HHS Releases Trustworthy AI Playbook, Website

The Department of Health and Human Services has released its artificial intelligence playbook to provide high-level information about trustworthy AI and guidance for deploying AI across its typical lifecycle.

HHS published its Trustworthy AI (TAI) Playbook, alongside the launch of its AI website, nearly a year after the agency released its AI Strategy and appointed Oki Mek as its first chief AI officer. The new playbook looks to meet requirements from OMB M-21-06 and Executive Order 13960, which call on agencies to encourage AI innovation and growth while establishing ethics and principles around the technology.

The playbook aims to guide HHS leadership to create policies around TAI and evaluate risks associated with AI investments, while also highlighting how program and project managers can incorporate it properly, work with teams before building AI solutions, oversee projects throughout their lifecycle and mitigate risks.

Rather than cementing a formal policy or standard, the playbook paints broad strokes for AI adoption to promote White House TAI principles, centralize relevant federal and non-federal resources, and set a framework for smart adoption throughout the AI lifecycle and future use.

“HHS has a significant role to play in strengthening American leadership in artificial intelligence,” Mek said in the playbook. “As we use AI to advance the health and wellbeing of the American people, we must maintain public trust by ensuring that our solutions are ethical, effective and secure.”

The playbook highlights six principles that the agency should apply across “all phases of an AI project,” which call on HHS’s AI to be fair and impartial, transparent and explainable, responsible and accountable, robust and reliable, safe and secure, and stewards of privacy.

From those principles, the playbook looks across the AI lifecycle, from initiation and conception through deployment and operational maintenance to underscore how to apply those principles across common and critical steps for AI adoption.

The playbook provides various use cases across various stages of the lifecycle to indicate how to apply the principles in steps like designing solutions and evaluating model risk factors. These use cases include automated medical document processing, AI for medical billing fraud detection, chatbots for customer service and more.

Mek hopes that this information the playbook provides will increase success rates of AI projects across HHS.

“As many studies have shown that about 85% of AI projects fail, the playbook helps guide the build of trustworthy AI and will help increase the success rates,” Mek said in a social media post announcing the HHS AI website and playbook. “To truly realize the benefits of AI, we must communicate, share and collaborate across the health sector. AI provides us with an immense instrument to keep the flame of intelligence glowing even brighter.”

With the playbook in place, HHS will focus on further collaboration and dialogue around AI in 2022. These include:

• Cultivating the HHS AI community of practice
• Hosting further AI lunch and learn sessions for HHS personnel
• Standing up the HHS AI Council to execute the HHS AI strategy
• Expanding the playbook as needed to promote ethical and trustworthy AI
• Developing an AI use case inventory