Responsible Artificial Intelligence (AI) Program
United State Code defines “artificial intelligence” (AI) as “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.” (15 USC 9401(3)). Other definitions include concepts such as algorithms and automated decision-making systems.
AI can certainly produce new opportunities to help solve the City’s urgent challenges, improve City services, and increase the City’s responsiveness. However, irresponsible use of this technology has the potential to exacerbate problems such as misinformation and bias, and could exacerbate societal harms such as fraud, discrimination, and infringements on privacy. Realizing the benefits of AI requires mitigating its substantial risks.
The City of Seattle is committed to using technology in a manner that upholds the City’s principles, policies, commitments, and all applicable laws and regulations. To that end, with regard to AI, we have enacted guiding principles and policies, and are implementing guidelines and programs to support City employees in using AI responsibly.
The City’s Policy on Generative AI
Generative AI is a class of artificial intelligence systems, including algorithms, deep-learning, and machine learning models, capable of generating new content, including but not limited to text, images, video, and audio, based on the inputs of training datasets. These also include systems capable of ingesting input and translating that input into another form, such as text-to-code systems.
The City of Seattle released its Generative Artificial Intelligence (AI) policy [PDF] in October 2023. The policy was developed with input from regional AI experts, university researchers, and members of the Community Technology Advisory Board (CTAB). The policy governs the use of Generative AI based on the City’s principles. This allows for flexibility for City employees to use new technology while ensuring City employees adhere to the guidelines.
The policy requires that City employees acquire generative AI technology through the City’s approved procurement channels, which will include a review step to determine if the AI technology is consistent with the City’s standards. It also directs City employees to ensure that there is a “human in the loop” to review output from generative AI technologies, and that the review process is documented. The policy also requires employees to attribute any content created by AI and published on behalf of the City to the AI system that produced it. Finally, the policy directs employees to ensure that any data or content submitted to AI systems or generated by AI systems be free of harmful bias, control for potential data privacy concerns, and comply with the State of Washington Public Records Act and all relevant City policies concerning records retention and disclosure.
Artificial Intelligence Principles
The City of Seattle has adopted the following principles regarding AI systems in general. These principles describe general codes of conduct that represent the City’s values and are aligned with our responsibilities to the residents we serve. These principles serve to guide City employees in their use of both generative and traditional AI technology.
- Innovation and Sustainability: The City values public service innovation to meet our residents’ needs. We commit to responsibly explore and evaluate AI technologies, which will improve our services and advance beneficial outcomes for both people and the environment.
- Transparency and Accountability: The City values transparency and accountability and understands the importance of these values in our use of AI systems. The City will ensure that the development, use, and deployment of AI systems are evaluated for and compliant with all laws and regulations applicable to the City prior to use, and will make documentation related to the use of AI systems available publicly.
- Validity and Reliability: The City will work to ensure that AI systems perform reliably and consistently under the conditions of expected use, and that ongoing evaluation of system accuracy throughout the development and/or deployment lifecycle is managed, governed, and auditable, to the greatest extent possible.
- Bias and Harm Reduction and Fairness: We acknowledge that AI systems have the potential to perpetuate inequity and bias resulting in unintended harms on Seattle residents. The City will evaluate AI systems through an equity lens, in alignment with our Race and Social Justice commitments, for potential impacts such as discrimination and unintended harms arising from data, human, or algorithmic bias to the extent possible.
- Privacy Enhancing: The City values data privacy and understands the importance of protecting personal data. We work to ensure that policies and standard operating procedures that reduce privacy risk are in place, and are applied to the AI system throughout development, testing, deployment, and use to the greatest extent possible.
- Explainability and Interpretability: The City understands the importance of leveraging AI systems, models, and outputs that are easily interpreted and explained. We work to ensure all AI systems and their models are explainable to the extent possible, and that system outputs are interpretable and communicated in clear language, representative of the context for use and deployment.
- Security and Resiliency: Securing our data, systems, and infrastructure is important to the City. We will ensure AI systems are evaluated for resilience and can maintain confidentiality, integrity, and availability of data and critical City systems, through protection mechanisms to minimize security risks to the greatest extent possible, in alignment with governing policy and identified best practices.
Next Steps
In 2024, Seattle IT will lead the development of an Artificial Intelligence Program that will provide further assistance and guidance to City employees on the responsible use of AI technology.
As part of that initiative, Seattle IT will host a Community of Practice for City employees with an interest in AI technologies to exchange information, best practices, use cases, and other experiences.
The program will also help develop specific evaluation and purchasing guidelines and frameworks for assessing AI technologies that are proposed for use in conducting City business.
Regional and National Collaborations
The City of Seattle is working with other government agencies, universities, and organizations across the country to develop and evolve approaches for using AI in a responsible manner.
Formed by the City of San Jose, CA, the Government AI Coalition is collaborating with government agencies across the country to develop standards that promote effective, equitable, trustworthy AI systems to better serve our communities. The GovAI Coalition is committed to using AI for social good, ensuring ethical and responsible AI governance, promoting vendor accountability, realizing improved efficiency, and fostering cross-agency collaboration and knowledge sharing.
MetroLab Network Generative AI for Local Governments Task Force
The City’s Interim Chief Technology Officer, Jim Loter, is co-chair of this initiative, which is composed of individuals from cities, counties, non-profits, universities, and metropolitan planning organizations from around the country. This task force is considering policy with a two-pronged approach: 1) What is the world of the possible and how can GenAI better city services; and 2) What processes and policies should be put into place to ensure a just, equitable, and accurate use of this technology.
State of Washington’s AI Community of Practice
Seattle IT staff are engaged with WaTech’s AI Community of Practice, which continues to work to provide guidance for the development, procurement and implementation of Artificial Intelligence (AI) technologies by Washington state agencies. further develop artifacts to guide AI use in the state. The State of Washington issued their AI guidelines in August 2023.
Other News and Resources
Seattle IT’s Interim CTO Jim Loter talks AI on GeekWire podcast