The Asset Aging Services Keeps Discarding

Aging services has spent years treating age as a liability. Now that assumption risks being embedded in the artificial intelligence systems entering healthcare, long-term care, and disability services.

The familiar story is that intelligence peaks early and then steadily declines: speed matters, technology belongs to the young, and older professionals are useful mainly as institutional memory. There is a sliver of truth underneath the stereotype. Processing speed can slow with age. Working memory can narrow. Learning an unfamiliar platform may take longer.

But the story leaves out what develops across a lifetime: the ability to integrate knowledge from many domains, recognize patterns across decades, weigh competing perspectives, and resist a premature conclusion. Researchers describe aspects of this capacity as post-formal reasoning or wisdom. Erik Erikson associated later life with generativity—an orientation toward continuity and contribution beyond immediate gain. More recent research suggests that older adults can perform especially well when a problem requires multiple perspectives and attention to long-term consequences.

In practice, this becomes late-life synthesis: cross-domain knowledge, ethical memory, and pattern recognition shaped by sustained exposure to consequences. Aging-services organizations rely on it every day, even when they fail to name or value it.

After four decades in healthcare administration, surveying, and long-term care advocacy, I do not read a nursing facility’s financial filings as a new analyst might. I recognize the architecture underneath them: management-fee structures, related-party transactions, and the familiar choreography of reform rhetoric that follows public exposure. That recognition is not cleverness. It is temporal depth, earned by watching the same failure return under different names often enough to know it on sight.

AI Can Extend Judgment—But It Cannot Supply It

Artificial intelligence has meaningfully increased what I can produce. It helps organize research, manage drafts, retrieve information, and translate nonlinear thinking into readable prose. It does not think for me. The expertise was already there; the technology gave it infrastructure.

That distinction matters. AI is effective at retrieval, summarization, organization, and pattern surfacing at scale. It cannot decide which patterns matter, distinguish structural reform from theater, or determine which tradeoffs are morally acceptable when a facility is short-staffed on a Tuesday night. Those judgments grow from lived exposure to consequences. No volume of training data is equivalent to having been present when those consequences landed.

For aging-services leaders, the operative question is not whether experienced staff can adapt to new technology. It is whether the technology is configured to draw on what they already know—or to route around them.

Historical Bias Can Become Automated Practice

AI systems learn from data, and data reflects human choices: whose experiences were recorded, which patterns were treated as normal, and whose needs were measured precisely. Older adults and people with disabilities have too often been excluded from those choices.

A University of Toronto-led analysis of 146 AI ethics guidelines found that only 23 percent mentioned age-related bias, even as AI was rapidly moving into nursing homes, hospitals, and disability services. The omission is not abstract. Healthcare algorithms trained on historical data can learn discriminatory practices as though they were clinical facts.

Older patients have historically been undertreated for pain. Disability status has been used to justify rationing care. Nursing home residents have sometimes received less aggressive treatment than patients in acute settings. An algorithm does not know that a consistent historical pattern may also be unjust. Without deliberate safeguards, it can repeat that pattern at scale, invisibly, and with the authority of data.

Obermeyer and colleagues documented this dynamic in a widely used healthcare algorithm. The system produced significant racial bias because it used cost as a proxy for medical need. The lesson extends beyond that specific case: whenever data reflects unequal access or treatment, a seemingly neutral proxy can encode the inequity.

The disability-rights movement offers the governing principle aging services needs: Nothing about us without us. Systems affecting a population should be built with that population at the level of values and authority—not designed first and tested on people afterward. Too much AI development still violates that principle.

Four Questions Before the Next AI Purchase

Aging-services organizations are making AI purchasing decisions now. Before adopting clinical decision support, staffing algorithms, automated eligibility tools, or similar systems, leadership should require written answers to four questions:

1. Has the system been tested for different performance across age and disability status, and can the vendor produce the results?
2. What does the training data represent—and who is missing from it?
3. What happens when an algorithmic recommendation conflicts with professional judgment? Does the system defer, or must the practitioner log an override as an exception?
4. Who governs the deployment? Do older adults and people with disabilities hold actual decision-making authority, including a vote, rather than an advisory seat alone?

These are not compliance checkboxes. They determine whether a system strengthens the judgment an organization already possesses or quietly overrides it.

The Disability Law Center’s two-year investigation of Bear Mountain at Worcester, Massachusetts  documented understaffing, overmedication, isolation, and neglect. KFF Health News separately found that more than half of Massachusetts nursing homes were operating below state staffing minimums. These are not novel failure modes. Experienced advocates had been tracking the underlying patterns long before investigations confirmed them in writing. That is exactly the problem. A system trained to reproduce historical norms may accept those conditions as ordinary. A person with decades of experience may recognize them as warning signs.

The Expertise Hiding in Plain Sight

Every aging-services organization has people who can read a situation before the metrics catch up: the charge nurse who knows a resident is declining before an assessment flags it; the administrator who can distinguish a genuine quality initiative from a public-relations exercise; the resident who knows which workflow fails after the day shift leaves; the advocate who recognizes a familiar promise attached to an unchanged structure.

This capacity is not sentiment or institutional nostalgia. It is knowledge, earned through repeated contact with consequences. Yet it is precisely the input that AI governance in aging services too often excludes.

The answer is not a diversity checkbox on a design team. It is to treat the experience of people who have navigated ageist and ableist systems—from inside those systems as staff members, residents, family members, and advocates—as foundational expertise. They need a place at the procurement and governance tables before decisions are made, not an invitation to share anecdotes after a system has already been selected.

Organizations that get this right will not simply deploy AI more ethically. They will deploy it more accurately. The people correcting an algorithm’s blind spots will be the same people who have spent decades learning to see around them.

That is a moral argument, but it is also an operational one. Aging-services leaders should act on it before the next procurement cycle—not after the next investigation.

James A. Lomastro, PhD, is a healthcare policy analyst, national CARF surveyor, and pro bono advocate with Dignity Alliance Massachusetts. He writes on nursing home reform, AI governance, and elder justice.

Photo credit: Shutterstock/Drazen Zigic