AI Has a Darker Side in Healthcare, Expert Says | MedPage Today

TEXT START: LOS ANGELES -- Artificial intelligence (AI) has its benefits in the healthcare sphere, but there are also pitfalls to watch out for, one expert said here at the Association of Health Care Journalists Fall Summit.

The Dissection

The text inventories AI’s immediate labor effects: surveillance, algorithmic management, deskilling, speed-up, wage suppression, data extraction, loss of autonomy, and the conversion of healthcare into a platform-mediated gig market. Its real function is to frame AI as a labor-governance problem that can be contained through worker power, legislation, and professional safeguards.

It identifies the operating machinery of extraction accurately. It does not follow that machinery to its terminal conclusion.

The Core Fallacy

The text treats worker degradation as the central danger while leaving the employment system itself intact. Under the Discontinuity Thesis, the larger break is not merely that remaining jobs become more coercive or deskilled. AI severs the mass employment-to-wage-to-consumption circuit and removes the majority’s economic necessity.

Its proposed remedies—bans, liability rules, worker protections, preservation of scope of practice—are lag defenses. They may slow deployment or preserve pockets of human labor, but they cannot defeat competitive pressure to automate cognitive work wherever AI is cheaper, faster, more scalable, or less legally exposed.

The article’s own evidence exposes this. Residents unable to interpret EKGs without AI demonstrate genuine deskilling and system fragility. But the competitive response will not automatically be restoration of human mastery. It will be more automation, narrower human roles, and institutional tolerance for a smaller number of highly supervised operators.

Hidden Assumptions

• Healthcare will continue to provide broad, economically necessary employment rather than shrinking into a smaller owner-controlled and maintenance-heavy system.
• Human-only clinical domains can be preserved at scale through law or professional norms.
• Worker power can be rebuilt quickly enough to counter employers, vendors, and states competing for lower costs.
• AI deployment is discretionary rather than a competitive requirement.
• Retraining can restore skills after automated systems have removed the practice needed to maintain them.
• Assigning liability will meaningfully restrain deployment instead of merely distributing the cost of failure among hospitals, vendors, insurers, and clinicians.
• Patient trust and professional autonomy will outrank throughput, documentation, margin, and labor substitution.
• The principal harm is bad treatment of employed workers, not the disappearance of productive participation for the broader population.

Social Function

This is a partial truth functioning as transition management and ideological anesthetic. It gives workers and institutions an accurate list of near-term injuries—surveillance, deskilling, speed-up, and gigification—while making the crisis appear governable within the existing labor order.

The closing claim that technology is “not good or bad, but neither is it neutral” is directionally correct but strategically incomplete. AI is not neutral because it enters institutions through ownership, bargaining power, and competitive cost logic. Those forces do not merely produce bad implementation. They determine who remains necessary.

The Verdict

The article is a competent early-warning memo about the first layer of the autopsy. It correctly shows AI turning healthcare workers into monitored, measurable, replaceable inputs. It fails to confront the second layer: once AI can perform enough cognitive clinical work, the system has no stable mechanism for preserving mass human participation.

The laws and safeguards mentioned are hospice care for the old employment model. They may preserve islands of human practice and create temporary servitor niches, but they do not reverse the structural death of post-WWII capitalism.