What happens when we introduce AI into a system that has not yet learned how to collaborate?
AI Doesn’t Eliminate Fragmentation
AI can analyze information. It cannot make agencies trust one another, resolve competing priorities or decide when sensitive information should be shared. Those are governance decisions.
In homelessness services, AI might identify patterns associated with housing mobility or suggest possible interventions. But what happens when housing, health and nonprofit partners maintain different records? What if information is missing? What if agencies define success differently?
AI may connect the data without connecting the organizations. Worse, fragmented collaboration can become automated fragmentation.
That is why AI governance cannot be the exclusive responsibility of IT departments. Program administrators, frontline employees, nonprofits, legal and privacy professionals, data specialists and communities should have a voice. AI governance should itself be collaborative governance.
Who Get to Define What AI Sees?
My recent study of artificial intelligence raised a question that public administrators should consider: Who gets to define what counts as evidence?
AI does not experience communities. It analyzes the information institutions collect about them. A large agency may contribute extensive data, while a small community organization may possess critical knowledge that never enters the system. People with fewer interactions with government may also be less visible administrative data.
An AI system could therefore produce a sophisticated recommendation from an incomplete picture.
Before asking what AI can predict, we should ask: Whose information is represented, whose knowledge is missing and who defined the outcomes?
AI Should Help Us Think
AI can identify patterns, reveal connections and help us ask better questions. But identifying a pattern is not the same as understanding what it means.
Government must preserve meaningful human oversight. A caseworker should be able to question an AI recommendation. A person affected by an important decision should be able to challenge it. Public officials must remain accountable for the outcome.
This approach aligns with the National Institute of Standards and Technology’s AI Risk Management Framework and the U.S. Government Accountability Office’s AI Accountability Framework. NIST emphasizes governance and clearly defines human roles in AI oversight, while GAO organizes its framework around governance, data, performance and monitoring.
Govern Collaboration First
The promise of AI in government is not replacing human judgment. It is helping humans see connections that fragmented institutions make difficult to see. But technology cannot manufacture trust.
Before agencies automate complex human-service decisions, they should build shared rules, transparent data practices, collaborative relationships, and clear accountability.
The principle is simple: Do not automate fragmentation. Govern collaboration first.
Otherwise, AI may help government move faster, while simply building a faster maze.
Dr. Denise D. Hendrix, DPA, is a director at New York City’s Human Resources Administration (HRA), where she advances collaborative governance and public service innovation. With more than 15 years of experience in public and nonprofit administration, her work focuses on homelessness, child welfare, organizational leadership and cross-sector collaboration. Dr. Hendrix is an adjunct professor, public administration scholar and contributor to PA TIMES. Her research explores trust, communication and collaborative governance networks. She is currently pursuing professional coaching certification and is passionate about developing leaders who build strong organizations through trust, effective communication and meaningful collaboration.
Leave a Reply
You must be logged in to post a comment.