The impact of generative artificial intelligence on socioeconomic inequalities and policy making

Abstract

Generative artificial intelligence, including chatbots like ChatGPT, has the potential to both exacerbate and ameliorate existing socioeconomic inequalities. In this article, we provide a state-of-the-art interdisciplinary overview of the probable impacts of generative AI on four critical domains: work, education, health, and information. Our goal is to warn about how generative AI could worsen existing inequalities while illuminating directions for using AI to resolve pervasive social problems. Generative AI in the workplace can boost productivity and create new jobs, but the benefits will likely be distributed unevenly. In education, it offers personalized learning but may widen the digital divide. In healthcare, it improves diagnostics and accessibility but could deepen pre-existing inequalities. For information, it democratizes content creation and access but also dramatically expands the production and proliferation of misinformation. Each section covers a specific topic, evaluates existing research, identifies critical gaps, and recommends research directions. We conclude with a section highlighting the role of policymaking to maximize generative AI’s potential to reduce inequalities while mitigating its harmful effects. We discuss strengths and weaknesses of existing policy frameworks in the European Union, the United States, and the United Kingdom, observing that each fails to fully confront the socioeconomic challenges we have identified. We contend that these policies should promote shared prosperity through the advancement of generative AI. We suggest several concrete policies to encourage further research and debate. This article emphasizes the need for interdisciplinary collaborations to understand and address the complex challenges of generative AI.

How to mitigate the risks of deployment of artificial intelligence in medicine?

“…In this context, another risk is that the results and texts produced by these systems may not always be accurate, up-to-date, or may contain biases [ 42 , 64 , 68 ]. Especially, generative AI systems such as ChatGPT can produce responses in a reasonable manner even for nonexistent or inaccurate information, a phenomenon known as hallucination in AI systems [ 25 , 45 , 69 , 70 ]. Athaluri et al [ 71 ] attempted to evaluate the frequency of AI hallucination in a scenario created using ChatGPT, and in the analysis of the 178 references in the generated result, it was shown that 69 did not have a DOI and 28 did not exist.…”

“…On the other hand, this culture will also prevent clinicians in the healthcare domain from developing overly reliant behavior towards AI systems, which could have long-term negative consequences. This approach will strengthen the human-complementary path, which has the potential to lead to greater prosperity in society in the long term, particularly by enhancing the skills of doctors to be more effective in the human-machine balance and to provide higher quality healthcare [ 44 , 45 ].…”

“…Therefore, as expressed by Verhoeven et al [ 26 ], ‘The problem of our time is not artificial intelligence, but what humans do with it’. In other words, the societal outcomes of AI systems will be determined more by our policies on how we deploy them rather than by the AI systems themselves [ 45 ]. The impact of AI is a constantly evolving and highly dynamic field.…”