The Future of AI For the Public Sector

Back in September of 2024, the IBM Center for The Business of Government gathered a group of 30 experts in the realm of Artificial Intelligence, to consider and discuss the benefits and challenges of AI in the public sector. Over the course of the last several months, the wisdom garnered from that exciting gathering has been distilled in a series of blog posts. The first covered the potential benefits of AI; the second delved into some of the concerns and solutions about its use and the third, which ran on January 8, focused largely on the issues of bias that are of concern to many observers.

The overarching sentiment expressed by the roundtable participants was this: The use of Artificial Intelligence as a driving force in the way the public sector functions is going to increase dramatically in the coming months and years. The benefits that are already on the horizon are dramatic, but no one has the ability to predict with any degree of reliability what this technology will be able to accomplish in the future.

In an environment where this kind of forceful change is moving at a rapid clip, with no certainty of where it will go, the participants pointed out that it’s critical to tread with caution and beware of the ways in which AI can conceivably create problems as well as to help solve them.  

Among the major benefits that the roundtable discussed were AI’s capacity to:

• Help translate complex documents into plain and easy to understand English, offering more access to information that might otherwise be indecipherable. 
• Aid governments’ quests to fully understand the needs of people. In the past, this effort was largely dependent on self-reporting by citizens through 311 or other means. But AI’s ability – coupled with technology like on-street cameras – can help to catch issues that beset the public and fix them before they become a major problem.
• When public sector staff is confronted with complex queries from the public, AI has high utility to provide staff at call centers with easy access to accurate answers.
• Not only can AI ease the interactions between governments and and the people they represent, it can also simplify their access to services that are provided by non-government support organizations.
• The critical search for grants, which has historically been a tangle of confusing funding and application processes, can be simplified in a way that will help individuals, organizations, and states and localities themselves to cut through the thicket of available resources, finding available funds, and helping to simplify the process of applying.
• Helping individuals to benefit from social programs before a small problem becomes a crisis. Early detection of mental health problems was an area that was seen as ripe for this kind of use.
• A host of other uses including providing faster call responses for taxpayers; making sure that the findings of administrative representatives, such as health facility inspectors, are supported by state regulation; enabling easier access to data for governments; educating citizens and uncovering discrimination.

Among the challenges that the participants singled out as significant included the expense of the implementation of AI. As one participant said, “Not many governments have $200 million to build their own large language model.” Additionally, the lack of interoperability and standardized definitions in the various data sets used in AI operations was mentioned repeatedly.

These challenges are far from insurmountable, however, and the panelists offered up a series of solutions to them, most notably being the need for a solid governance structure that will help to make sure that sufficient oversight is in place for AI, particularly since advances in the field will largely be driven by private sector firms that may not be aware of the particular concerns of a public sector organization.

This kind of governance helps create efficiencies that reduce expenses and provide incentives to others to make investments in AI that can hasten its development in government.

For governance like this to work well, it’s clear that centralization is going to be key, whether under the leadership of chief AI officers or other means. While such positions are still relatively rare, chief data officers are likely in a position to take that role. Fortunately, federal agencies have created those jobs, and, as they develop, people in them will likely have the capacity to consider AI as a major part of their roles.

But although expenses and interoperability were significant areas of concern to participants, the most significant challenge was the potential of unintentional biases finding their ways into the algorithms that underly AI.

Members of the roundtable indicated that governments must be proactive in ensuring that biases aren’t influencing the decisions they reach with the use of AI – even when the actual systems have been developed by outside contractors.

One way to accomplish this is by ensuring that human beings don’t yield important decisions to AI carte blanche but use this new technology primarily as a source of information – which can be vetted for biases before it’s used to come to any conclusions that influence the work of government.

Naturally, given the speed with which AI is advancing, this is going to require dedication to training any public sector employees who are using AI about the potential for hidden biases, and how to detect them.

One area of particular concern brought up by the roundtable group is the use of AI to turn text into images, notably photos of human beings. Current predictions find that a huge percentage of content on the internet will be produced by AI – including such photographs. In policing, for example, using biased text-to-image AI can be used to create sketches of suspected offenders that could lead to wrongful arrests and even convictions.

As the use of AI in the public sector accelerates, more benefits and challenges will undoubtedly emerge. Keeping close watch over them will be essential for public sector leaders. As Shakespeare wrote, “It is not in the stars to hold our destiny, but in ourselves.”