Wisdom and philosophy
Via negativa
Improve things by removing what does harm before adding anything new, because we usually know more about what is wrong than about what is right.
Christian apophatic (negative) theology, which describes God only by what God is not; Nassim Taleb carried the term over to decisions and risk in Antifragile (2012).
Via negativa is Latin for "the negative way". Theologians who believed nothing true could be said directly about God listed what God is not and proceeded by elimination. Taleb applies the same move to action: the most reliable improvements are often subtractions. Stop smoking, drop a dependency, cancel a meeting, withdraw a drug that does harm, before reaching for a supplement, a framework or a new process.
It works because knowledge of what fails is sturdier than knowledge of what works. A claim that something helps can be overturned by the next study, while a well-documented harm rarely turns into a benefit. Taleb calls this "subtractive epistemology": the most robust way to add to knowledge is removing what we think is wrong. Removal also tends to have fewer side effects than addition, because anything you add interacts with everything already there in ways nobody has tested.
Some things have to be added: a vaccine, a test suite, a backup. And removing a structure you do not understand can be the mistake itself, which is the point of Chesterton's fence. The common misuse is treating "do less" as a blanket rule, or as a polite name for doing nothing or for cutting costs. It is strongest when the harm of the thing you remove is documented and its benefit is only assumed.
Example
For years doctors prescribed combined hormone therapy to postmenopausal women, partly to prevent heart disease. The Women's Health Initiative tested it in 16,608 women aged 50 to 79, and on 31 May 2002 its safety board stopped the estrogen plus progestin arm after an average of 5.2 years. Per 10,000 women each year, the treated group had 7 more coronary heart disease events, 8 more strokes, 8 more pulmonary embolisms and 8 more invasive breast cancers. Hormone use fell after the results came out, and in 2007 Peter Ravdin and colleagues reported that the age-adjusted breast cancer rate among US women dropped 6.7% in 2003 compared with 2002, almost entirely in women aged 50 and over. They described the link to stopping hormone therapy as a timing association, not proof.
Where it applies
Before adding a cache, a retry loop or another library to fix a flaky service, look for what to take out: dead feature flags, an unused dependency, a config option nobody sets. With a large language model (LLM) agent, try removing tools it rarely needs and prompt instructions that contradict each other before writing new ones.
Think about it
What could you stop doing this month that would help more than anything you are planning to start?
Applied it to something real? Log it.
Test yourselfWhy does Taleb argue that removing something is often a safer bet than adding something?
We know what harms more reliably than what helps, and an addition interacts with the existing system in untested ways, while removing a known harm has fewer hidden side effects.
Related models
- InversionWork a problem backwards by asking what would guarantee failure, then make sure you avoid it.
- Chesterton's fenceBefore you remove a rule, a process or a piece of code, find out why someone put it there.
- The Lindy effectFor things that do not wear out, such as ideas, books and technologies, every year they survive raises how long they can be expected to last.
- EntropyLeft alone, an isolated system drifts into its most probable, least ordered state, so keeping anything ordered costs continual work.
Sources and further reading
- Farnam Street, A Simple Heuristic to Recognize Charlatans (quotes Taleb's Antifragile on via negativa)
- Writing Group for the Women's Health Initiative Investigators, Risks and Benefits of Estrogen Plus Progestin in Healthy Postmenopausal Women, JAMA, 2002
- Ravdin and colleagues, The Decrease in Breast-Cancer Incidence in 2003 in the United States, New England Journal of Medicine, 19 April 2007