AI and Inequality - EconReads

Whether AI will widen or narrow gaps between workers, firms and countries, and what early evidence and theory suggest.

Past technologies have often widened inequality, rewarding highly skilled workers and owners of capital. Will AI do the same?

Reasons AI might widen inequality

Capital owners gain: if AI replaces human work, profits may flow to the owners of AI systems and data centres rather than workers.
Superstar firms: large firms with data and capital may capture most gains.
Skilled workers: workers who can use AI effectively, or build it, may see rising pay.
Global divide: countries with AI companies, chips and computing capacity may pull ahead of those without.

Reasons AI might narrow some gaps

Early research found that AI tools often help less experienced or lower-performing workers most, narrowing gaps within occupations. In a study of customer support agents, AI assistance raised the productivity of novice workers substantially, while having little effect on the most skilled. Similar patterns appeared in some writing and consulting tasks.

This suggests AI could compress wage differences within some jobs by spreading expert knowledge.

Which jobs are exposed

Unlike earlier automation, which mainly affected routine manual and clerical jobs, generative AI affects many higher-paid, knowledge-based tasks. Some economists argue this could reduce the premium for certain professional skills, while others expect new high-paid roles to emerge.

Policy choices

Outcomes depend on policy:

Education and training so more people can use AI.
Tax policy, such as how capital and labour are taxed.
Competition policy to prevent excessive concentration.
Access: affordable AI tools in many languages can spread benefits widely, including to people with disabilities.

Economists Daron Acemoglu and Simon Johnson have argued that the direction of technology is a choice, and that AI should be steered toward augmenting workers rather than simply replacing them.

The novice and the expert

A newly hired customer service agent struggles with complex queries, while an experienced colleague handles them easily. With an AI assistant suggesting responses drawn from the best agents' past answers, the newcomer quickly performs much better. The expert gains little. Within this job, the gap between them narrows.

Thinking AI's effects on inequality are predetermined

AI's effects depend on how it is designed, adopted and governed. Policy choices on education, taxation, competition and access shape whether gains are widely shared.

Key takeaways

AI could widen inequality by rewarding capital owners, superstar firms and AI-rich countries.
Early evidence shows AI often helps less experienced workers most, narrowing some gaps.
Generative AI affects many higher-paid knowledge tasks.
Education, tax, competition and access policies shape AI's distributional effects.