What AI Could Mean for Accessing Public Benefits

For decades, technological evangelists have heralded each flashy new tech advancement as transformative and world-changing. From videophones to flying cars, the true impact has almost never lived up to the hype. However, it’s becoming increasingly clear that recent developments in generative artificial intelligence, or AI, are likely no over-promise. The technology has already begun profoundly impacting numerous aspects of our lives, from content creation to customer service, scientific research, and the arts.

One critical area that will almost certainly see dramatic changes in the years ahead is that of accessing public benefits programs—a realm already fraught with systemic barriers. While AI brings great potential to streamline burdensome processes and expand access to critical services, it also poses significant risks if deployed without care. As we stand at the precipice of an AI revolution, it’s critical that policymakers, tech companies, and advocates work collectively to steer these technologies towards empowering people in need, rather than further isolating them. The promise is real, but far from guaranteed.

These developments come at a moment of growing recognition of how critical a robust, accessible safety net is to building and maintaining a prosperous society. Analysis on the expansion of the child tax credit during the Covid-19 pandemic found that increasing the benefit amount and making it available to more people lifted 3.7 million children out of poverty. In contrast, the addition of work requirements to enroll in the Supplemental Nutrition Assistance Program (SNAP, formerly known as food stamps), as proposed by Congressional Republicans early this year, would have reduced enrollment by an estimated 52% if it had been enacted, by making it more difficult for families in need to access the benefits. Given that SNAP is credited with keeping 3.3 million children out of poverty every year, this change would have drastically increased the number of children in poverty. These outcomes have major ramifications, not just for the families involved, but for society at large, as the long-term impact of child poverty has been calculated to cost more than $1 trillion every year.

The message is clear: Seismic changes are afoot, and we must take action now to steer AI towards empowering people in need.

The potential for AI to help Americans access such public benefits as the child tax credit and SNAP is immense. Chatbots empowered by natural language processing could provide guidance for benefits applicants 24/7, offering assistance far beyond what’s available from the often hour-long hold times on benefits hotlines. AI chatbot systems have already shown great potential for exhibiting empathy, something that overburdened social workers may struggle to do, and could be designed to offer support in whatever language applicants are most comfortable in writing or speaking.

AI systems could also greatly simplify the task of digesting the lengthy, complicated eligibility rules and paperwork requirements that make enrollment so daunting, synthesizing the morass of support programs and red tape into straightforward explanations. This could improve the experience both for benefits applicants and for social program eligibility workers, who themselves may struggle to stay abreast of program benefits and guidelines, particularly as new rules come into effect at both the state and federal levels.

AI programs are already being employed to reduce paperwork in the medical arena, and the same methodology could be adapted to the benefits application space as well, asking applicants a handful of questions in plain English and completing the necessary forms based on their responses. This could dramatically expedite the application process, as well as reduce the chance of paperwork errors, which often lead to further delays and the rejection of applications.

Despite this promising potential, relying on AI as a panacea risks overlooking deeper systemic problems underlying challenges to accessing benefits. While new technology could help with streamlining, issues like intentionally burdensome enrollment processes, understaffed benefits offices, and ingrained stigma around receiving public benefits would still need to be addressed. There’s also the serious concern of encoded bias. If AI is trained on historically biased sources, it will reproduce that bias. Without thoughtful design, AI risks providing offensive or misleading guidance to women, people of color, and other marginalized groups.

Perhaps the greatest near-term challenge is that AI-driven job loss could overwhelm already overstretched benefits systems. With automation threatening many low-wage jobs, demand for benefits like unemployment insurance and food stamps may skyrocket. But without a substantial increase in the staff and technological capacity of benefits offices (not to mention increased funding for social safety net programs), applicant wait times and improper denials could worsen dramatically. Unionized jobs could be replaced by AI programs. And people losing employment to automation may also struggle to re-qualify for benefits amid restrictive eligibility rules like work requirements.

The message is clear: Seismic changes are afoot, and we must take action now to steer AI towards empowering people in need. With governmental systems slow to change, we risk being wholly unprepared for AI’s benefits access impacts. But by embracing this technology thoughtfully and pairing it with deeper reforms, we can begin to turn the tide against poverty and the lack of opportunity that for too long have plagued our society.