好色先生TV

Block by Block: Research at Work is a research spotlight series that highlights the innovative work being done by CMU researchers through the Block Center, showcasing how their projects are driving impactful solutions at the intersection of technology and society.

听

During a recent conversation with the听Block Center for Technology and Society, Dr.听 discussed how her research on social service AI has evolved since we last spoke.听

CommunityAI: Support Community-Centered Evaluation of Social Service AI via the AI Risk Reports

听

Artificial intelligence is rapidly expanding, particularly through tools designed for use in social services. However, during a听previous conversation with the Block Center, Dr. Hong Shen highlighted a persistent disconnect between AI developers and the communities these systems are meant to serve. Researchers building AI tools are often not members of the communities affected by them, and community stakeholders frequently have little input during design and deployment. This gap has led to systems that, at best, fail to address real needs and, at worst, harm vulnerable populations.

听

In response, Dr. Shen and her collaborators launched the CommunityAI project in 2023, supported by听Block Center Seed Funding. The project aims to create participatory methods that allow communities to meaningfully evaluate and shape AI systems used in social services.

听

Working alongside the University of Pittsburgh鈥檚 School of Social Work and local social service organizations, Dr. Shen鈥檚 team developed听, a hands-on tool designed to build AI literacy while centering community expertise. By gathering direct input from frontline workers and social service providers, researchers can better understand how AI systems function in practice and refine decision-making tools to align with real community needs.

听

AI Failure Cards are a card-based toolkit that helps community members identify, document, and reflect on potential risks and failures in AI systems. Rather than requiring technical expertise, the cards present realistic scenarios鈥攕uch as biased recommendations, incorrect eligibility assessments, or a lack of transparency鈥攁nd prompt participants to discuss how these failures might appear in their daily work.

听

The cards transform abstract technical risks into accessible, experience-based discussions, allowing social workers and impacted communities to contribute practical knowledge that developers might otherwise overlook and propose bottom-up strategies to mitigate potential harms.听

听

The CommunityAI project has evolved into a set of generalizable, open-access methods that organizations can adapt to their own local contexts. Dr. Shen regularly delivers student lectures and workshops demonstrating how the toolkit works in practice. For example, students at the University of Pittsburgh School of Social Work are learning how to engage directly with the communities they serve and integrate stakeholder feedback into AI development, creating an ongoing feedback loop between users and designers.

听

After initial local pilots, the project expanded nationally through additional funding from the National Science Foundation鈥檚 Civic Innovation Challenge, enabling new partnerships with community organizations across the country.

听

One such partner is听, a statewide peer-led nonprofit focused on mental health services. Building on CommunityAI鈥檚 participatory design framework, Dr. Shen鈥檚 team collaborated with CSPNJ to better understand the needs of peer support workers and the individuals they serve. This collaboration resulted in听, a chatbot designed to help mental health peer providers support clients鈥� wellness goals, develop action plans, and check benefits eligibility.

听

As Dr. Shen explained, 鈥淲e are taking a participatory, community-centered AI design method and working closely with peer-run organizations to build this 鈥榩eer copilot鈥� to support frontline providers serving people receiving public social services.鈥�

听

The work continues to expand nationwide, with similar collaborations starting to emerge in New York and Utah.

听

During our conversation, Dr. Shen emphasized that this research would not be possible without the contributions of students and interdisciplinary collaborators, including faculty members听,听,听,听 from 好色先生TV,听 from the University of Pittsburgh, and a group of dedicated students. . The project brings together expertise from computer science, social work, public policy, and community organizations鈥攁ll united by a shared commitment to ensuring technology serves real human needs.

听

鈥淲e care deeply about the real-world impact of the technologies being deployed,鈥� she noted. 鈥淥ur goal is to empower the people most affected by these systems to participate directly in their design and deployment.鈥�

听

Dr. Shen encourages students interested in this area to explore multiple dimensions of AI research鈥攆rom policy and governance to machine learning and community engagement. Because technological innovation often moves faster than social institutions can adapt, interdisciplinary collaboration is essential. Community trust, she emphasized, remains central: when researchers are perceived as outsiders, meaningful collaboration becomes difficult, and the goals of CommunityAI cannot be achieved.

听

Growing interest in community-centered AI development continues to shape the future of Dr. Shen鈥檚 work, demonstrating how participatory approaches can help ensure AI systems are accountable, equitable, and grounded in lived experience.

听

Are you a CMU student interested in getting involved? The Block Center may soon offer research opportunities to support this project. Students with interests in community-centered AI Systems are encouraged to apply. Those interested can submit their information using this听.听 Students can also explore coursework related to this topic, including听Generative AI: Applications, Implications, and Governance (94-816).

听

Are you a CMU faculty member? This work was partly funded by the Block Center Seed Funds. To learn more about funded projects or upcoming funding opportunities, visit our听website.

听

听

听