Steps for federal agencies to consider before using artificial intelligence
The use of artificial intelligence in the federal government is becoming more common, but many agencies are still trying to decide how this technology will create new efficiencies and help them better serve the public.
Earlier this year, the Partnership for Public Service and Booz Allen Hamilton hosted a workshop to discuss the state of AI in government as well as the benefits and tradeoffs of adopting this technology.
During that event, Suzette Kent, who until recently served as federal chief information officer, said the Office of Management and Budget and its Office of the Federal Chief Information Officer are supporting use of AI to help agencies achieve their missions and improve the delivery of citizen services. Kent said it will be necessary for agencies to make foundational investments in AI, and in particular, to make good use of data if they are to be successful at achieving mission goals in the future.
“The federal government has some of the best data in the world,” Kent said. “Everyone needs to understand the problem that your agency wants to solve while understanding the data inventory you have and the boundaries of using that data.”
The term artificial intelligence refers to machines and software that perform tasks typically associated with humans, such as recognizing speech or images, predicting events based on past information or making decisions. AI continues to improve at tasks such as transferring information from paper to computers, answering questions by finding relevant information in large databases, detecting patterns in troves of data and predicting someone’s behavior based on past conduct.
AI technology is evolving, with engineers finding new ways to train algorithms and create hardware architecture to optimize data science techniques. Panelists said this will require gaining a greater understanding of how machines learn and making continued investments in the technology.
The Partnership-Booz Allen Hamilton event featured a wide ranging discussion of issues agencies should be thinking about as they look to adopt AI:
- There is no one blanket definition of data science, but the goal should be getting from data to decisions. Data science is, broadly, the art of turning data into actions through a multi-disciplinary approach to identify patterns and make well-informed decisions.
- The most successful artificial intelligence projects have strong data. The role of AI should be curating data and recommending options. It frees people to be critical thinkers and decisionmakers.
- Understand the risks. When deciding the role of AI, agencies should consider the risks relating to workforce preparedness, the potential for adversaries to use AI for harm, the possible impact of future regulations and historical bias in the data that could skew the results.
- Know when it is appropriate to use AI. The idea of using AI to solve any problem is tempting, but agencies should evaluate whether it is excessive or too early to adopt. The “goldilocks zone” for AI solutions includes identifying objects and people in videos or pictures, translating speech and text from one language to another, and detecting fraud and anomalous behavior.
- Consider six key areas before you implement AI. Agencies should verify that they are ready to address such issues as analytics, data, analytic techniques, people, technology and culture before beginning the AI journey.
In addition to artificial intelligence, a suite of emerging technologies is also helping government become more effective. These technologies include immersive technologies and edge computing, and all three are featured in the Partnership and Booz Allen’s latest report on emerging technologies in government, “Cracking the Code.”
This post is authored by Katie Malague and Kathleen Featheringham. Katie Malague is vice president for government effectiveness at the Partnership for Public Service. Kathleen Featheringham is a principal/director of AI strategy and training at Booz Allen Hamilton.