Artificial intelligence could improve social care by connecting fragmented systems and flagging risks, but it will not replace the human elements of trust and compassion.
Connecting the Dots
Social care today relies on manual steps that slow down service delivery. People often fill out the same forms multiple times, and various agencies see only parts of a person’s situation. AI tools might analyze clinical, claims, and social needs data to identify risks before they reach a hospital emergency room. These systems could also support better matching between community resources and individual needs.
Organizations using this technology would ideally spend less time chasing information and more time helping people. Closed-loop referrals and improved follow-up are possible goals. AI might summarize notes and suggest next steps to reduce the administrative burden on case managers.
However, organizations must avoid replacing existing systems with new ones. Social care networks are complex, and adding disconnected tools can create more friction than it solves. AI should help make the ecosystem easier to handle, not add another disconnected tool into the mix.
Risks and Reality
Social care data is often messy, incomplete, and collected inconsistently. When algorithms are trained on historical data, they can repeat existing inequities. Bias in health AI is a documented concern, and poorly designed tools could worsen disparities instead of reducing them. If an algorithm acts as a gatekeeper for services, it might deny assistance based on a model’s prediction of low probability, leaving people without food, housing, or transportation when they need it most.
Transparency and governance are essential. ONC’s HTI-1 final rule established transparency requirements for AI and predictive algorithms in certified health IT. HHS section 1557 rules prohibit discrimination through patient decision support tools. A higher level of scrutiny is necessary; AI should not quietly make decisions that people cannot see, question, or correct, and it cannot replace human judgment.
For the people receiving these services, the margin for error is dangerously small. A model might suggest transportation support when the real barrier is fear, language, or domestic violence, or it might screen a person as low-risk today while they need help next month. AI can help with context, but it cannot fully understand the nuance of a deeply personal situation.
