Information, Classification, and Epistemic Adequacy in Medicine

In the modern world of medicine, the word "information" is everywhere. It is the data in a patient's chart, the results of a blood test, the notes a doctor writes, and the statistics used to shape public health policy. We treat this information as a simple accumulation of facts, something that grows larger and more detailed as technology improves. However, the concept of information has a deeper history. In the mid-twentieth century, scientists developed rigorous ways to measure information, treating it not just as knowledge, but as a specific property of how we describe the world. They realized that to communicate or store anything, we must compress a vast, complex reality into a shorter, finite set of words or numbers. This compression is necessary; we cannot carry the entire universe in our pockets. The question that arises is whether this necessary shortening ever strips away something vital. If we squeeze a complex human experience into a simple medical label, do we lose the very details that determine how to help that person?

This is the central puzzle explored by Vladimir Zaichenko in a new research article. The paper investigates what happens when the open-ended, messy reality of a patient's life is forced into the rigid boxes of medical classification systems. Zaichenko argues that while we cannot avoid simplifying a patient's condition to make it manageable, there is a specific kind of failure that occurs when we simplify too much. He suggests that a medical description is only truly "adequate" if it keeps the specific distinctions that a doctor needs to make the right decision, even if it throws away everything else. The research focuses heavily on psychiatry, where the gap between a person's complex inner life and a diagnostic label is often widest, but the logic applies to all of medicine. The core finding is that the ethical duty of a doctor to "do no harm" includes an intellectual duty: to ensure that the process of turning a patient into a diagnosis does not discard the very clues needed to treat them safely.

To understand the stakes, one must look at how medical systems actually work. A human being is a complex system with an almost infinite number of possible states. A patient might feel a certain way, have a specific history, react to stress in a unique manner, and live in a particular social environment. Every detail could potentially matter. Yet, hospitals and clinics cannot operate on infinite detail. They rely on classification systems, such as the International Classification of Diseases, which offer a finite list of categories. When a doctor diagnoses a patient, they are performing a translation. They take the open-ended, continuous flow of the patient's reality and map it onto a limited set of available labels. This process is not a mistake; it is a structural necessity. Just as a map must leave out every single tree and stone to show the road, a medical diagnosis must leave out many details to show the condition.

The problem, Zaichenko explains, is that this reduction is not always neutral. In the rush to categorize, doctors and systems might discard distinctions that seem unimportant at the moment but become critical later. For instance, two patients might receive the exact same diagnosis because they fit the same criteria. However, the details that were left out of that label might be the very things that dictate whether one patient needs a specific medication while the other needs a different kind of therapy. If the classification system treats them as identical because it discarded those nuances, the resulting care could be ineffective or even dangerous. The paper suggests that the goal of a medical description should not be to capture every possible fact, nor should it be to make the description as short as possible. Instead, the goal should be to find the shortest description that still keeps all the distinctions necessary for the specific clinical task at hand.

This idea draws on a concept from computer science known as minimum description length, which asks how briefly one can describe an object without losing the ability to reconstruct it. Zaichenko adapts this for medicine, proposing that a diagnosis is "epistemically adequate" only if it preserves the distinctions required for the doctor to act correctly. If a description is too compressed, it becomes a source of error not because it is false, but because it is incomplete in the wrong way. The paper argues that we often mistake the label for the reality. Once a patient is assigned a category, that category can begin to substitute for the richer, more complex picture of the person. This is particularly dangerous when medical information moves from the clinic to other settings, such as legal courts or administrative offices. In those places, the people making decisions rely entirely on the compressed label, unaware of the vital details that were stripped away during the initial diagnosis.

Psychiatry serves as a sensitive test case for this argument because mental health involves a field of information that is incredibly difficult to pin down. A person's behavior, thoughts, feelings, and history blend together in ways that do not fit neatly into isolated variables. When a psychiatrist assigns a diagnosis, they are compressing a vast, heterogeneous field of experience into a single term. The paper notes that this is not inherently wrong; a finite vocabulary is the only way to communicate and treat. The danger arises when the system assumes that the label contains everything that matters. Zaichenko points out that two people with the same diagnosis might have vastly different needs for treatment, prognosis, or risk assessment. If the classification system discards the features that distinguish their needs, the resulting care plan may fail. The issue is not that the system is imperfect, but that we forget which parts of the reality were left behind.

The research also highlights how modern medicine's increasing specialization contributes to this problem. Different doctors and instruments focus on specific parts of a patient's condition, each optimizing their own piece of the puzzle. A specialist might produce a highly precise measurement for their specific area, but that precision does not guarantee that the information remains useful for the overall treatment plan. The output of one specialist becomes the input for another, and in that handoff, crucial distinctions can be lost. The paper suggests that this is a structural risk of specialization: the more precise a local view becomes, the more it can become detached from the broader context of the patient's life. The result is a collection of accurate but fragmented descriptions that, when combined, may still miss the point.

Ultimately, the paper connects this technical issue to a fundamental ethical principle: the obligation to do no harm. Traditionally, this rule is understood as a warning against physical injury or bad medical procedures. Zaichenko argues that it must also apply to the production of knowledge itself. A doctor can cause harm not by saying something false, but by producing a description that is formally correct yet omits a distinction that is relevant to the patient's future care. If a diagnosis leaves out a detail that changes the treatment, the harm is real, even if the diagnosis was technically accurate according to the rules. The ethical responsibility, therefore, extends to what is left unstated. Doctors and systems must be careful not to discard distinctions that could foreseeably lead to harm, even if those distinctions seem minor or irrelevant to the immediate task of classification.

The conclusion of the paper is that epistemic adequacy—the quality of being a good description for a specific purpose—is not an optional refinement for doctors. It is a necessary condition for safe practice. A classification is only as good as the function it serves. If the goal is to treat a patient, the description must keep the distinctions that make treatment possible. If the goal is to assess risk, it must keep the distinctions that make risk assessment possible. The paper does not propose a new way to measure information or a new mathematical formula for doctors to use. Instead, it offers a way of thinking about the trade-offs that happen every time a patient is diagnosed. It asks us to recognize that every time we turn a complex human life into a medical code, we are making a choice about what to keep and what to lose. The challenge is to ensure that we never lose what matters most.