Unmet Expectations and Aspirations: Community Organizations' Perspectives on Artificial Intelligence for Social Good Partnerships

Core Concepts

Despite significant intellectual contributions and labor from community organization members, their goals are frequently sidelined in favor of funders' and technology teams' priorities in AI for Social Good partnerships.

Abstract

The study explores the perspectives of community organization members involved in Artificial Intelligence for Social Good (AI4SG) partnerships. Key findings:

1. Participation in AI4SG projects is heavily influenced by external factors, especially funding agendas and the broader belief in the promise of AI technologies. Funders often determine the problems to be addressed, the solutions to be explored, and the project timelines, which do not always align with community organizations' goals.

2. Community organization members provide critical data access, community connections, domain expertise, and technical feedback that are essential for the success of AI4SG projects. However, their contributions are often overlooked.

3. While many community organization members expected tangible project deployments, only two out of the 14 projects studied reached the deployment stage. When projects fell short of expectations, community organization members sustained their belief in the potential of the projects, still seeing diminished goals as valuable.

4. To enhance the efficacy of future collaborations, participants shared aspirations for co-leadership starting from the early stages of projects, technical capacity building for end users, and a relationship-first approach that centers community organizations' needs and expertise.

The findings highlight the power asymmetries within AI4SG partnerships and call for stakeholders, especially funders and technology teams, to shift focus from the tool (AI) to the social issues at hand. The study proposes "data co-liberation" as a guiding principle to center community organizations' co-leadership in the ethical development of AI for social good.

Stats

"There's this perverse incentive of responding to the donor. There might be someone at a foundation who loves the idea of this platform and really wants to see it built. [...] And we end up building this platform for the funder with all the things the funder wants to see in it."
"For [our organization], the other idea was a priority. But in the context of the interaction with the [industry partner] on how to pick an idea, it was not prioritized in that way."
"To collect data, you need the ethical approval from the government, from the health industry, we have done that. We've gone through the pain. It is basically expensive to collect."
"I always point out how the training was doing because the first time we met, they brought very good metrics. I don't know, like, an accuracy rate of more than 90%. So I was very like, that sounds weird. Are we over-fitting? I feel like we are over-fitting."

Quotes

"Come to us first": Centering Community Organizations in Artificial Intelligence for Social Good Partnerships
"Data Feminism begins by analyzing how power operates in the world today."
"Embracing pluralism in data science means valuing many perspectives and voices and doing so at all stages of the process—from collection to cleaning to analysis to communication."

Key Insights Distilled From

"Come to us first": Centering Community Organizations in Artificial Intelligence for Social Good Partnerships

by Hongjin Lin,... at arxiv.org 09-12-2024

https://arxiv.org/pdf/2409.06814.pdf

Deeper Inquiries

How can funders and technology teams better align their priorities and incentives with the social impact goals of community organizations in AI for Social Good partnerships?

To better align their priorities and incentives with the social impact goals of community organizations in AI for Social Good (AI4SG) partnerships, funders and technology teams should adopt a collaborative and inclusive approach that emphasizes the needs and aspirations of community organizations. Here are several strategies to achieve this alignment:

Engage in Early and Continuous Dialogue: Funders and technology teams should initiate conversations with community organizations from the outset of project development. This engagement should not be limited to the ideation phase but should continue throughout the project lifecycle. By fostering open communication, stakeholders can ensure that the goals of community organizations are prioritized and integrated into project objectives.

Flexible Funding Models: Traditional funding models often impose rigid timelines and deliverables that may not align with the complex nature of social issues. Funders should consider providing flexible funding that allows community organizations to adapt their projects based on evolving needs and contexts. This flexibility can help ensure that the projects remain relevant and impactful.

Co-Design and Co-Implementation: Technology teams should adopt a co-design approach that actively involves community organization members in the development of AI technologies. By recognizing community organizations as equal partners, technology teams can leverage local knowledge and expertise, leading to more effective and contextually appropriate solutions.

Shared Success Metrics: Establishing success metrics that reflect the goals of both funders and community organizations is crucial. Funders should work with community organizations to define what success looks like in terms of social impact, rather than solely focusing on technical outputs. This shared understanding can help ensure that all stakeholders are working towards common objectives.

