Is AI destined to widen or shrink the gap between haves and have-nots, or do rollout choices decide?

This explores whether AI tilts toward more or less inequality by nature, or whether deployment decides. The corpus leans toward deployment: the technology can go either way, and the choices around it pick the direction.

The most direct statement is an interdisciplinary review across information, work, education, and healthcare. In every one of those areas, generative AI can both worsen and reduce inequality, and the direction is set by access, integration, and incentive structures, not by how capable the model is Does generative AI inevitably worsen or reduce inequality?. Asking whether AI narrows inequality is a bit like asking whether roads do. It depends on where they go and who can use them.

The optimistic case has real evidence behind it. In a randomized experiment with 1,174 adults, generative AI cut the advantage of the higher-educated on a business problem-solving task from 0.548 to 0.139 standard deviations, closing about three-quarters of the gap. Lower-education participants also kept part of their gain after the AI was removed, so the leveling wasn't only borrowed competence Can AI narrow the education performance gap?. One caution is that this is a single, tidy task. A separate analysis of 960 real occupational workflows found that agents that ace benchmark contests still fail at long-horizon professional work Why do agent benchmarks not predict real economic value?. A gap that closes on a clean task may not close in an actual job.

The worry in the corpus is less about task performance and more about who ends up with the capability and the power. One note argues that generative AI models crystallize humanity's collective output, so restricting access to them risks privatizing shared knowledge and creating new inequality Should restricting AI access create new kinds of inequality?. Another describes a quieter mechanism. Societies stay aligned with human interests partly because they depend on workers who care about outcomes. As AI replaces that labor, ordinary people lose influence gradually, and the drift across institutions could become irreversible Does incremental AI replacement erode human influence over society?. That is inequality of power, not just of skill or wages.

Together these separate two questions. At the level of one person and one task, AI can plainly help the less-advantaged catch up. At the level of a society, that only translates into narrower inequality if access stays broad and human contribution stays economically and politically necessary. The retrieved material has no long-run field data on who ends up better off, so the deployment-decides claim rests on one review and one experiment, and the rest is structural argument.