The Partnership for Public Service, in collaboration with the IBM Center for the Business of Government, convened federal service delivery leaders for a roundtable conversation on ethical artificial intelligence and benefits delivery.
Blog Author: Cole von Glahn, Manager, Technology and Innovation, Partnership for Public Service. This blog was first published on the Partnership for Public Service website.
During the conversation, we heard from a range of leaders on how they are approaching the possibilities and challenges AI presents for government. Here are the main takeaways from their discussion.
AI is an impact multiplier
AI can enable federal programs to deliver benefits more quickly and accurately, and its potential for code generation could revolutionize government’s ability to build and maintain IT infrastructure. By automating rote tasks, AI could also clear space for civil servants to think creatively and interactively with their customers on how to hone processes for a more effective government.
However, just like the humans who make and use it, AI’s impact can be negative and its unintended consequences harmful. These harms are multiplied and scaled similarly to its benefits. Irresponsibly implementing AI systems may result in systemic bias, inaccurate outputs and performance degradation.
We have the opportunity to thoughtfully and decisively address questions of fairness, explainability, security, transparency and privacy. Standing in the way are constraints on workforce development, institutional knowledge needs and the capital required to execute.
AI-ready government
A well-trained and empowered federal workforce is critical to an effective government, and AI implementation is no exception. Agency leaders identified organizational culture, internal knowledge gaps, staff engagement and restrictive hiring authorities as key barriers to the ethical implementation of AI for service delivery.
Remedying organizational culture starts at the top. Equipping leaders with a technical and ethical understanding of AI’s potential empowers them to direct their organizations on its responsible implementation with confidence and clarity.
The same attention should be paid to developing individualized learning pathways for civil servants. Training would support current staff as they become partners in AI implementation, and applying their institutional, domain and program expertise to AI use cases would save jobs, improve outcomes, mitigate harms and ease the transition for technical hires.
In addition, participants consistently reiterated the need to loosen hiring restrictions or risk losing high quality candidates. Cross-functional hiring teams with technical, HR and policy knowledge can improve our government’s ability to attract and retain technologists. Loosening hiring requirements and leveraging cross-functionality would help government compete with private firms in this talent market.
In each of these cases, collaboration is critical. AI is an all-of-government opportunity that can only be ethically and impactfully achieved by leveraging the full diversity of skills and experiences available to the federal government.
Resources
Leaders underscored that they lack the money and mandate necessary to achieve responsible AI goals.
While President Biden’s AI executive order calls for adopting many critical workforce measures and sets admirable ethical implementation goals, it does not provide the funding needed to act. The order alone will not unify AI’s fragmented policy and legal landscape, which hampers agencies that are ready to move forward. To achieve their AI goals, agencies need Congress to appropriate long-term, predictable funding and lay out clear frameworks for use.
An urgent call to action
As participants discussed these barriers, they emphasized the level of urgency that surrounds operationalizing AI.
Moving forward without resolving these foundational issues limits our government’s capacity to consider deeper concerns regarding responsible delivery. Solving them means having the talent and time to address those concerns.
Government can’t move fast and break things. It must instead hold itself and its partners accountable for the responsible implementation of AI that improves services and protects the public.
Read our next blog, "Artificial Intelligence Development for Government Agencies: Strategies for Ethical and Effective Implementation" in the series.