JULY 6 — In any hospital, intensive care unit beds are limited. This is not unusual, but a reality that medical teams must work within every day. 

When demand exceeds capacity, difficult decisions have to be made about which patients receive critical care.

These decisions are never taken lightly. They are guided by clinical protocols, medical judgement and, often, careful deliberation among the team. 

Even so, they can be challenging, especially when several patients present with serious conditions and the differences between them are not always clear.

In such situations, it is not difficult to imagine how artificial intelligence might be used to assist. 

An AI system could analyse large volumes of clinical data, compare current patients with thousands of past cases, and estimate probabilities of survival or recovery. 

It could then generate a ranked list, indicating which patients are most likely to benefit from intensive care.

Consider two patients. The first is a younger man with no underlying conditions. His illness is severe, but his overall prognosis is relatively strong if given intensive care. 

The second is an older woman with a more complex medical history. Her chances of recovery are lower, and her condition is less predictable.

Based on available data, the system recommends prioritising the first patient. The reasoning is clear. The decision appears justified. And yet, something in this moment remains unresolved.

The second patient has been under the care of the hospital for some time. The medical team knows her condition well. 

Her family is present, waiting outside, hoping for any sign of improvement. The recommendation does not ignore these realities, but it cannot fully account for them either.

The question is no longer only which patient has the higher probability of recovery. It is also how that decision is to be understood, explained and executed.

Rules are essential in guiding such decisions. They provide structure, reduce arbitrariness and help ensure a degree of fairness. 

In the context of artificial intelligence, governance frameworks play an important role in defining how systems should be designed and used. 

They establish boundaries and set expectations for accountability.

But rules operate within limits. They can tell us what is permitted. They can guide us toward outcomes that are efficient or statistically favourable. 

What they cannot do is fully account for the particularities of each situation, especially when those situations involve individuals whose lives cannot be reduced to variables alone.

In the intensive care unit, each patient represents more than a clinical profile. 

There are relationships, histories and hopes that do not appear in datasets, but nonetheless shape how decisions are experienced by those involved. 

These dimensions do not negate the importance of clinical judgement, but they remind us that decisions of this kind are not solely technical.

This is where judgement becomes necessary. Judgement does not reject rules. It works with them, but is not confined by them. 

It involves holding together different considerations, some of which may not be easily reconciled, and arriving at a decision that one is prepared to stand by.

Artificial intelligence can assist in this process. It can highlight patterns, reveal possibilities and support analysis in ways that go beyond human capability. But it does not carry the decision through the moment in which it must be made.

It does not bear the consequences. That responsibility remains with those who act. 

As systems become more capable and their recommendations more persuasive, there is a natural tendency to treat them as authoritative. Their conclusions, presented with clarity and confidence, can give the impression that the decision has already been made.

Yet receiving a recommendation is not the same as making a final decision. 

There remains a space, often quiet and easily overlooked, where responsibility must be taken. This is the space where judgement operates, where one considers not only what can be done, but what ought to be done in that particular moment.

Returning to the two patients, the system’s recommendation does not remove the dilemma of the decision. It refines it.

The data may guide the discussion, but it does not conclude it. The final decision still rests with those who must explain it, justify it and live with its consequences, not only in clinical terms, but in human ones.

And it is in that moment, standing between what the rules suggest and what the situation demands, that we begin to understand why artificial intelligence needs more than governance.

* Ng Kwan Hoong is an Emeritus Professor of Biomedical Imaging at the Faculty of Medicine, Universiti Malaya. A 2020 Merdeka Award recipient, he is a medical physicist by training but also enjoys writing, drawing, listening to classical music, and bridging the gap between older and younger generations. He may be reached at ngkh@ummc.edu.my

** This is the personal opinion of the writer or publication and does not necessarily represent the views of Malay Mail.