simplifies reality. It assumes that AI directly replaces labour, that workers spend a larger fraction of income than capital owners, that re-employment after layoffs is incomplete, that firms compete in the same market , that consumer demand depends heavily on labour income.
They also discuss situations where these assumptions may be weakened, such as faster worker re-employment, stronger capital-income recycling, or broader general-equilibrium effects.
The paper's central message is not that AI is inherently harmful. Rather, it argues that competitive markets can systematically over adopt labour replacing AI because each firm ignores part of the demand it destroys. Left unchecked, this creates an "automation arms race" that may reduce both employment and long run profits. The authors therefore conclude that policy should address not only the social consequences of AI displacement, but also the competitive incentives that drive firms to automate beyond the socially efficient level.
The paper is available here
The AI Layoff Trap
If AI displaces human workers faster than the economy can reabsorb them, it risks eroding the very consumer demand firms depend on. We show that knowing this is not enough for firms to stop it. In a competitive task-based model of a transitioning economy, each firm captures the full cost saving...
arxiv.org
Some news about the topic
The AI Layoff Trap: Companies May Be Harming Themselves — Enterprise DNA
A peer-reviewed paper from Penn and Boston University shows businesses caught in an 'automation trap' that creates deadweight loss for everyone.
enterprisedna.co
