AI Ethical Principles
The Department of Defense (Department or DoD) adopted the AI Ethical Principles in February 2020.
The Five AI Ethical Principles for the Department are:
DoD personnel will exercise appropriate levels of judgment and care, while remaining responsible for the development, deployment, and use of AI capabilities.
The Department will take deliberate steps to minimize unintended bias in AI capabilities.
The Department’s AI capabilities will be developed and deployed such that relevant personnel possess an appropriate understanding of the technology, development processes, and operational methods applicable to AI capabilities, including with transparent and auditable methodologies, data sources, and design procedure and documentation.
The Department’s AI capabilities will have explicit, well-defined uses, and the safety, security, and effectiveness of such capabilities will be subject to testing and assurance within those defined uses across their entire life-cycles.
The Department will design and engineer AI capabilities to fulfill their intended functions while possessing the ability to detect and avoid unintended consequences, and the ability to disengage or deactivate deployed systems that demonstrate unintended behavior.
What is Responsible Artificial Intelligence (RAI)?
RAI is a holistic, integrated, and disciplined approach that focuses, not only on the technology, namely operationalizing the DoD AI Ethics Principles via good engineering practices, but on organizational operating structures and culture necessary to advance the responsible design, development, procurement and use of AI. Responsible AI provides Service members, the American public, and DoD partners and allies the confidence that DoD’s AI-enabled systems will be safe and reliable and will adhere to ethical standards.
Implementing Responsible Artificial Intelligence in the Department of Defense
In May 2021, Deputy Secretary of Defense Dr. Kathleen H. Hicks executed a memorandum outlining the Department of Defense’s commitment to responsible AI. In this memo, Secretary Hicks outlined tenets to advance and operationalize the DoD’s AI Ethic’s Principles. With an objective to build a trusted ecosystem of AI tools and subject matter expertise, the tenets outlined in the memo will enable the Department to enhance military capabilities and build confidence with our workforce, warfighters, and the American public by utilizing AI in a responsible manner.
RAI Implementation Tenets:
- RAI Governance
Ensure disciplined governance structure and processes at the Component and DoD-wide levels for oversight and accountability and clearly articulate DoD guidelines and policies on RAl and associated incentives to accelerate adoption of RAI within the DoD.
- Warfighter Trust
Ensure warfighter trust by providing education and training, establishing a test and evaluation and verification and validation framework that integrates real-time monitoring, algorithm confidence metrics, and user feedback to ensure trusted and trustworthy Al capabilities.
- AI Product and Acquisition Lifecycle
Develop tools, policies, processes, systems, and guidance to synchronize enterprise RAI implementation for the AI product throughout the acquisition lifecycle through a systems engineering and risk management approach.
- Requirements Validation
Incorporate RAJ into all applicable AI requirements, including joint performance requirements established and approved by the Joint Requirements Oversight Council, to ensure RAI inclusion in appropriate DoD AI capabilities.
- Responsible AI Ecosystem
Build a robust national and global RAI ecosystem to improve intergovernmental, academic, industry, and stakeholder collaboration, including cooperation with allies and coalition partners, and to advance global norms grounded in shared values.
- AI Workforce
Build, train, equip, and retain an RAI-ready workforce to ensure robust talent planning, recruitment, and capacity-building measures, including workforce education and training on RAl.
JAIC Pilots a Responsible AI Procurement Process with Responsible AI Institute
JAIC is working with RAII to pilot a procurement review process to ensure AI acquired via Tradewind is aligned with the DoD’s AI Ethical Principles. The intent is to identify entry points for RAI integration into the AI Acquisition process with a combination of tools and reviews for oversight and accountability. Details of the pilot are forthcoming.
Responsible AI RFI 2021
In late-June, the Joint Artificial Intelligence Center (JAIC) issued a Request for Information (RFI), through its AI procurement ecosystem, Tradewind. The RFI called on responsible AI (RAI) subject matter experts and organizations to provide input on knowledge, tools, services, and solutions to aid in supporting and advancing the six RAI Implementation Tenets set forth by the Deputy Secretary of Defense in May. Below is a summary and corresponding infographic illustrating the findings from the RFI process and responses.
After a 3-week open request period, the RFI closed in July, with 75-plus responses, 60-plus attendees over two information sessions, and a host of questions received via email. Additionally, the RFI resulted in a diverse set of respondents from five different industries: technology and solutions providers, consulting and advisory groups, academia and non-profits, individual subject matter experts, and a law firm.
A preliminary review of all RFI responses yielded a set of findings to aid in advancing pragmatic and process-based avenues for RAI integration in the Department:
- The findings also revealed that the ecosystem of RAI organizations and subject matter experts was wide and responsive and that they have diverse backgrounds, profiles, and expertise – further evidencing the multidisciplinary nature of the field. Findings also showed that respondents had a breadth of experience serving the Government, including representatives from all Service branches and a dozen other DoD offices such as the White House, the Department of Justice, the Federal Bureau of Investigation, the Department of Homeland Security, and the National Institute of Standards and Technology (see Finding 3).
- Most respondents showed support for at least one of the DoD RAI Implementation Tenets, with responses spanning all six: RAI Governance, Warfighter Trust, AI Product and Acquisitions Lifecycle, Requirements Validation, Responsible AI Ecosystem, and AI Workforce (see Finding 4).
- Many respondents addressed more than one focus area identified in the RFI, spurring confidence in focus areas ranging from research and development to documentation and information management (see Finding 5).
The information gleaned from the RFI provides robust market research and an understanding of the state of the art in this space, allowing the Department to build a strong RAI ecosystem. The information gathered will help guide the direction and efforts for the Department in advancing RAI implementation, including future development of AI capabilities and solution sets.