After you have developed a clear vision, established a clear owner and a Hub, and published a formal Equitable AI policy, your organization will have formalized its commitment to address equity and fairness in AI implementations. But how do you effectively engage all stakeholders you identified to make Equitable AI an organization-wide priority? Stakeholder engagements will vary based on which stakeholder types are being engaged. For example, engaging a Champion might involve having them promote the Equitable AI Policy and resources in their business unit. Engaging Implementers might involve having teams complete training in Equitable AI and conduct a work session with their project teams.
Leaders need to go in with “eyes wide open” and should recognize the realistic challenges to integrating Equitable AI into the fabric of their organization. As you prepare to engage your stakeholder community, the staff in your Hub should use a community-driven approach to shape your organization’s strategy and put resources in place. The goal is to establish an internal culture of trust, transparency, and accountability to effectively implement Equitable AI.
Enlist the commitment, expertise, and support of the Advisors and Champions that you identified in Play 4 to help develop your approach and resources. You may also find success in recruiting external researchers, consultants, and organizations who can guide you, especially when you don’t yet have in-house resources in place.
Tips to Get Started
- Shape your community. Before engaging the stakeholders you’ve identified, decide how you want to structure your community. Your Champions, working with Implementers on implementation teams, will serve as the “spokes” of your “Hub-and-spoke” model. They will provide feedback and ideas back to the Hub discussed in Play 4. There are different ways to organize the implementation teams. You could create teams organized according to the Equitable AI principles you adopted (for example, one team could focus on Data Privacy and another on Equitable Design) or you could align the teams so that different teams are focused on different types of AI or potential use cases for the AI (for example, one team could focus on talent acquisition use cases). Consider having a Sensitive Use Cases working group to handle high-risk, highly sensitive use cases of AI-enabled technologies.
- Start small. Keep your scope manageable by prioritizing any high-priority use cases you want your implementation teams to focus on. You can always expand which use cases your implementation teams will tackle over time as your Equitable AI initiative expands.
- Prioritize stakeholders. You may want to engage stakeholders with more expertise first to help shape your community, lead implementation teams, and add research to your knowledge base or repository. This could include engaging external consultants who have experience in Equitable AI.
- Enlist your employee resource groups (ERGs). Be sure to prioritize your recruitment of a diverse, cross-sector group of stakeholders with diverse abilities and lived experiences—whether based on disability, race, gender, or other identities. Involve your existing ERGs—including your disability ERG—to help engage diverse employees in your initiative. ERG contributors can provide overall advice, help plan and design AI implementations, and participate in audits or pilots. (The most successful ERGs compensate their leaders or contributors).
- Consider motivations: Before you engage stakeholders, identify what motivates them to participate in the Equitable AI initiative. Defining motivations ahead of time will help you win quicker buy-in. For example, shareholders or investors might have concerns about the company’s public image, while consultants might hope the organization’s Equitable AI efforts will help provide a blueprint for others in the industry to implement a similar Equitable AI program.
- Cultivate your knowledge base and gather tools. Build and maintain an easily accessible knowledge base where AI implementation resources can be stored, including specific resources and tools related to AI equity and fairness. Link to internal policies, procedures, and implementation records, as well as external resources specific to mitigating bias and improving AI fairness for diverse users, including people with disabilities.
- Provide employees and other impacted parties with a safe way to ask questions and raise concerns. Offer a way for employees and other impacted parties such as potential employment applicants to ask questions or raise potential ethical concerns about a particular AI-enabled technology being implemented. Make sure that such questions or concerns can be communicated without risk of retribution or retaliation, and consider how to handle the feedback, including when it is about an AI implementation that the organization is profiting or realizing cost savings from.
Tips for Ongoing Operations
- Create a professional development program for Implementers. Design an incentive-based, self-paced professional development program to help Implementers gain new skills, understand user experiences, build confidence around implementing Equitable AI, and gain mentors. Professional development programs provide a means of recognizing strong performers via employee digital badges, congratulatory spotlights, and more. Examples of professional development programs include:
- Prepare a training curriculum and materials. Put staff training in place to ensure all employees understand their organization’s vision, policies, initiative structure, and resources for implementing Equitable AI. Integrate the training as part of your Equitable AI Champions Program and link it to larger organization-wide initiatives such as Diversity Equity, Inclusion, and Accessibility (DEIA) and Data Privacy. See the Tipsheet: Developing Staff Trainings for Equitable AI.
- What stakeholder communities can best shape your Equitable AI initiative based on your organization size and structure?
- Does your organization have ERGs in place already? Do you have an ERG for employees with disabilities and their allies?
- What staff training and support resources are already in place in your organization? Do you have a repository where staff training, research and other resources are stored?
- Does your organization have an incentive-based training and professional development program?
- Has your organization developed or procured training in AI or AI Ethics?
- Christopher Pappas, 7 Interesting Ways to Incorporate Badges in Online Training, eLearning Industry, 2021.
- Kathy Baxter, How to Build Ethics into Design, Salesforce Design (Medium), 2018.
- Christina Zhang, Why we invest in Ethics by Design, Salesforce Blog, 2021.
- Reid Blackman, “A practical guide to building ethical AI”, Harvard Business Review, 2020