Does AI close the gap between more- and less-educated people, or mostly hand a head start to those already ahead?

This explores whether AI closes or widens the gap between people with more and less formal education. In the most direct test in the collection, it closes most of it. In a randomized experiment with over a thousand adults working on a business problem-solving task, generative AI shrank the performance advantage of more-educated participants by about three-quarters Can AI narrow the education performance gap?. The more surprising detail is that less-educated participants kept part of their improvement after the AI was taken away. So the tool did more than cover for missing skills while it was on. It seems to have taught them something.

That retained gain matters because other parts of the collection point the opposite way. A four-month EEG study found that relying on an LLM was linked to weaker brain connectivity and poorer memory of one's own recent work Does AI assistance weaken our brain's ability to think independently?. Separate work shows that AI suggestions can hurt reasoning even when they're correct, because they break the user's concentration Does AI assistance always help reasoning or does it carry hidden costs?. Put together, these suggest a better question than 'narrow or widen.' Does a given use of AI build up the person's own skill, or does it replace it? A gap that closes only while the tool is running is a weaker kind of equality than one that remains when it's gone.

The pattern carries over to hiring. In a hiring experiment, listing AI skills on a résumé partly or fully offset the interview penalty for candidates with associate degrees rather than bachelor's degrees, and for older candidates Can AI skills help older or less-educated job candidates?. The effect was strongest for office assistant roles and weaker for graphic designers. At the level of the whole labor market, though, job-posting data from ten countries shows demand for AI skills clustering in a technical core (Python, SQL, machine learning) inside STEM jobs, with non-technical jobs drifting away from that core rather than toward it Is AI creating common skills across jobs or deepening divisions?. So AI can level the field within a single task or job application while a new split opens between occupations.

The interdisciplinary review in the collection gives the plainest answer: it depends on how AI is deployed, not on the technology itself Does generative AI inevitably worsen or reduce inequality?. Across information, work, education, and healthcare, the same tools both worsened and reduced inequality depending on who had access, how the tools were built into workflows, and what incentives surrounded them. The collection has strong evidence for one key result, the experiment on a single task, but little long-term evidence on whether those catch-up gains last over years or across a career. That gap is worth keeping in mind before treating the three-quarters figure as settled.