In the sixth episode of the CBS legal drama Cupertino, titled “Harper,” the series turns its lens to the high-stakes intersection of artificial intelligence (AI) and life-or-death medical allocation.
In the episode, Harper Gervin, 27, the roommate of Price, Siffre & Associates manager Christy Palladino, loses her long-awaited liver transplant to Albert Ensler, a wealthy venture capitalist, just before surgery. With the organ viable for four hours and Michael Price, Olivia Siffre, and Judge Hornstein stranded in U.S. 101 traffic, they litigate by video. Assuming AI bias disadvantaged Harper, the attorneys learn that the allocation algorithm selected her, but a hospital physician-executive overrode it after Ensler’s foundation offered a cancer-research donation. Christy exposes the quid pro quo, restores Harper’s claim, and reveals that human corruption poses a greater danger than AI bias.
This fictional courtroom scramble mirrors a real, rapidly evolving debate in modern transplant medicine over how AI might help or hurt patient ranking. To my chagrin, the television producers decided not to credit my novel Transplant, which examined the nefarious side of transplant allocation. Self-promotion aside, in real-world transplant settings, AI has shown more promise than harm as a decision-support tool, though many proposed applications still require prospective clinical validation.
Transplant waitlists have long relied on mathematical formulas, such as the Model for End-Stage Liver Disease (MELD) score, adopted for U.S. liver allocation in 2002, to estimate medical urgency and prioritize candidates. MELD is a statistical risk score, not an autonomous AI; its current adult version, MELD 3.0, was implemented in 2023 to address disadvantages faced by women.
Real physician voices, twice a week
Free, and one click to unsubscribe.
Determining eligibility for a transplant waitlist differs from ranking candidates for a particular donated organ. The U.S. transplant system is transitioning toward continuous distribution models that replace rigid allocation categories with weighted point systems. Lung allocation adopted this framework in 2023; continuous distribution is a policy framework, not inherently an AI system.
National allocation policies also consider factors such as blood type compatibility and distance from the donor hospital. Modern machine-learning models can analyze vast amounts of complex clinical data to uncover patterns that traditional equations or human committees might miss. Researchers are also developing deep-learning tools to help organ procurement organizations identify hard-to-place organs, such as kidneys, early, so they can be considered for accelerated placement, an approach that could speed placement and reduce organ discard rates.
Furthermore, an experimental multi-agent AI system, where separate large language model agents simulate the perspectives of a transplant surgeon, hepatologist, cardiologist, and social worker, achieved approximately 92 percent accuracy in predicting one-year post-transplant survival and 98 percent in identifying contraindications in a retrospective proof-of-concept study. The study combined real transplant-recipient data with artificially assigned contraindications; it did not demonstrate safe or equitable performance within a functioning transplant selection committee. Despite these impressive technical capabilities, handing the final ranking of transplant candidates to a fully autonomous AI without human oversight remains unrealistic because of ethical, social, and regulatory barriers.
One major obstacle is that algorithms do not always align with human moral values when making complex ethical trade-offs. A study examining how large language models weigh competing kidney transplant candidates found that AI systems often prioritized variables differently than human research participants, whose medical training was unspecified. Models sometimes fixated on a single factor (such as drinking habits) and rarely expressed indecision, even when given that option. These hypothetical comparisons reveal differences in value judgments; they do not establish that AI performed worse than physicians making real transplant decisions.
Equally concerning is the risk of amplifying historical bias. Because machine-learning models are trained on past medical records, and historical transplant data reflects longstanding racial, socioeconomic, and geographic disparities in who is referred and approved for organs, an unmonitored algorithm risks automating and entrenching systemic inequities. This challenge is compounded by the “black box” nature of complex neural networks, which often cannot provide transparent, easily understandable explanations for why one patient was ranked above another.
Organ allocation depends heavily on public trust, and patients, families, and clinicians require clear accountability for life-altering decisions. The urgency of end-stage organ disease and the scarcity of donated organs pose practical and ethical challenges to conducting traditional randomized clinical trials of allocation tools. These challenges do not eliminate the need for rigorous prospective evaluation, including carefully designed trials where feasible, before entrusting AI systems with decisions that affect access to transplantation.
Consequently, the more defensible near-term approach to AI in transplant medicine is neither purely human decision-making nor complete algorithmic control, but rather a collaborative “human-in-the-loop” model. Just as Cupertino illustrates that both algorithms and human administrators have distinct vulnerabilities, AI should serve as a clinical assistant rather than an independent arbiter, with transparent oversight of both its recommendations and human overrides.
In practice, this could mean integrating validated algorithms to flag high-risk candidates who might otherwise be overlooked by conventional metrics, estimate likely outcomes for specific donor-recipient combinations, and streamline administrative bottlenecks before an organ loses viability. By pairing the data-processing speed and consistency of machine learning with the ethical judgment, contextual awareness, and accountability of multidisciplinary transplant teams operating under national allocation policies, health care systems can aim to make organ allocation fairer and more efficient while keeping ultimate responsibility for human life firmly in human hands.
Arthur Lazarus is a physician-author whose work spans narrative medicine, physician leadership, artificial intelligence, health care ethics, medical culture, and fiction. He has published more than 500 articles and essays across scientific journals, professional publications, and online platforms.
He is the author of numerous books on narrative medicine, AI in medicine, career development, and the changing moral landscape of health care, as well as fictional series including Rounds Never End, Sick and Systemic, and Real Medicine, Unreal Stories. His writing explores the forces reshaping modern medicine while preserving a central commitment to story, meaning, judgment, and the human relationship at the heart of care.
He shares updates on LinkedIn.
Your KevinMD author page is one of the top results for your name. Turn it into a premium profile, publish in 1 to 3 days, and get a guaranteed podcast slot.

