In order to bring your Equitable AI vision to life and create an enforceable, internal mandate for change, your organization should create and publish a formal Equitable AI Policy. The policy should be communicated internally, and organizations could consider publishing an appropriate version externally.
The policy will provide a set of consistent rules or guidelines for your organization and employees to follow to achieve your business goals as outlined in your business case. The policy can also help your organization understand key motivations for adopting an Equitable AI Policy, such as for competitive advantage, strategic planning, strategic intervention, social responsibility, or industry leadership. An effective policy should provide clear guidance on how the policy supports market requirements, which executive(s) or organization(s) are accountable, how employees are empowered and protected, and knowledge sharing.
Organizations who do not implement a policy run the risks of implementing multiple AI technologies simultaneously in different departments, utilizing different approaches and business strategies.
Tips to Get Started
- Learn from early AI technology implementations and from other organizations. Use what you’ve learned from your organization’s early implementation of AI to inform your new or revised policies. Research how other organizations have developed and implemented their policies.
- Use your existing effective policies as models. Don’t recreate the wheel. Leverage policies already in place in your organization that have been effective, such as those related to data privacy, information security, or technology/product accessibility.
- Identify related IT policies, procedures, and infrastructure. Understanding current IT governance structures and infrastructure will make it easier to understand the points of intersection with your organization’s Equitable AI Policy.
- Include a section with examples of specific roles for leaders in your organization. The policy should delineate where key responsibilities fall. Consider the roles that Advisors, Champions, and day-to-day Implementors will play as all stakeholders are engaged.
- Align your policy to organizational social responsibility imperatives. Your Equitable AI Policy should align to your organization’s other social responsibility efforts. This is important for both internal reasons—such as an organizational commitment to Diversity, Equity, Inclusion, and Accessibility (DEIA)—as well as external reasons.
- Collaboratively develop and pilot the policy before full rollout. Your Hub should work with your Advisors and Champions to collaborate on development of the Equitable AI policy. Be sure to include stakeholders responsible for DEIA efforts in your organization. Pilot the policy in early AI implementations or in one department before rolling the policy out across your organization.
- Keep the policy simple: Avoid making the policy too academic. The policy should be easy to understand and written at the operational level.
- Bridge the information gap. AI and its implications may not be familiar topics for many within your organization. As you roll out your policy, make sure to provide basic training about AI, AI Ethics, and Equitable AI so everyone affected by the Equitable AI policy can share a common baseline understanding. Consider different ways to get the word out ranging from online documents, key company All Hands meetings, onboarding or other key channels.
Tips for Ongoing Operations
- Seek feedback on the policy. Be sure to include key stakeholders (usually as part of your organization’s Hub) who can field questions and accept feedback from the organization as it implements the policy.
- Update the policy as needed over time. The pace of change for AI technologies, as well as social responsibility expectations, can be fast. Your organization should plan to revisit and update your Equitable AI Policy annually or every couple of years.
Guiding Questions
The following questions should be considered as you develop your Equitable AI policy. These can also serve as a framework for the sections in your Equitable AI policy.
- What are the purposes and motivations for your Equitable AI initiative? For example, is your organization motivated by competitive advantage strategic planning, social responsibility, and/or by the desire to follow best practices as determined by industry leadership?
- What types of AI technology implementations are covered by the policy, and are certain types considered sensitive or high-risk?
- Which of your internal and external stakeholders are impacted by the policy? (Refer to your stakeholder mapping to help identify this.)
- What other organizational policies, principles or standards have been adopted that relate to your Equitable AI Policy?
- Who is ultimately responsible for Equitable AI in your organization? Mention your Hub and any leaders who own responsibility in your organization.
- What governance approach and structure will your organization use to help ensure equity in AI implementations, and how will it be maintained?
- How will your organization monitor and report on Equitable AI implementations, including training, process, procurements and audits?
- What training and support is being offered to the organization to help ensure successful implementation of Equitable AI? What training, support systems and incentives are needed to empower employees?
- How often will the organization revisit and update the Equitable AI policy, to aid in continuous improvement and to meet emerging internal workplace or social responsibility needs?
Resources
- Machine Learning (ML) Garage Ethics Framework, created by Digital Catapult’s Ethics Committee (licensed under a Creative Commons CC BY 4.0 international license)
- PlainLanguage.gov Guidelines
