Other Mental Models · OM-38
Improvement often comes more reliably from removing what is harmful, unnecessary, or fragile than from adding something new — subtraction is frequently a more robust and less risky path to a better outcome than addition.
A principle, with roots in theology and philosophy, holding that understanding or improving something is often better achieved through subtraction — identifying and removing what is false, harmful, or unnecessary — than through addition, since what remains after genuine removal tends to be more robust, and removal carries fewer unintended consequences than adding something new whose full effects aren't yet known.
The term originates in negative theology (describing God by what God is not, rather than by positive attributes), and was adapted as a general decision-making and risk-management principle by Nassim Nicholas Taleb, particularly in Antifragile (2012), where he argued subtraction is generally a more robust improvement strategy than addition under genuine uncertainty.
The Mechanism
What you remove is usually more certain in its effect than what you add
Taleb frequently illustrates via negativa with health: removing known harmful habits (smoking, chronic sleep deprivation, excessive sugar) tends to have more reliably positive and better-understood effects on long-term health than adding a new supplement or intervention whose full long-term effects are far less certain — subtraction of a known harm carries a more predictable benefit than addition of an unproven, unfamiliar new element to an already-complex system.
01 · REMOVAL GENERALLY CARRIES A MORE PREDICTABLE, BETTER-UNDERSTOOD RISK PROFILE THAN ADDITION
This is the core reasoning behind preferring subtraction under genuine uncertainty
When a system's components and their interactions are not fully understood — true of most complex systems, from the human body to organizations to financial markets — adding something new introduces an additional variable whose full downstream interactions are genuinely difficult to predict in advance, whereas removing an already-identified harmful or unnecessary element tends to simply return the system toward a state closer to its own prior, often already reasonably robust, baseline.
02 · IT COUNSELS SKEPTICISM TOWARD THE DEFAULT BIAS TOWARD ADDITION IN MOST HUMAN SYSTEMS
Organizations, bureaucracies, and individuals tend to default to adding, rather than removing, when trying to improve something
Empirically, both individuals and organizations show a documented default bias toward addressing problems by adding something new — a new rule, a new feature, a new process step — rather than by considering whether removing an existing harmful or unnecessary element might achieve the same or better result at lower risk; via negativa specifically counsels checking the removal option first, given its typically more favorable risk profile.
03 · IT DOESN'T CLAIM ADDITION IS NEVER WORTHWHILE — ONLY THAT SUBTRACTION DESERVES SERIOUS CONSIDERATION FIRST
The principle is a prioritization heuristic under uncertainty, not an absolute prohibition on adding anything new
Via negativa doesn't argue that adding new elements to a system is never beneficial — many genuine improvements do require addition — it specifically argues that under conditions of real uncertainty about a complex system's true behavior, the removal option should be seriously considered and often attempted first, given its generally more predictable and less risky effect, before committing to an addition whose full consequences are harder to fully anticipate.
Where It Fails / Inversion
Where it fails / inversion
Not every problem can be solved by removal — some genuine improvements require a positive, additive intervention that no amount of subtraction alone could achieve (building an entirely new capability rather than removing an existing deficiency), and treating via negativa as a universal preference for subtraction over addition, rather than as one heuristic to weigh alongside others, can lead to under-investing in genuinely valuable new additions.
How To Use It
Worked example · improving an organization's process without adding more bureaucracy
A team trying to improve a chronically slow approval process should first consider which existing steps, approvals, or requirements could simply be removed as unnecessary or redundant, rather than defaulting immediately to adding a new tracking tool or an additional oversight step — removing an identified unnecessary step tends to have a more predictable, immediately beneficial effect on speed than adding a new process element whose own downstream effects and adoption challenges are harder to fully anticipate.
How to use it
When trying to improve a system, consider what could be removed — a harmful habit, an unnecessary rule, a fragile dependency — before defaulting to what could be added; removal of a genuinely unnecessary or harmful element tends to have a more predictable, less risky effect than adding something new to an already complex system whose full interactions aren't yet well understood.
See Also