Staff note: Artificial intelligence (AI) is changing fast, bringing exciting opportunities and some tricky challenges. One key issue is making sure AI is developed responsibly and inclusively, with meaningful input from diverse stakeholders. At CSCW 2024 in San José, Costa Rica, TIPI’s Azza El-Masri joined a workshop organized by the Partnership on AI (PAI). This event was a chance to discuss important challenges and share ideas on how to engage various stakeholders in AI development. Participants included academic teams, non-profits, tech companies, and government entities from around the world. Read below for some insights from Azza.
Partnership on AI: A Hub for Multistakeholder Collaboration
Founded in 2016, PAI is a non-profit organization dedicated to addressing some of the most pressing questions surrounding AI. Operating remotely with 107 partners worldwide, PAI focuses on five key program areas:
- AI & Media Integrity
- AI, Labor & The Economy
- Fairness, Transparency & Accountability
- Inclusive Research & Design
- Safety-Critical AI
PAI’s mission is rooted in fostering multistakeholder collaboration, with a commitment to cultivating community, identifying ethical tensions that impact marginalized groups, and producing actionable resources to guide AI practitioners. For instance, their efforts in the AI & Media Integrity program have led to impactful collaborations that address the spread of misinformation, ensuring that vulnerable communities are not disproportionately harmed by AI-driven content. A landmark development in this journey was the establishment of the Global Task Force on AI in June 2023, which played an active role in this workshop.
Guidelines for Inclusive AI Development
The workshop focused on applying draft guidelines by the Global Task Force that help practitioners turn ethical principles into everyday practices, particularly in engaging marginalized communities. These guidelines underscore that stakeholder engagement and participatory design are not a one-size-fits-all endeavor but a nuanced process requiring adaptability, empathy, and careful planning. Key recommendations include:
- Specify Stakeholder Groups: Tailor engagement strategies to align with specific stakeholder needs and avoid treating representative samples as universally applicable.
- Design from the Margins: Center the needs of those least likely to be included in conventional conversations.
- Prepare in Advance: Build foundational knowledge about stakeholder communities before initiating engagement.
- Emphasize Community Voices: Prioritize insights from individuals with lived experiences within the communities being addressed.
- Align Expertise and Knowledge: Recognize diverse forms of expertise and actively recruit participants who reflect the needs of the project.
- Maintain Clear Communication: Be transparent about the scope and limitations of stakeholder engagement and ensure consistent follow-up.
- Adapt and Iterate: Remain flexible and invest in refining engagement methods based on participant feedback.
- Document Thoroughly: Keep comprehensive records for accountability and future reference.
Bridging the Gap: Workshop Activities and Scenarios
The workshop also featured scenario-based group activities, facilitated by members of the Global Task Force on AI. Each scenario presented real-world challenges, such as:
- Expanding natural language processing (NLP) to underserved languages like Kiswahili and Zulu.
- Designing AI tools for sensitive use cases, such as chatbots supporting survivors of domestic abuse.
- Mitigating harmful outputs in generative AI models to ensure inclusivity for people with disabilities.
- Ensuring the usability and adoption of AI-driven scheduling systems by small businesses.
Through these exercises, participants explored the complexities of stakeholder engagement, such as balancing limited resources with the need for thorough community consultation and addressing barriers to participation. For example, in the scenario focused on expanding natural language processing to underserved languages, participants grappled with challenges like sourcing diverse and representative datasets while navigating resource constraints and cultural sensitivities.
Key Takeaways
One of the most resonant messages from the workshop was that “every moment is a moment to talk to people.” Stakeholder engagement is not a discrete event, but an ongoing process of dialogue and learning. By incorporating diverse perspectives at every stage, AI practitioners can create systems that better serve society as a whole. This might involve engaging with underrepresented groups to understand their unique challenges, co-designing solutions with community members, or iterating based on feedback to ensure the technology addresses specific needs effectively.
Figure 1. Common Applications of Stakeholder Engagement during the AI Development Lifecycle. Diagram by the Partnership of AI’s Global Task Force.
Looking Ahead
The PAI workshop at CSCW 2024 reinforced my belief that responsible AI development requires more than technical expertise; it demands empathy, collaboration, and a steadfast commitment to inclusivity. As TIPI continues to engage with these critical issues, we look forward to contributing to ongoing conversations and fostering connections that drive meaningful change in the AI landscape, particularly on as we continue to engage with municipal governments on ethical AI practices.