AI accountability in public administration
Federal agencies in the United States are embracing the use of Artificial Intelligence (AI) to improve service delivery and public trust, while ensuring data security and privacy. This shift is in line with recent directives and executive orders from the Office of Management and Budget (OMB) and the Government Service Delivery Improvement Act.
The Social Security Administration (SSA) has been at the forefront of this change. By using AI to analyze website and contact center insights, the SSA uncovered a root cause of calls: difficulty returning to saved applications. To address this issue, the SSA added a "return to my saved application" button, which has helped reduce complaints and calls.
The OMB has issued directives and executive orders requiring federal agencies to adopt AI as a core component of their service delivery strategies. Agencies are now required to develop and maintain a minimum foundational knowledge of how to use AI. The AI adopted by federal agencies should be responsible AI, promoting transparency, accountability, and security.
To effectively implement responsible AI, federal agencies are adopting a multifaceted approach. Strengthening workforce capacity and governance involves recruiting and empowering AI experts with operational experience, designating Chief AI Officers, and providing extensive workforce training to ensure personnel understand AI’s capabilities and risks.
Adopting AI procurement and deployment frameworks is another key strategy. Using tools like a GSA-led AI procurement toolbox can guide agencies through AI acquisition and deployment processes in a way that emphasizes reliability, security, and ethical considerations. Agencies are also encouraged to set procurement standards that exclude AI models with ideological biases, promoting neutrality and public trust.
Balancing deregulation and innovation with security and privacy safeguards is another crucial aspect. The federal AI Action Plan advocates reducing overly burdensome regulations that hinder AI innovation while simultaneously expanding security assessments. Regulatory sandboxes enable real-world testing under controlled environments to identify risks and improve AI systems before wide deployment.
Building robust AI infrastructure and data quality standards is also essential. Investment in scalable AI infrastructure and assuring minimum data quality standards safeguards AI model training and helps prevent misuse of sensitive information.
Prioritizing transparency, trust, and incident reporting is critical. Building trust can be fostered by clear transparency about AI uses, establishing frameworks for reporting AI incidents, and involving multi-agency coordination to respond to AI-related challenges.
Ensuring data security and privacy is another crucial aspect. Agencies must integrate strong cybersecurity practices, conduct rigorous risk assessments, and comply with data privacy standards in all AI applications to protect sensitive personal and governmental data while maintaining public confidence.
By combining these strategies, federal agencies align with current U.S. government AI policy priorities that emphasize responsible deployment to enhance public service delivery and trust, balanced with safeguarding security and privacy considerations in an evolving technological landscape.
Agencies are also encouraged to use AI to balance the best of both AI and human interactions. For example, proactive outreach to website users experiencing issues and detecting citizen feedback related to scams can enhance the citizen experience.
Moreover, agencies must choose American-made AI technologies from vendors committed to responsible AI, with safeguards to protect sensitive data and personally identifiable information. Agencies should train their AI models to ensure the right outcomes and intervene when necessary to prevent unintended results.
The IRS's smart callback option, which was rolled out during the 2023 tax filing season, is in line with consumer expectations for shorter wait times during customer service interactions. The Department of Veterans Affairs (VA) has also implemented AI to better understand the veteran experience, identify veterans at risk of self-harm, suicide, and homelessness, and improve service delivery based on AI-derived insights.
Most consumers prefer a call-back option over waiting on hold, according to a 2023 Medallia study. Federal agencies can identify friction in the user experience to uncover improvements that can be made using AI, such as removing bureaucratic bottlenecks or enhancing the end-user experience. A 2024 Medallia study found that shorter wait times and call-back options are key to customer satisfaction during customer service interactions.
In conclusion, the adoption of AI by federal agencies can help optimize operations, reduce inefficiencies, and strengthen public trust. Agencies are required to leverage commercial off-the-shelf American-made, cost-effective AI solutions to accelerate federal efficiency, remove unnecessary bureaucratic requirements, improve self-service, enhance overall federal services, eliminate "avoidable waste" and "costly delays", and ensure they're listening to the public.
Technology, fed by artificial-intelligence (AI), has become a core component in the service delivery strategies of federal agencies. This shift, towards adopting AI, aims to promote transparency, accountability, and security in the AI solutions being utilized, while ensuring data security and privacy.
Agencies are encouraged to choose American-made AI technologies that come with safeguards to protect sensitive data and personally identifiable information, and to train their AI models to ensure the right outcomes and intervene when necessary to prevent unintended results. This focused application of AI serves to optimize operations, reduce inefficiencies, and strengthen public trust.