Capacity Building: Funders should invest in capacity-building initiatives that empower community organizations to engage effectively in AI4SG projects. This includes providing training and resources that enhance the technical skills of community organization members, enabling them to take on leadership roles in the development and implementation of AI technologies.

By implementing these strategies, funders and technology teams can create a more equitable partnership framework that aligns their priorities with the social impact goals of community organizations, ultimately leading to more successful AI4SG initiatives.

What strategies can community organizations employ to assert their leadership and co-ownership in the development of AI technologies intended for social good?

Community organizations can employ several strategies to assert their leadership and co-ownership in the development of AI technologies intended for social good:

Establish Clear Objectives: Community organizations should articulate their specific goals and objectives for AI4SG projects. By clearly defining what they hope to achieve, they can communicate their priorities to technology teams and funders, ensuring that their needs are recognized and addressed.

Leverage Local Knowledge and Expertise: Community organizations possess valuable insights into the social issues they address. By emphasizing their local knowledge and expertise, they can position themselves as essential partners in the development process. This can involve sharing data, providing context, and guiding technology teams in understanding the nuances of the communities they serve.

Advocate for Co-Leadership Roles: Community organizations should advocate for co-leadership roles in AI4SG projects. This can be achieved by negotiating terms that allow them to participate in decision-making processes, project design, and implementation. By asserting their role as co-leaders, they can influence project outcomes and ensure that their goals are prioritized.

Build Alliances and Networks: Forming alliances with other community organizations, advocacy groups, and stakeholders can amplify their voice and strengthen their position in AI4SG partnerships. By collaborating with like-minded organizations, they can share resources, knowledge, and strategies for asserting leadership.

Engage in Capacity Building: Community organizations should invest in building their own technical capacity and understanding of AI technologies. This can involve training staff members in data science, machine learning, and project management. By enhancing their technical skills, they can engage more effectively with technology teams and contribute meaningfully to the development process.

Document and Share Success Stories: By documenting and sharing their successes and lessons learned from AI4SG projects, community organizations can demonstrate their impact and expertise. This visibility can help establish their credibility and attract further partnerships and funding opportunities.

By employing these strategies, community organizations can assert their leadership and co-ownership in AI4SG projects, ensuring that their voices are heard and their goals are met.

How might the principles of data co-liberation be applied in other technology-for-social-good domains beyond AI, such as digital platforms, data infrastructures, or algorithmic decision-making systems?

The principles of data co-liberation can be effectively applied in various technology-for-social-good domains beyond AI, including digital platforms, data infrastructures, and algorithmic decision-making systems. Here are several ways to implement these principles:

Inclusive Data Governance: In digital platforms and data infrastructures, stakeholders should establish inclusive governance frameworks that involve community members, users, and marginalized groups in decision-making processes. This can ensure that data practices reflect the needs and values of diverse communities, promoting equitable access and use of data.

Participatory Design Processes: Applying co-liberation principles in the design of digital platforms involves engaging users and community members in the design process. By incorporating their feedback and insights, developers can create platforms that are more user-friendly, accessible, and aligned with the needs of the communities they serve.

Transparent Data Practices: Organizations should prioritize transparency in their data practices, including how data is collected, stored, and used. By openly sharing information about data practices, organizations can build trust with users and community members, empowering them to understand and influence how their data is utilized.

Empowerment through Data Literacy: Promoting data literacy among community members can empower them to engage with data infrastructures and algorithmic systems effectively. Training programs that enhance understanding of data concepts, privacy issues, and data rights can enable individuals to advocate for their interests and participate actively in data-related discussions.

Feedback Mechanisms: Establishing robust feedback mechanisms allows users and community members to voice their concerns and suggestions regarding digital platforms and algorithmic systems. By actively soliciting and incorporating feedback, organizations can continuously improve their systems and ensure they serve the intended social good.

Ethical Algorithm Development: In the context of algorithmic decision-making systems, applying co-liberation principles involves ensuring that algorithms are developed with input from affected communities. This can help mitigate biases and ensure that algorithms are designed to promote fairness, accountability, and transparency.

By applying the principles of data co-liberation across these domains, organizations can foster more equitable, inclusive, and ethical technology practices that prioritize the needs and rights of communities, ultimately enhancing the social impact of their initiatives.