"This new era of AI has presented a once-in-a-century chance to wipe away a lot of the damage and renew the mission. Not to the moon, but to a more perfect union," writes Josh Tyrangiel in his recent Washington Post article Let AI Remake the Whole of Government.
Tyrangiel correctly goes on to argue that few have appreciated what AI could do to transform how government works. "Large-language-model-powered chatbots could already be providing better service — at all hours, in all languages, at less cost — for people who rely on the federal government for veterans benefits, student loans, unemployment, social security and Medicare."
This is just the tip of the iceberg. AI is a set of tools for processing large quantities of data. The piece makes the case for taking advantage of this power to get real-time awareness of how the government delivers services in an effort to streamline and improve efficiency.
It is great to have so many column inches devoted to the topic of improving governance, which rarely rises above the fold in media coverage focused on the sprint of politics, rather than the marathon of governing. I appreciate the fact the Tyrangiel (citing Eric Schmidt) demystifies AI, which is just another form of software that we have to keep improving, learning as we go. It's not a one-and-done. With more and better data, enabled by AI, we can define the problem we are trying to solve and iteratively tackle it with technology.
What the piece misses, however, is that understanding a problem cannot come from data alone. Central to the tenets of agile software development is the need to talk to those most affected, if we want to understand why a problem is actually happening and how people experience it. Design and development must be human-centered as much as data-driven.
The piece is strangely silent about much of the exciting experimentation underway with AI, especially at the state and local level, where agencies are using it to listen to and understand comments and input from residents. In New Jersey, the Office of Innovation takes advantage of the power of generative AI to summarize and synthesize to allow us to hear from residents and workers alike with greater efficiency.
He invokes the work of Jen Pahlka but without mentioning that she improved service delivery in California, not with a data dashboard, but by talking to and listening to Californians about their experiences applying for government services (Her human-centered work features centrally in our course on innovation skills for public sector professionals here.)
This abstract view of data partly explains the focus on the work of the data-mining company Palantir as his model for agile, data-driven practice but one that is devoid of human participation or collaboration. There's also no mention of the ethical concerns arising from Palantir's projects that involved, for example, mining data to drive Trump era immigrant deportations, that have made the company so controversial.
The article also lets stand unexamined a quote from General Perna (who ran Operation Warp Speed) questioning: "Why do we need 80,000 new people at the IRS?" inadvertently playing into a political narrative that government-is-the-problem and therefore we should shrink the public sector. This notion that data dashboards can replace workers fundamentally misunderstands the need for human insight, creativity, intelligence and experience. If anything, government today is woefully understaffed and lacking in time and capacity to handle today's complex and myriad challenges. Anyone who has ever been on hold with an automated phone system knows that we need to use AI as a helpmeet to free up time for public servants to focus on helping people and solving problems, not to replace them.
