First Do No Harm
metaphor established
Source: Medicine → Ethics and Morality, Decision-Making
Categories: health-and-medicineethics-and-morality
From: Schein's Surgical Aphorisms
Transfers
“First do no harm” — often rendered in Latin as primum non nocere — is the most widely invoked principle in medical ethics and one of the most thoroughly migrated medical concepts. It encodes a specific priority ordering: before considering whether your intervention will help, verify that it will not make things worse. The constraint is asymmetric: it treats the prevention of iatrogenic harm (harm caused by the healer) as lexicographically prior to the pursuit of therapeutic benefit.
Key structural parallels:
- The intervention-harm calculus — in medicine, every treatment carries risks. Surgery can cause infection. Antibiotics can cause allergic reactions. Chemotherapy destroys healthy cells alongside cancerous ones. The physician’s first obligation is to weigh these risks against the disease’s natural course and ensure the treatment is not worse than the condition. This imports directly into technology ethics (will this feature cause more harm than it prevents?), policy design (will this regulation’s side effects exceed the problem it addresses?), and software deployment (will this hotfix introduce regressions worse than the bug it patches?). The structural transfer is the discipline of comparing intervention outcomes to baseline outcomes, not to ideal outcomes.
- Endogenous healing versus exogenous intervention — a key medical insight underlying the principle is that the body heals itself in many cases. Unnecessary surgery, over-prescription, and premature intervention can interfere with natural recovery. This maps onto organizational life, where many problems resolve themselves through natural turnover, shifting priorities, or gradual adjustment. The manager who intervenes in every team conflict, the regulator who legislates every market imperfection, and the developer who refactors every code smell may all be violating the principle by disrupting endogenous repair processes.
- Commission-omission asymmetry — the principle encodes a moral asymmetry: actively causing harm through intervention is worse than passively allowing harm through inaction. A surgeon who operates unnecessarily and causes a complication bears responsibility in a way that a surgeon who recommends watchful waiting does not, even if the patient deteriorates. This transfers into legal liability (malpractice attaches to action more easily than to inaction), technology ethics (companies that deploy a harmful AI model are judged more harshly than companies that fail to deploy a beneficial one), and policy (regulators who approve a dangerous drug bear more blame than regulators who delay a beneficial one).
- Epistemic humility as a design constraint — the principle is strongest when uncertainty is high. When the physician is confident in the diagnosis and the treatment, the injunction is easily satisfied. When the diagnosis is uncertain and the treatment experimental, “first do no harm” becomes a powerful brake on overconfident intervention. This transfers to any domain where actors must decide under uncertainty: it argues for conservative action when you do not fully understand the system you are intervening in.
Limits
- “Harm” is not self-defining — in medicine, harm is relatively (though not perfectly) clear: death, disability, pain, infection. In policy, technology, and ethics, harm is deeply contested. Does restricting speech cause harm? Does not restricting speech cause harm? Does economic disruption from regulation count as harm? The principle offers no guidance on how to define harm, and its apparent clarity conceals the fact that in most domains of application, the threshold question — what counts as harm? — is the real disagreement.
- It privileges the status quo — “do no harm” takes the current state as the baseline. But the current state may itself be harmful. A physician treating a patient with a chronic condition cannot return to a pre-disease baseline; they can only choose among imperfect futures. In policy, “first do no harm” can be invoked to block regulation of an already-harmful industry, because any regulatory intervention introduces new costs and risks. The principle, taken literally, can become a conservative bias against necessary disruption.
- It imports a one-to-one clinical frame into systemic contexts — the physician has one patient. The principle asks: will this treatment harm this patient? But policymakers, platform operators, and organizational leaders affect thousands or millions of people simultaneously. Vaccination programs harm some individuals (adverse reactions) while benefiting populations. Content moderation harms some speakers while protecting others. The principle’s one-patient structure provides no framework for these distributional trade-offs.
- The omission bias it encodes can be its own harm — the principle’s asymmetric treatment of action and inaction means that the harms of doing nothing are systematically underweighted. The regulator who delays approving a life-saving drug to avoid the risk of side effects causes invisible deaths — people who would have been saved die without anyone noticing. The technology company that declines to deploy a beneficial tool because of potential misuse costs are real but never counted. “First do no harm” makes inaction feel safe, but inaction has its own body count.
Expressions
- “First do no harm” — the standard English formulation, used across medicine, technology ethics, policy, and regulation
- “Primum non nocere” — the Latin rendering, sometimes attributed to Hippocrates but actually of uncertain medieval origin
- “Move fast and break things” — Facebook’s early motto, explicitly constructed as a rejection of the principle in favor of speed
- “The precautionary principle” — the regulatory analog in environmental and public health policy, encoding a similar intervention-skepticism
- “Defensive medicine” — the practice of ordering unnecessary tests to avoid liability, an over-application of the principle that generates its own harms
- “Above all, do no harm” — variant phrasing common in technology ethics discussions and AI safety discourse
- “Hippocratic oath for engineers” — periodic calls to formalize the principle for software and technology practitioners
Origin Story
The phrase “first do no harm” is routinely attributed to the Hippocratic Oath, but it does not appear there. The Oath says “I will abstain from all intentional wrong-doing and harm” and “I will use treatment to help the sick according to my ability and judgment, but never with a view to injury and wrong-doing.” The closest passage in the Hippocratic corpus is from Epidemics (Book I, Section XI): “As to diseases, make a habit of two things — to help, or at least to do no harm.” The Latin primum non nocere appears in medical writing by the 17th century but its exact origin is uncertain — it is sometimes attributed to Thomas Sydenham, sometimes to the Hippocratic tradition generally.
The phrase achieved its modern prominence through its adoption as a shorthand for medical conservatism: the physician’s first obligation is to not make things worse. Its migration into technology ethics accelerated in the 2010s as Silicon Valley’s “move fast and break things” ethos generated public backlash, and “first do no harm” became the rallying principle for a more cautious approach to technology deployment, AI development, and platform governance.
References
- Hippocrates. Epidemics, Book I, Section XI — the closest original source for the sentiment
- Smith, C.M. “Origin and Uses of Primum Non Nocere — Above All, Do No Harm!” Journal of Clinical Pharmacology 45(4), 2005 — traces the phrase’s uncertain provenance
- Beauchamp, T. and Childress, J. Principles of Biomedical Ethics (1979, 8th ed. 2019) — formalization of non-maleficence as one of four bioethical principles
- Jonas, H. The Imperative of Responsibility (1979) — philosophical argument for precautionary ethics in technology
Related Entries
Structural Neighbors
Entries from different domains that share structural shape. Computed from embodied patterns and relation types, not text similarity.
- Everyone Goes Home (fire-safety/mental-model)
- No One Profits from Their Own Wrong (governance/mental-model)
- System Resilience vs. Fragility (architecture-and-building/mental-model)
- Homeostasis (/mental-model)
- Psychological Flexibility (materials/metaphor)
- Resilience (resilience/mental-model)
- Culture as a Control System (physics/paradigm)
- Emotional Stability Is Maintaining Position (embodied-experience/metaphor)
Structural Tags
Patterns: balanceforceboundary
Relations: preventrestore
Structure: equilibrium Level: generic
Contributors: agent:metaphorex-miner